Datasets:

ai-aerospace

/

ams_data_full_2000-2020

like 0

188 role. Use of a 270-volt bus for launch vehicle applications is based partly on the belief that design and construction standards, once developed and validated, can be applied to manage this issue. The availability of approved high voltage, high current power electronics presents another challenge to high power ETVC systems for use in human rated launch vehicle applications. An effort by NASA to update requirements and its list of suitable parts for this application is underway. Addressing this concern and that of corona discharge susceptibility is a priority for proponents wishing to gain acceptance of powerful ETVC on large manned launch vehicles. The use of FPGA by avionics has already gained wide acceptance for manned launch vehicle applications. Within the common controller architecture FPGA, allow control parameters to be tuned to specific applications and the needs of specific actuator classes. When the testing presented in this report was performed, the developmental EMA actuator was the highest powered example to which the common controller had yet been configured. Actuator classes already covered by the common controller in previous developmental tests include: an electromechanical rotary engine control valve actuator, an electromechanical launch abort system valve actuator, and a less powerful EMA. Control of an EHA is planned as future work. The common controller leverages a modular architecture to increase its flexibility across dissimilar actuator platforms. As seen in the discussions of EMA and EHA mechanisms, actuator types differ not only in power requirements but also in instrumentation and control schemes. In a modular architecture, signal conditioning and instrumentation drive circuitry can be swapped out as required by a particular actuator type. Laboratory Demonstration of Integrated EMA System The integrated system test brought together all the elements of a single channel for one axis of a complete ETVC system. These system elements included the two-channel single fault tolerant developmental EMA, four high rate lithium ion battery modules and the cross platform extensible common controller. Testing was performed during August and September of 2011. It was during the first week of testing, when system integration and check out was performed, that it was necessary to modify control parameters for more optimal performance under test conditions. This modification satisfied the objective of demonstrating that control parameters were not merely factory preset constants but could be potentially changed in the field. Developmental hardware was used for these tests rather than the more flight-like battery module and controller shown in Figure 3 because they were not available at the time these tests were performed. As can be seen in Figure 7, four developmental battery modules, each containing many individually matched cells, were combined to form a single bus. Figure 7. Developmental Lithium Ion Ba ttery Supply (left) and Common Controller

Document

AMS_2012.pdf

189 Test Setup and Procedure The integration of the developmental EMA with the common controller and battery modules was done as illustrated by the interconnection diagram of Figure 8. In addition to the developmental hardware, this test setup also includes a Power Control and Regeneration circuit and the Universal Test Interface (UTI) needed to communicate with the common controller and translate telemetry into a form that could be recorded by laboratory data acquisition systems. Telemetry included commanded position, actuator position and force, motor current and motor velocity. The UTI and Regen circuit are Moog, Inc., Space and Defense Group proprietary hardware. The EMA, and Battery Modules used in this demonstration are NASA owned assets. The common controller is ATK and Moog jointly held proprietary hardware. Figure 8. EMA System Test Interconnect Diagram Battery cell temperatures at 38 separate locations were monitored by an Agilent model 34970A Data Acquisition / Switch Unit using Type-T thermocouples. Temperature logs were made during charging of the battery modules, as well as during testing of the integrated system. Logs would often cover several tests performed on a single day to track thermal data across consecutive runs. This also demonstrated the objective of repeated operation on a single battery charge. A voltage probe, visible in Figure 7, and a current probe applied to the Power Control and Regeneration circuit were used to verify that the integrated system was able to control peak power draw and voltage droop to acceptable levels. One of two large single axis inertial load simulators was used to provide representative inertial and spring loads for EMA system testing. These test stands were designed by Marshall Space Flight Center (MSFC) engineering and are located at the Marshall Thrust Vector Control Research, Development and Qualification Laboratory in high bay 110 of building 4205 at the NASA Marshall Space Flight Center.

Document

AMS_2012.pdf

190 Figure 9. Schematic of large inertial load simulator The developmental EMA was mounted between the two clevises indicated in Figure 9, as shown in the photograph of Figure 1. Each Inertial Load Simulator is instrumented to measure tension and compression forces applied by an actuator using a Honeywell model 3156-150K load cell that has a capacity of 667 kN (150,000 lbf). Pendulum position is measured to a system accuracy of 2 arc seconds using a Heidenhain RCN 729 absolute angular encoder. Figure 10. ETVC Demonstration Duty Cycle Two programmed duty cycles were created to exercise the actuators through a series of steps, ramps, discrete frequency sine waves and frequency sweeps. The ETVC Demonstration Duty Cycle shown in Figure 10 is somewhat aggressive. Lasting 111 seconds, this duty cycle created a relatively high demand on the power source and controller during use, and as such far exceeded the demands typically placed on TVC systems in a realistic launch scenario. The ETVC Demonstration Duty Cycle consisted of the following sequence: Large amplitude steps in the extend and retract directions; A single period of a 1 Hz sine wave at the same large amplitude; A stepwise return to null done in 5 discrete steps; A continuous sine-sweep from 1 Hz to 16 Hz, inclusively; A set of large amplitude ramps; A set of four discrete sine waves, chosen to excite a structural resonance within the test stand; Another large amplitude single period sine wave at 1 Hz. This duty cycle was used for both EMA and EHA testing.

Document

AMS_2012.pdf

191 Lessons Learned For EMA testing, it was necessary to first command the actuator away from null and then to displace the duty cycle accordingly. This was done because the type of LVDT used in this particular actuator created a control discontinuity as the actuator passed through null. The preferred solution would have been to change out the LVDT for one more optimally suited. However, schedule constraints did not allow for the exchange of this LVDT, so it was decided to simply avoid duty cycles that commanded the actuator through the null position. Figure 11. Discrete Frequency Duty Cycle The discrete frequency response duty cycle shown in Figure 11 was the other programmed duty cycle used. It was generated to assess the frequency response of the integrated EMA TVC system. This duty cycle consisted of a set of 17 sinusoidal waveforms covering the frequencies 0.5 Hz, 1.0 Hz, 2.0 Hz, …, 16 Hz. Unfortunately, a fixed number of 30 cycles was programmed for each frequency. The preferred method would have programmed a fixed duration for each frequency so that the number of data samples acquired would be equal. The duty cycle used resulted in a broadening of the individual spectral response curves with increasing frequency. Although this error was not identified at the time of testing, sufficient spectral data was present in the recorded command and response signals to perform the needed assessment. To each of the 17 individual sinusoidal waveforms of the discrete frequency duty cycle a Hanning window was applied, creating a smooth, gradual rise in amplitude to a maximum followed by an equally gradual reduction of amplitude. A wait time of 1 second was introduced between each waveform. This technique was used as a result of lessons learned during a previous modal assessment in which it was observed that whenever abruptly starting and stopping sinusoidal waveforms an impulse component was introduced with its associated broadband excitation. This technique was verified by Fourier analysis to be a very effective means of applying monochromatic (single discrete frequency) excitation. The recorded command signal and the actuator response, acquired via LVDT, were analyzed by applying a discrete Fourier transformation individually over time intervals corresponding to each successive frequency in the duty cycle. The highest four frequencies (13 Hz, 14 Hz, 15 Hz and 16 Hz) were discarded because the transform of the response did not meet the necessary criteria for inclusion. The coefficient of variation (⁄ߤߪ=ܸܥ representing the ratio of the standard deviation to the mean of a computed discrete Fourier transform, was used as this criterion. Only those responses for which the calculated CV was greater than 1 were selected.

Document

AMS_2012.pdf

192 Results and Conclusions All test objectives were accomplished without any unexpected outcome. Functional integration of all developmental ETVC components was performed. Controller parameters were adjusted for smoother operation and to account for differences due to the larger inertial load at the MSFC laboratory facility compared to that at Moog facilities. This need provided an opportunity to demonstrate the ability to tune the controller “in the field”. Peak power draw and voltage droop was demonstrated to be controlled to acceptable levels as was repeated operation on a single battery charge. The dynamic behavior of the EMA system was determined for step and frequency response. Figure 12 shows the response of the EMA actuator to a large commanded step in the extend direction under the test stand inertial load. This step occurred at the start of the aggressive ETVC Demonstration Duty Cycle. Response as recorded by the LVDT and the optical angular encoder, which provided a much cleaner signal, are plotted along with the recorded command signal. Figure 12. Step response of EMA actuator Gain and phase response spectra of the underpowered EMA are shown in Figure 13. Poor performance above 4 Hz was expected under the inertial load and power limitations imposed by the test. Frequencies above 12 Hz were discarded because the response did not meet the acceptance criteria for inclusion. This is equivalent to the magnitude of the response being below the effective noise floor of the data acquisition system. The highest battery cell temperature was recorded on 2 September 2011. The temperature log for that day, shown in Figure 14, captures a total of three EMA tests. A single thermocouple measured a single out of family temperature event at 2,010 seconds, elapsed from the start of the log, which corresponds to the end of the 2 nd EMA test performed that day. At the same time, the temperature nearest that value, recorded by another thermocouple, was compared. The difference is not significantly greater than the standard deviation measured across all temperature data at this same time. This means that the amount by which this single value is out of family is insignificant. It should be noted that even for this worst case example, the rise in temperature was not appreciable, the temperature returned to an in family value post test and, in no case, did any cell temperature ever approach a level of concern.

Document

AMS_2012.pdf

193 Figure 13. Gain and Phase Spectra of Underpowered EMA Figure 14. Lithium Ion Battery Temperature Log (worst case) References 1. Garrison, Michael and Scott Steffan. "Two Fault Tolerant Electric Actuation Systems for Space Applications." Association Aeronautique et Astronautique de France (AAAF) Changes in Aeronautical and Space Systems , Challenges for On-Board Energy Conference, Avignone, France (26 - 28 June 2006). 2. Garrison, Michael and Scott Steffan. "Two Fault Tolerant Electric Actuation Systems for Space Applications." AIAA-2006-4939, 42nd AIAA/ASME/SAE/ASEE Joint Propulsion Conference, Sacramento, CA. (9 - 12 July 2006). 3. McMahon, William A. "Apollo Experience Report – Guidance And Control Systems: CSM Service Propulsion System Gimbal Actuators." NASA TN D-1969 (July 1975). 4. Wayne, Duke. "OMS/RCS Subsystem Presentation." NASA GNC Presentation at NASA Kennedy Space Center, (4 June 1996).

Document

AMS_2012.pdf

194 5. White, Robert A. “High Voltage Design Criteria.” Marshall Standard MSFC-STD-531, George C. Marshall Space Flight Center (September 1978). 6. Dunbar, William G. “Corona Onset Voltage of Insulated and Bare Electrodes in Rarefied Air and Other Gases.” Air Force Aero-Propulsion Laboratory Technical Report AFAPL-TR-65-122, Air Force Systems Command, Wright-Patterson Air Force Base (June 1966). 7. United States Environmental Protection Agency. Hydrazine Hazard Summary-Created in April 1992; Revised in January 2000.

Document

AMS_2012.pdf

195 Development of High Temperature High Current Contact Technology in slipring Assemblies for the BepiColombo MPO & MTM Spacecrafts Fabrice Rottmeier, Mikaël Krummen* and Mickaël Miler* Abstract RUAG Space Nyon has been selected to design, develop and test the slipring assemblies of the Mercury Planetary Orbiter (MPO) and the Mercury Transfer Module (MTM) spacecrafts for ESA’s BepiColombo Scientific mission. The exposure of the spacecrafts to the harsh thermal environment of this mission to Mercury has inhibited the use of standard high current contact technology in the design of MPO and MTM cylindrical slipring assemblies. In order to sustain the particularly high thermal requirements resulting from the combination of the thermal environment with the significant thermal dissipation of high current transfer at high temperature, new contact and electrical transfer technologies were developed and tested. Validation tests were performed on breadboard models (BBMs) with flight representative contact system at temperatures ranging from -33°C to +186°C. This paper first presents the new developments integrated in the electric contact system design to meet the simulated thermal environment. It is then followed by a detailed presentation of the objectives and results of the validation tests. Lessons learned and the optimization of the design for the flight configuration are presented in the last sections of this article. Introduction Standard high current contact technologies used for space applications at RUAG consists of composite brushes soldered to a flexible copper beryllium blade. This flexible element provides a contact preload force that is designed to not only compensate the abrasive wear of the composite brush during its lifetime but also compensate for the preload loss due to the inherent creep of this material. The current is directly transferred from the composite brushes through the flexible blade. The composite brushes slide on hard gold-coated cylindrical tracks. For the BepiColombo slipring assemblies, the high temperature environment inhibits the use of standard solder which has been replaced by a high temperature gluing compound. Consequently, current from the composite brush is not transferred through the blade, but through copper braids directly sintered into the composite brushes. This standard industrial process is for the first time used in a space application. Additionally, composite brushes material has been adapted to the high-temperature environment while considering the need for reduced friction and low electrical resistance of the sliding contacts in vacuum conditions. Since this material differs from standard ones used by RUAG for space applications, validation tests have been performed to validate the new contact technology and the composite brush assembly for space applications. Technology overview and main features For the BepiColombo slipring assemblies, the high temperature environment inhibits the use of standard solder which has been replaced by a high temperature gluing compound. While providing a fixation system compliant with the local thermal environment, this new gluing system is relatively resistive, such that the electrical conduction through the flexible blade is not effective. In order to re-establish a low * RUAG Switzerland Ltd, RUAG Space, Nyon, Switzerland Proceedings of the 41st Aerospace Mechanisms Symposium, Jet Propulsion Laboratory, May 16-18, 2012

Document

AMS_2012.pdf

196 resistive electrical path between the composite brushes to the brush holder’s electrical distribution system, ultra-flexible copper braids (two per composite brush) are directly sintered into the composite brushes. This standard industrial process is used for the first time in a space application. Sintering is directly performed by the composite brush manufacturer. This process implies very high temperatures, thus preventing the use of standard space cables with polymer insulation material. In order to minimize the parasitic stiffness of the cable assembly in the composite brush suspension system, ultra flexible copper braids are used. The reduced stiffness of these cables is obtained by using small diameter strands (Ø0.05 mm) specially woven in 3 braids of 43 strands. As for all space applications, copper braids have to be silver plated. While standard space-grade cables commonly have a silver thickness of 2 μm, it was not possible to procure ultra-flexible braids in standard space grade with a silver coating thicker than 1 μm for this special type of braid. At the other extremity of the flexible braids, standard soldered electrical connections could not be implemented due to the high local thermal environment. This connection has therefore been replaced by a terminal lug with the pair of flexible braids crimped to it. In order to electrically insulate the flexible braids, PTFE heat-shrinkable insulating sleeves have been used. This material has been selected for two main purposes. While PTFE is fully compatible with the local thermal environment, it also has the advantage of shrinking at higher temperatures than the maximum predicted local temperature. This permits a control of the forming of the shrink sleeves in the production process while guaranteeing no further shrinkage during the operational life of this component. The control of the dimensions of the insulating sleeves has been considered important based on the observation of the ultra-flexible braid under the microscope. This particular braid construction has a very loose braid structure to minimize strand contacts during flexion. Therefore, a too tight sleeving will augment the stiffness of the insulated braids, affecting the stiffness of the flexible blade system. The diameter and thickness of the shrunk sleeves also plays a role in the stiffness of the system. The final definition of the PTFE sleeves for the ultra-flexible braids has been based on numerous stiffness tests in order to minimize the parasitic stiffness brought on by this electrical connection to the composite brush. A 3D CAD view of the composite brush assembly is provided in the left side of Figure 1. A picture of a completed composite brush assembly for the production of the flight units is also shown on the right side of Figure 1. Figure 1. Composite brush assembly (CAD and final flight configuration) Each composite brush assembly is designed to transfer 8.25 A in the worst case thermal configuration of the mission. The material of the composite brushes has been adapted to this high-temperature environment while considering the need for reduced friction of the sliding contacts in vacuum conditions.

Document

AMS_2012.pdf

197 The selected composite brush material basically consists of high silver content composite brushes with MoS 2 lubricant. Since this material differs from standard ones used by RUAG for space applications, validation tests have been performed to validate the new contact technology for space applications. Validation Test Program In order to validate the high temperature contact technology selected for the BepiColombo slipring Assemblies, the following validation test program has been established. The main objective of this validation test program was to verify that the selected composite brushes technology is adequate for the BepiColombo mission. Validation tests have been performed at component level as well as at system level on breadboard models fully representative of the newly developed contact system. Component-level validation testing aimed specifically at the validation of the new gluing system, the validation of the sintered connection between the flexible braid and the composite brushes as well as the validation of the electrical performances of the new composite brush assembly. Component-level validation tests specifically consisted of: o Shear tests on the glued interface o Microsections of the glued assembly o Pull test of flexible braid sintered into composite brush o Verification of electrical continuity between composite brush and blade o Anthony-Brown test on silver coated flexible braids (red plague risk assessment) In addition, the inability to procure ultra-flexible braids in full compliance with European Cooperation for Space Standardization (ECSS) requirements on silver coating thickness of copper wires (1 µm coating instead of 2 µm) has necessitated additional testing for the assessment on red plague contamination control. Performance of the composite brushes in terms of electrical dynamic resistance, friction torque and wear has been characterized on breadboard models and compared to the initial design values. The validation tests have been conducted in parallel with the development of both MTM and MPO slipring assemblies for programmatic reasons. Both breadboard models have been used to validate the overall slipring assembly concepts with main focus on the new contact technology. These have been tested with respect to the various representative mission environments, which consisted of random vibrations, shocks, thermal vacuum cycling and accelerated life tests with intermediate electrical and mechanical functional tests. Component-level validation tests Mechanical validation tests of the new gluing system of the composite brushes mainly consisted of post thermal-cycling shear tests of composite brushes samples as well as analyzing microsections views of the composite brushes after performing the validation test campaigns. These two tests aimed specifically at the verification of the mechanical properties of the gluing system after aging. These tests are described hereafter, along with other components tests. Shear test on glue interface Shear tests were performed after 100 thermal cycles on numerous samples of the composite brush/flexible blade assembly by Aerospace & Advanced Composites GmbH - AAC. Table 1 summarizes the shear test results performed on two series of samples and reference samples (without thermal cycling). These results have demonstrated that the aging through thermal cycling has not impacted the structural integrity of the adhesive system that is fully compliant to the design requirements.

Document

AMS_2012.pdf

198 Moreover, the margin of safety with respect to shear force is significant, as in the real application, the composite brush experiences a shear force approximately 1000x smaller than the minimum ultimate shear force resistance when the slipring’s motion is initiated. Table 1. Shear test results summary Series 1 Shear Force Std deviation Thermocycled Samples (8 samples) 442 N 14 N Reference Samples (2 samples) 448 N 4 N Series 2 Thermocycled Samples (8 samples) 427 N 20 N Reference Samples (2 samples) 448 N 20 N Figure 2 provides pictures of shear test samples after shear testing to illustrate the fracture mode of the composite brush glued assembly. The same fracture mode has been observed on all test samples. Figure 2. Shear test samples Figure 3 illustrates the shear fracture propagation mode. A crack initiates on the side of the composite brush where it is held in the shear test fixture, and then propagates up to the glue interface, underneath the blade. The glue interface surface then breaks very quickly as the contact surface between the blade and composite brushes reduces rapidly. The initiation of the crack in the composite brush suggests that this component has a shear-stress limit lower than the glued interface. Glue layer microsections Examples of transverse micro-sections views of the composite-brush/flexible blade assembly made after life tests are presented in Figure 4. The Copper Beryllium blade section is shown on the top-half of the micro-section views (uniform color) and the composite brush sections on the lower half (marbled section). All micro-section views have shown that the integrity of the gluing system of the composite brushes is not compromised by both the thermal cycling and life tests performed on the breadboard models.

Document

AMS_2012.pdf

199 Figure 3. Shear test setup and fracture mode description Figure 4. Example of composite brush micro-section views ( © Laboratoire Dubois SA ) Pull test on composite brushes sintered cables Flexible braids are directly sintered into the composite brush as part of the composite brush manufacturing process. In order to validate the proper mechanical fixation of the braids, pull tests have been performed by RUAG.

Document

AMS_2012.pdf

200 To assess the potential impact of thermal vacuum cycling on the mechanical fixation of the sintered braids, composite brushes from both breadboard models have been tested and the results compared with the ones obtained on the QM/FM composite brushes fabrication batch. The flexible braid pull test results are summarized in Table 2. Table 2. Pull test results summary Pull Force Std deviation composite brushes from BBMs (thermal vacuum cycled) 58.3 N 4.8 N composite brushes from QM/FM production lot (not cycled) 52.5 N 2.8 N No major impact of the thermal cycling on the resistance of the sintered connection can be observed from the pull test results. In fact, in most cases the rupture mainly occurred on the flexible braid rather than on the sintered joint. This is illustrated in Figure 5 on a composite brush assembly that encountered both fracture modes during the flexible braid pull tests. Figure 5. Pull test results on composite brushes sintered cables with both rupture modes The maximum design loads on the flexible braids are generated by the vibration and shock environment and principally due to the inertial loads on the flexible braids. These loads are marginal with respect to the ultimate traction load. These pull tests have proven the suitability of the sintered connection between the flexible braids and the composite brush for the BepiColombo slipring assembly design. Electrical grounding between composite brush and blade In this new composite brush Assembly the spring blades are not part of the electrical line anymore. Since the flexible blades are connected to the composite brush using electrically insulating glue at one end, and supported by a brush holder made of insulating material at the other end, the electrical grounding of the blades needed to be insured by tests or by the use of an additional grounding lug. Continuity tests have been performed on the composite brushes assembly of both breadboard models after the validation tests. The tests were fully successful and proved that the electrical continuity was guaranteed between all blades and composite brushes, despite the use of electrically insulating glue. Using a purely theoretical approach on the insulating properties of the thin layer of glue used, the spring blades should have been electrically insulated. However, the electrical grounding tests have enlightened

Document

AMS_2012.pdf

201 that the assumption of a uniform barrier of glue between the composite brush and the flexible blade is not realistic. In fact, a good electrical conduction between the blade and the composite brush is ensured by discrete electrical contact points arising from the high surface roughness of the composite brush in combination with the gluing process performed under preload. In order to guarantee the electrical grounding of the flexible blades during QM/FM production, an electrical continuity check is performed on every composite brush assembly. Anthony-Brown tests on silver plated flexible braid The applicable requirements from the ECSS call for a 2- μm silver plating thickness on copper conductors. While standard space grade cables commonly have a silver thickness of 2 μm, it was not possible to procure ultra-flexible braids in standard Space grade with a silver coating thicker than 1 μm for this special type of braid. “Red Plague” is a well-known phenomenon that causes corrosion of copper on silver plated copper cables. The risk of occurrence is increased in the presence of humidity, particularly in condensed form. With the selected composite brush assembly, this is typically the case considering the thermally insulating sleeving of the flexible braids in humid atmosphere and submitted to temperature variations. Samples of the flexible braid production batch have been equipped with the PTFE shrink-sleeves and provided to ESA’s Test Center (ESTEC). Anthony-Brown tests are currently being performed for the assessment on red plague contamination control. Breadboard Model Validation Tests The new composite brush sliding contact system has been validated at system level on two breadboard models (MTM and MPO slipring assemblies). Figure 6 provides a picture of one of these breadboard models for illustration purposes. Both models shared the same composite brush contact system and were fully representative of the flight hardware in terms of materials, contact configuration and dimensions, with the exception of the number of tracks that was reduced in the breadboard models. The main implic ation is that the breadboard models have a reduced overall length with respect to the QM/FM design.

Document

AMS_2012.pdf

202 Figure 6. BepiColombo BBM 1 (configuration with contacts not fully populated) The validation test sequence for both breadboard models is schematized in Figure 7. Breadboard model validation tests mainly consisted of mechanical vibration and shock tests, thermal cycling and life tests with numerous electrical functional tests to demonstrate the adequacy of the contact technology to this high temperature application. Figure 7. Breadboard models validation sequence In addition, the preload force drop due to the combination of the wear of the composite brushes and flexible blade material creep has been measured and assessed. These measurements were then used to compute the friction coefficient of the contact system from torque measurements performed both in air and in vacuum as well as with and without brush holders in order to remove the contribution of the bearings on the overall measured torque. The main results of these BBM validation tests program are described in the next subparagraphs. Thermal vacuum cycling, mechanical and accelerated lifetime tests BBMs have been tested under representative mechanical, thermal and accelerated lifetime (number of cycles including qualification margin) of the BepiColombo mission. During the lifetime testing, monitoring of the dynamic resistance and friction torque was performed, in order to measure the evolution of these critical parameters. Figure 8 and Figure 9 present the static and dynamic RMS contact resistance evolution during lifetests (measured with a current of 1A). Both graphs show a slight increase with the number of revolutions, which reflects the impact of the generated wear particles on the sliding contact electrical characteristics. However, these resistance values remain very small and well within the specification.

Document

AMS_2012.pdf

203 Figure 8. Static contact resistance evolution – RMS value Figure 9. Dynamic resistance – RMS value The design of the composite brush preloading system is inherently critical with respect to the mechanical environment, mainly due to the configuration of this system that has a concentrated mass at the end of a spring blade. The weight of the composite brush design has been minimized and the preloading system designed to avoid eigenfrequencies in the vibration frequency spectrum. Moreover, the mechanical dimensioning of the preloading system is such that lift-off of the composite brush is prevented as well. Vibration and shock tests have been performed on both breadboard models to validate the design of the composite brush suspension system and demonstrate that the composite brushes and tracks were not affected by these mechanical environments. An illustration of the contact system after vibration and shock tests on a breadboard model is shown in Figure 10. Slight traces were visible on all tracks near the composite brushes. After close inspection, it 05101520 Before vibrationsAfter vibrations0 revolution 14000 revolutions28000 revolutions42000 revolutions56000 revolutions70000 revolutionsContact static resistance [mOhm]in AIR in VACUUM 012345 Before vibrationsAfter vibrations0 revolution (1rpm)14000 revolutions (0.007rpm)28000 revolutions (1rpm)42000 revolutions (1rpm)56000 revolutions (1rpm)70000 revolutions (0.007rpm)Dynamic resistance - RMS measured with 1A [mV/A] in AIR in VACUUM

Document

AMS_2012.pdf

204 turns out that these marks were coming entirely from the local composite brush wear during vibration and shock tests. As a matter of fact, the rotation of the rotor was not rigidly blocked such that a reduced back and forth sliding movement was enabled. This local sliding created deposition of the composite brush wear particles on the tracks. These residues could easily be wiped off and a visual inspection of the gold-coated tracks under magnification did not reveal any damages or wear of the track surface. It should be noted that the dimensioning of the composite brush preloading system is such to avoid any lifting of the contacts under the applicable vibration and shock environment. The observed agglomerates of wear particles are typical of composite brush contact slipring designs and have already been previously observed by RUAG. These traces do not affect the performances of the contact system. Figure 10. Wear traces due to composite brush sliding effect during vibration In terms of accelerated lifetime testing, one of the breadboard models was subjected to more than 70’000 revolutions, which corresponds to more than seven times the specified number of revolutions for the BepiColombo mission. Thorough visual inspections of the tracks and composite brushes after life test and under magnification have shown that the overall conditions of the tracks and composite brushes were fully satisfactory. As a matter of fact, very limited amount of wear particles were found, and slight wear traces were visible on the tracks. The analysis of the total wear of the composite brushes demonstrated that the wear of the composite brushes was extremely limited, despite the number of revolutions experienced on top of the design life of the contact system. Moreover, the wear was observed to be uniform on all the composite brushes. It should be noted that the design life of composite brushes in industrial applications is at least more than one order of magnitude higher than the one tested in this application. Finally, the complete visual inspection of the breadboard model slipring assembly showed that it was still in a very good condition. Friction coefficient The friction coefficient of the electrical contact system plays an important role in the friction torque of a slipring assembly, mainly due the high number of contacts, and is, of course, an important parameter for the design of the drive assembly. In order to validate the friction coefficient used for the design of the BepiColombo slipring assemblies, torque measurements were performed in air and in vacuum as well as with and without brush holders in order to reject the torque contribution from the bearings. Measurements of the preload force of the composite brushes allowed the actual friction coefficient of the electrical contacts in air and vacuum to be calculated. Figure 11 summarizes the friction coefficient measurement results. These data, besides enabling the validation of the total torque calculations for the slipring assemblies, have also shown that measuring the torque in air can be considered as worst case conditions for the definition of the qualification and acceptance test programs. These results furthermore confirm the adequacy of the selected composite brushes technology with respect to space applications.

Document

AMS_2012.pdf

205 Figure 11. Friction coefficient measurement results Dynamic contact resistance Examples of the dynamic electrical contact resistances measured over a full revolution and after run-in of the composite brushes are presented in Figure 12. These measurements were performed at the nominal current of 12.5 A per track on four different tracks in air and vacuum. All results have shown a low electrical contact resistance in both vacuum and air conditions with great consistency between the different measured tracks. Moreover, the performances of the composite brushes are significantly better in vacuum environment, which, again, confirms the adequate selection of the composite brush material for space applications. The RMS values of the dynamic electrical contact resistances are of particular interest, as contact resistance is a key driver for the thermal dissipation at the contact interfaces due to the electrical power transfer. This is particularly critical considering the thermal environment of the BepiColombo mission. Figure 13 summarizes the RMS values of the dynamic contact resistance curves presented in Figure 12, with the addition of measurements performed before run-in of the composite brushes. These results show that with the selected composite brushes technology, RMS values of the contact resistance are very low, especially under vacuum. Composite brushes run-in also shows its benefits, with an overall reduction of the dynamic contact resistances. These results show that composite brush technology becomes optimal under vacuum and is furthermore improved by the effect of the run-in, which correspond to the flight configuration. in vacuum in air0.30 0.25 0.20 0.15 0.10 0.050.00friction coefficient

Document

AMS_2012.pdf

206 Figure 12. Dynamic contact resistance curves over 360°– VACUUM & AMBIENT (after run-in) Figure 13. RMS dynamic contact resistance – VACUUM & AMBIENT (after run-in) Lessons learned and improvements Validation test programs performed at component and breadboard model levels have shown that the selected composite brushes contact technology is fully compliant with the BepiColombo requirements. In view of QM and FM slipring assembly production, some improvements are being carried out, namely:  For the breadboard models, the composite brushes were manufactured as closely as possible to the cylindrical shape of the track, with the objective of performing a run-in directly on the assembled slipring assemblies. However, breadboard models tests have showed that wear is extremely low, and that a significant number of revolutions were required to obtain a perfect fit of the composite brushes to the tracks. Figure 14 illustrates the composite brushes contact surface before and after run-in tests performed on the breadboard models. Track A 0 mΩ5 mΩ10 mΩ15 mΩ20 mΩ 0 ° 30 ° 60 ° 90 ° 120 ° 150 ° 180 ° 210 ° 240 ° 270 ° 300 ° 330 ° 360 ° angular positioncontact resistancevacuum ambientTrack B 0 mΩ5 mΩ10 mΩ15 mΩ20 mΩ 0 ° 30 ° 60 ° 90 ° 120 ° 150 ° 180 ° 210 ° 240 ° 270 ° 300 ° 330 ° 360 ° angular positioncontact resistancevac uum ambient Track C 0 mΩ5 mΩ10 mΩ15 mΩ20 mΩ 0 ° 30 ° 60 ° 90 ° 120 ° 150 ° 180 ° 210 ° 240 ° 270 ° 300 ° 330 ° 360 ° angular positioncontact resistancevac uum ambientTrack D 0 mΩ5 mΩ10 mΩ15 mΩ20 mΩ 0 ° 30 ° 60 ° 90 ° 120 ° 150 ° 180 ° 210 ° 240 ° 270 ° 300 ° 330 ° 360 ° angular positioncontact resistancevacuum ambient BepiColombo - Carbon brush technology RMS Contact resistance - AMBIENT 0 mΩ10 mΩ20 mΩ 12.5 A 12.5 A 12.5 A 12.5 A track A track B track C track Drms valueambient before run-in ambient after run-in BepiColombo - Carbon brush technology RMS Contact resistance - VACUUM 0 mΩ10 mΩ20 mΩ 12.5 A 12.5 A 12.5 A 12.5 A track A track B track C track Drms valuevacuum before run-in vacuum after run-in

Document

AMS_2012.pdf

207 In order to accelerate the run-in process for the QM and FM production lines, the brush holder assemblies are run-in on a dedicated abrasive tooling representative of the final track dimension. In order to further accelerate the run-in process, wear particles are removed by a continuous flow of demineralized water during the run-in procedure. Figure 14 – Overview of composite brushes before and after run-in  After thermal tests, a slight brownish coloration has been observed on the Kynar sleeves covering the ultra-flexible braids of the composite brushes. Initially translucent, with a pale yellowish coloration, they have slightly darken most probably due to the harsh thermal environment of the experienced thermal test (see images on the left of Figure 15), which were close to the Kynar temperature limits. Additional tests performed on the Kynar sleeves of the breadboard models did not reveal any degradation of the mechanical and electrical insulation properties. However, for the QM/FM design, it has been decided to switch to a PTFE-type heat shrink sleeve, which could sustain higher temperatures and avoid further shrinkage during the operational life of the slipring assemblies. Additional thermal tests have shown that this shrink sleeve was not affected by the exposition to the thermal environment defined for the BepiColombo mission (see images on the right of Figure 15).  In the initial design of the QM/FM slipring assemblies, the flexible blades were grounded using an additional cable fixed to the opposite end of the composite brush using a terminal lug. However, continuity tests performed between the composite brushes and the flexible blades have all shown that the grounding of the flexible blades is ensured through the glued interface between the composite brushes and the suspension blades. The grounding cables have been removed in the QM/FM contact system design and replaced by a continuity test performed on each composite brush assembly during their production.

Document

AMS_2012.pdf

208 Figure 15. Change of heat shrink sleeve type (left: old, right: new)  The gluing process carried out on the composite brushes flexible blade assembly during breadboard model production has shown that it requires careful attention. In order to enhance the QM/FM gluing process, the amount of glue and the location of its application have been defined and documented. Additionally, acceptance criteria have also been defined to guarantee a proper distribution of the glue, a precise alignment of the composite brush and in general a full repeatability of this process. Figure 16 illustrates the final configuration of the QM/FM composite brush as part of the acceptance criteria. Figure 16. Close-up view of glued interfaces

Document

AMS_2012.pdf

209 Conclusions Validation tests results have demonstrated the full adequacy of the high temperature high current contact technology implemented in the BepiColombo slipring assemblies. Moreover, the electrical dynamic resistance which has an important impact on the local thermal dissipation and consequently on the temperature distribution, appeared to be much lower than anticipated, validating the thermal simulation results with a comfortable margin. Additionally, the demonstrated low friction and low wear properties of the composite brush technology are well above the required lifetime for the mission. Moreover, the selected material for the composite brush is well-adapted to space applications, as it behaves better in vacuum than in air. Both validation campaigns (at component and at breadboard model levels), have helped in identifying important improvement points and lessons-learned. The production processes for the QM and FM manufacturing have been consequently enhanced. More specifically, the gluing and run-in processes, which are specific processes to the current application, have been updated and thoroughly detailed following these validation test campaigns. Finally, it has been demonstrated that the selected composite brush assembly, including blade gluing, sintered flexible braids and heat shrink sleeves fulfills all mechanical, electrical and lifetime requirements of the BepiColombo mission. References 1. Heinrich B, Zemann J, Rottmeier F. "Development of the BepiColombo MPO Solar Array Drive Assembly." Proceedings of the 14 th European Space Mechanisms & Tribology Symposium, September 2011.

Document

AMS_2012.pdf

211 Lessons Learned to Avoid Coax Cable Failure in Moving Mechanical Mechanisms Sheah Pirnack* Abstract Several programs have experienced anomalies due to cold welding of coaxial cables within Moving Mechanical Assemblies. Cold welding occurs when similar clean adjacent metal surfaces (like silver on silver or gold on gold) molecularly bond to one another given sufficient cleanliness, time, and contact pressure; as the name implies, it can and will occur in the absence of heat. This paper addresses the effects of cold welding in coaxial cables, outlines specific test results, suggests how to test for cold welding, and how to avoid the problem of cold welding in the design and material selection phase of a program. Introduction Within moving mechanical assemblies (MMAs), where any item is expected to move relative to another, cold welding can result in failure or severe performance degradation. Cold welds create a constraint where either the cold weld is strong enough such that the two surfaces are unable to move relative to each other or where movement / strain on the surfaces will cause early fatigue failure (such as local yielding, buckling or other low cycle fatigue damage) when the cold welded joint is flexed. In a coaxial (or coax) cable, the condition of concern begins when cold welds develop a local constraint within the multiple layers of the cable; then through subsequent motion the induced strain initiates low cycle fatigue failure in the proximity of the cold welds including tensile overload and buckling. Additional motion has the potential to result in severe damage to the conductors intended to carry the signal/data. Not every test specimen showed cold welding between the silver-coated copper layers common to coax cable design and cold welding does not always lead to a damage level that affects Radio Frequency (RF) performance. Within this paper the construction of coax cable will be introduced, photos of where cold welding has been found will be shared, and contributing factors for cold welding will be discussed. Additional comments will be given on specific application susceptibility. Figure 1 has photos of two different cables whose failure was initiated by cold welding. Figure 1. Failed Coax Cable: failure initiated by cold welding * Lockheed Martin Space Systems, Denver, CO Proceedings of the 41st Aerospace Mechanisms Symposium, Jet Propulsion Laboratory, May 16-18, 2012

Document

AMS_2012.pdf

212 Background Within space flight Moving Mechanical Assemblies (MMA’s) the cabling design, construction, and management should be an integral part of the design process: the cable should be chosen for its RF/data/power performance, space-compatible materials, flexing capability, and the cable management system should consider the cables’ capability for bending through life without fatigue or instability compromising the functional integrity of the cable. When a coaxial cable is part of a MMA, specific features of the cable design as well as the mechanical features that manage the flexing and containment of the cable must be carefully considered. Insufficient attention to these mechanical design details can cause cold welding and/or subsequent fatigue failures in space hardware. Coaxial cables are used as transmission lines for radio frequency signals; their applications include feed lines connecting transmitters and receivers with their antennas (data transmission / receipt). In coax cables the signal travels between the center conductor and the surface of the outer conductor separated by the dielectric. A typical flexible RF coax cable construction is shown and annotated in Figure 2. Figure 2. Construction of a typical Flexible Coax Cable (Design 1) Within a MMA application a coax cable must maintain its insertion loss / frequency performance throughout flexing and life. The construction details are critical to a flexible coax cable’s dynamic function. To maintain a high level of performance, the separation between the conductors has to be relatively constant through flexing/bending, the dynamic system (driving the motion) must be able to bend the cable, and the conductor’s have to maintain electrical continuity when flexed (no gaps). For high performance flexible cable the jacket and dielectric are plastic (low modulus/bend easily) and the metal conductive layers are multi-layer components themselves: the woven braided shield, the spirally wrapped outer-conductor foil wrap, and the stranded center conductor. The assembly is held together with the protective jacket sleeve to both protect the cable and ensure no radial gaps are generated at the outer conductor (maintaining cable insertion loss stability). The majority of aerospace coax cable designs’ incorporates silver-coated copper center conductor strands, outer conductor foil, and braid (Ag coated Cu). Recall that cold welding occurs when similar clean adjacent metal surfaces (like silver on silver or gold on gold) molecularly bond to one another given sufficient cleanliness, time, and pressure. Within the typical construction for flexible coax cable there are multiple opportunities for the various layers to cold weld to themselves or each other because of their similar material construction.

Document

AMS_2012.pdf

213 Results of Testing Cold Welds Observed in Cable In a series of cyclic cable tests performed in vacuum, post-test destructive physical analysis (DPA) revealed evidence of cold welding in multiple specimens and multiple coax cable designs [Acknowledgments 1c]. Both silver to silver and copper to copper cold welds were observed. Cold welding was found between metal to metal surface layers within various coax cables - see Figure 3 and Figure 5 for evidence of cold welding within the cable constituents. Cold welding in the center conductor can be identified by pulling individual strands apart; the evidence of cold welding is one or more strands are attached to the strand that is being separated from the group. As can be seen in the photo in the upper right hand side of Figure 3 what appears to be two strands bonded together is actually two center conductor strands cold welded along the whole length of the center conductor. Cold welding in a stranded center conductor reduces fatigue life properties of this critical component effectively acting as a deeper beam section that must conform to the same cable bending radius, producing a higher stress state than a single strand would [Acknowledgments 1b]. In multiple cable cyclic test specimens fretting, plowing, and material transfer due to the outer conductor foil layers cold welding and sliding relative to each other was observed [Acknowledgments 1c]. Not all cold welding affected the cable performance. When the foil was unwound from itself during DPA, intact cold welds were discovered in several specimens along the spirally wound length of the cable. The extent of the cold welding was dependent on the location along the cable, the type of cable, and the relative bending profile of that section of cable. An example of cold welding is pointed out in the lower right hand picture in Figure 3. Recall that an RF signals travel along the surface of the outer conductor. This type of wear and material transfer will not affect the cable’s function if the outer conductor layer remains intact and continuous. Figure 3. Cold Welds and effects of Cold Welds observed on all metal layers of coax cable

Document

AMS_2012.pdf

214 The photo in the lower left hand frame of Figure 3 reveals braid distortion along the length of the cable. The geometry and path of the seven (7) strand groups of woven braid are not uniform in their form; some strands are pulled away, or bent in an irregular shape, or, in an extreme case, an entire group of strands under local stress tensile yielded. The braid distortion is indicative of the braid strands sticking to each other or to the outer conductor foil layer underneath. With movement, the cold welding distorts or pulls the strands away from the woven group of strands that they originally followed. An X-ray of untested cable from the same lot cable is shown in Figure 4 (without distortion). An X-ray can be completed after test without DPA to look for evidence of cold welding in a cable by discerning if there is braid or conductor distortion. Figure 4. CT Scan / X-Ray Image of undistorted cable (braid visible) Figure 5 is a photo of the same section of braid distorted cable within Figure 3 with sections of braid peeled away to reveal the cold welds and outer conductor damage/stretching and tearing. Evidence of braid to outer conductor foil cold welding is clear in Figure 5. Figure 5. Image of Cold Welds observed between the braid and outer conductor (foil wrap) during DPA (some areas purposely cut away to make observation) [1c] Recall that RF signals travel between the center conductor and the inner surface of the outer conductor layer with a precise separation set by the dielectric within coax cable. In the construction outlined above, the outer conductor foil is the main functional layer along with the center conductor. If either of these functional layers is damaged, a corresponding affect in insertion loss within the cable is realized: the

Document

AMS_2012.pdf

215 extent of insertion loss increase is dependent on the operational frequency range and the level of damage. Susceptibility of Cold Welds in Cable: Choosing a flexible cable design Not every cable test sample showed cold welding between the silver coated copper layers, and not all cold welding led to damage that affected the RF performance of the cable. So what is the potential for damage due to the layers cold welding? And what kind of cold welding is most damaging? How can it be adequately screened for? And if a cable management system was successfully qualified and life tested already, is that enough, regardless of seemingly small changes between cable production lots? Multiple coaxial cables of different designs were tested in various conditions to determine the impact to the performance of the cable. Development testing, fatigue testing, life testing, X-ray inspection, material analysis and various other techniques were used to characterize the cable; RF performance parameters were used to monitor test units during testing. Different manufacturers and types of cables were tested with a construction outlined in Figure 2- will be referred to as Design 1. Some cables with an additional layer between the outer conductor spiral wrap and the braided shield were also tested and designs with this feature will be referred to as Design 2. A schematic representation of a Design 2 cross section is shown in Figure 6. Figure 6. Design 2 of cable with addition of an intermediate layer Three specific tests are summarized in Table 1: each test was conducted in an equivalent strain cycle and strain rate profile that encompasses appropriate life test margins in a vacuum chamber. Design 1 of cable went through multiple tests but one lot failed life testing (Lot C) and one lot successfully completed life testing without degradation (Lot A). DPA of the cables revealed very subtle differences in material hardness, contamination levels, jacket pressure, and overall manufactured process / tightness but significant differences in cold welding and damage. A key lesson learned from the results of the testing is a recommendation to complete vacuum life testing on every material lot of cable to ensure that manufactured lot differences won’t have an impact on cable life performance. Subtle changes in contamination, material lot and manufacturing processes can impact the results of life testing even with the same design of cable.

Document

AMS_2012.pdf

216 Table 1 Design 1: Lot A Design 1: Lot C Design 2 Baseline Construction / design 1: Lot A initial build & lot of materials, no performance affecting damage observed Construction / Design 1 same as Lot A, different build date and lot of materials; Cold Welding and damage observed in multiple layers along length of cable Design 2 incorporated intermediate layer, multiple Design 2 cables tested, Sporadic Cold Welding between Outer Conductor layer observed with minimal damage. Life Tests Successful (2 of 2 cables passed) Life Tests Failed (2 of 2 cables failed) Life Test Successful (6 of 6 cables passed) Design 2 cables went through multiple tests, multiple permutations of profiles, and various other stress and fatigue tests in addition to the same level of testing of Design 1. Design 2 cable outperformed Design 1 cable. Two key features differentiate Design 2 Cable from the Design 1 cable: the intermediate non-silver coated copper layer between the outer conductor and the braided shield; and alloyed center conductor strands with a thicker outer silver coating. The failed Design 1 cable incorporated relatively pure copper strands coated with a thinner layer of silver within the center conductor strands. In regard to cold welding, the intermediate layer prevents cold welding between the outer conductor and the braid by inserting a different material between the silver coated copper layers. This intermediate layer feature is believed to be the key difference preventing the most damaging occurrences of cold welds : cold welds that form between two separate layers in the outer portion of the cable. When the cable is put in bending, the outer layers are forced to move/slide relative to each other (more than the inner layers); if cold welds are present, instead of moving relative to each other by sliding, local yielding or pushing of surface materials occurs. Referring back to Figure 5, it is evident that braid to outer conductor cold welds initiated tearing and yielding in both layers through motion / bending of the cable. The alloyed center conductor mitigates some level of the similar metal interaction required for cold welding, has better metal fatigue properties, and with a thicker layer of silver on each strand the silver is likely to transfer between strands acting as a lube or additional wear layer through plowing and compliance. To mitigate the effects of cold welding in coaxial cables, the design and material of the cable for flexing applications should incorporate the two features mentioned: 1. Incorporate an intermediate layer between the outer conductor and braid of a different preferably non-metallic material. It will act as both a wear buffer and will prevent cold welding between the two outer most metallic layers of the cable. It also has the potential to mitigate insertion loss instability within cables (where the outer conductor loosens up over time developing a radial gap between adjacent wraps). 2. Ensure the center conductor is a higher strength alloyed copper and that the silver coating on the strands has a well controlled thickness (thicker likely is better based on the test observations). Because of the many variables involved, a life test must be performed for each new design of cable even with the recommended features incorporated. In addition to design verification, an important lesson-learned was that coax cable construction and implementation can vary in subtle ways between production lots that can affect cold welding and fatigue capability, and therefore a life test should be performed for each production lot regardless of design similarity.

Document

AMS_2012.pdf

217 Monitoring RF Parameters for indicators to screen for a problem during Coax Cable Vacuum Testing During vacuum testing of the Design 1 lot C cable RF data (TDR, IL, & VSWR monitored continuously) indicated that something was happening almost immediately after a test pause due to a chamber issue. Up until that point there was no indication of a cable performance change. Similar pauses occurred with the other cable lots/designs but the leading indicator (RF performance) was unaffected. Soon after the restart of the test, the degradation of the Design 1 lot C cable was rapid. The initial indicator was a time-domain reflectometer (TDR) test where a discontinuity / spike identified damage in the bending portion of a cable. Through additional cycling an insertion loss “suck-out” or increase started to form in the highest monitored frequency range and moved towards the lower frequency ranges with additional cycles (RF performance monitored from 100 MHz – 20 GHz). The test was stopped to determine the cause of RF degradation. The RF performance was cable motion and position dependent indicating opening and closing of functional layers (impedance changes affecting loss). An in-situ x-ray inspection was completed with the test unit inside the vacuum chamber to determine the condition of the cable without incurring the risk of moving it. One of the x-rays images taken (courtesy of Aerospace Corporation) [Acknowledgments / Reference 2, 3] is shown in Figure 7. This x-ray happens to be same cable as shown in Figure 1 (left photo). Figure 7. In-Situ X-Ray inspection photo taken of Test Unit in Chamber [2] reprinted with permission of The Aerospace Corporation At the onset of the degradation the RF indicators implied a gap forming on the outer conductor functional layer (impedance change showing up as a TDR spike/ reflectance). Additional cycling caused the gap to propagate to a full circumferential crack around the cable and after a full breach was realized, additional cycling increased the separation along the length of the cable (gap grew larger). As the gap began to expand around the circumference of the cable, the insertion loss became evident in the higher frequency range and moved to the lower frequency ranges as the gap expanded and grew larger. Based on bench-top testing completed at Lockheed Martin and repeated by Aerospace Corporation, (Acknowledgments 1a /References 2), the initiated crack would have had to have been at least 90 degrees circumferentially around the cable to show up in the highest frequency range (20 GHz). A crack development of 270 degrees about the circumference would impact the 10 GHz range. A full breach will

Document

AMS_2012.pdf

218 impact multiple frequency ranges with variable RF performance impacts dependent on the cable geometry, conductor condition, bending requirements, size of the gap, etc.. The bench-top testing completed at Lockheed Martin [Acknowledgments 1a] and Aerospace Corporation [2] mapped the rapid physical degradation to the RF indicators. Prior to the significant RF degradation resulting from the damage as shown in Figure 7, there was no warning of an issue in the cable. Unfortunately this means if you can see a change in the RF, an issue is likely already present indicating some level of damage to the cable. There is no leading indicator of actual cold welding in the RF data until the damage is done. The effect of the cold welding condition occurs after a cable is moved and the damage is within the active cable layers. RF position dependent performance was due to the breach in the outer conductor foil layer opening (higher impedance) when bending the cable and closing (recovering) by reversing the bending / motion. On a similar Design 1 Lot C cable, a partial breach of the outer conductor manifested itself in a similar manner, but was in a much lower strain area. Prior to DPA, a CT scan of the break area (this cable did not have a full circumferential break) was completed. The CT scan revealed that the center conductor had failed. The center conductor strands appear to have failed due to accelerated fatigue caused by increased bending from a partial breach. A likely contributor to the accelerated fatigue in the cable is an adjacent stiffness increase due to cold welded cable layers. The CT scan image of the break area is shown in Figure 8 as a section view so that the center conductor is visible; this image is of the same cable as shown in Figure 1 (right photo). Figure 8. CT Scan inspection photo taken of a partially breached cable with center conductor condition shown (p art of outer conductor / shield taken out of visible frame) As can be inferred from the photos of damaged cables and the trailing RF indicator, it is imperative to test cables well to screen for this condition. Experience has shown that cold welding has led to a progressive damage process and, while not proven experimentally, cumulative fatigue damage is expected to be a non-linear function of cyclic flexural amplitude times the number of cycles typical of metallic fatigue. Consequently, the design of the coax cable and the cable management system must be within proven fatigue capabilities for an expected mission angular movement profile [Acknowledgments 1b].

Document

AMS_2012.pdf

219 Possible contributors and/or additional areas of concern References on cold welding indicate time is necessary for cold welds to form. Although it does not take very long for cold welds to form, continuously running a test without pausing has been shown in past tests to prevent cold welds from occurring. It is believed that in the test with the Design 1 Lot C cable that material conditions were right within the cable layers to cause cold welds, the time and opportunity was given in the test pause, and after restarting the test and inducing motion on the cold welded cable the damage progressed to a severe level. Completing a cable test conti nuously without pauses impacts the cable layers ability to cold weld, potentially masking an issue in the cable. Mechanism idle periods during a mission promotes the development of cold welds (to become more prominent or stronger), in effect accelerating the degradation in the cable when mechanism operation resumes. Complete cable testing as flight-like as possible to ensure that the operation and stopping or non-operation is adequately simulated. Conversely if the application requires the cable to continually move this may actually prevent severe cold welds from occurring. Several vacuum life tests were run with Design 2 test cables. The most apparent advantage of the intermediate layer to cable performance is preventing the braid to outer conductor cold welding. It does not, however, prevent the various layers from cold welding to themselves (outer conductor wrap to adjacent outer conductor wrap or braid to itself). Some Design 2 cables were tested with a profile consistent with the other test units but with small angle dithers incorporated at each extreme (while similar cables did not undergo the small angle dithers). These dithers were not analytical contributors to fatigue so were not initially considered to affect the overall performance of the cable, but they did, however, have an effect on the condition of the cable noted in DPA. The cables that had small angle dithers incorporated into their test profile had cold welding between the outer conductors foil layers (lower right picture on Figure 3). The cold welds within the Design 2 test cable did not impact the RF performance of that specific cable but their significance in a design susceptible to multilayer cold welding may be severe. What was concluded from this result is that small angle motion likely exacerbates cold welding as it contributes to cleaning the surface between the metals (but is not necessary for cold welding to occur if the surfaces are already clean). Small angle dithers, although not a large contributor to fatigue can polish a surface, pushing barrier films away and cleaning the surfaces through movement. What is also theorized is that the cold welding did not impact the performance because it was only between the outer conductor and itself and not between the outer conductor and the braid. The motion that was being stifled at the sporadic cold welds between the outer conductors was made up for with adjacent wraps and layer motion. Clearly the speed/motion profile used in the life test significantly affects the results of the life test. However a surprising result of the investigation indicated that the cables that were vacuum life tested at slower speeds, with pausing and dithers incorporated within the profile (versus a generic encompassing designed profile simulating accelerated cumulative fatigue) degraded more rapidly (with less motion and cycles). Understanding the subtle parameters of a critical contact surface is essential to make informed decisions on how to design a life test given typical time considerations. Critical interfaces should be tested in as flight like manner as possible (Test like you fly) . It has been asked whether a vacuum environment is necessary for cold welding to occur. In the testing completed during this investigation, cold welding in cables was observed for cables tested both vacuum and atmospheric conditions; however all of the critical testing for our investigation was completed inside a vacuum chamber. All DPA activities were performed under ambient laboratory conditions. It is believed by the author that the space vacuum environment exacerbates the condition of cold welding simply because it is a cleaner environment and without air to replenish a potential contamination source (ex. oxidation, nitrides, and water), cold welding is more likely to occur. One test result, although not statistically significant, between two similar cable specimens demonstrated that cold welding occurred more readily in the specimen tested in a vacuum environment (sooner within a similar cyclic test profile) versus the test specimen run in ambient laboratory conditions. Two previously published papers on cold welding within cables [4,5] also mention the affect of vacuum on cold welding.

Document

AMS_2012.pdf

220 Given sufficient time and pressure, two metals of similar composition in clean contact will form cold welds. Applications of concern Depending on the amount of motion required of the cable, cold welds may or may not be a problem. Depending on the geometry of the system, fatigue may or may not be an issue, but additional consideration to local low cycle fatigue of layers due to cold welding should be given. If cold welds form and the cables are not required to move during operation, then cold welding is not a concern. The following are application considerations regarding cold welding to keep in mind throughout the design, verification, or qualification life cycle of a program or MMA design: Cables in or across single deployment mechanisms are generally NOT at risk:  Cable primary failure mode is fatigue (yielding and buckling). Cold welds reduce the number of cycles / amount of motion required for local fatigue failure to occur. Single deployment mechanisms that have to operate once will not generally have excessive cumulative damage from fatigue (even low cycle local fatigue).  Deployments generally occur at beginning of life (BOL). With relatively little time / motion relative to last “successful on-ground” test, damage from cold welds likely will not have the chance propagate to a failure within a single motion / deployment. However:  If the system has low force / torque margins there is some risk of higher force required to move / bend cable relative to what was tested initially if cold welds formed post launch / pre-deployment.  If cold welds form prior to deployment and a very large motion/bending is required for the initial / first and only motion some damage might occur at deployment. The likelihood of cold welds forming and cumulative damage in one motion is low but should be considered in the context of the application. Cables that are expected to operate and move over life (especially high cycle long life mechanisms) ARE at risk:  Dithering motion and vibration can exacerbate cold welds by abrading the surfaces and creating atomically clean surfaces. A space vacuum environment also contributes to maintaining clean surfaces.  Pausing, idle periods and non-operation between operation / motion cycles can exacerbate cold welds by allowing the two adjacent surfaces time to molecularly bond to one another and form stronger cold weld bonds.  Bending the cable through any angular motion has the potential to cause damage to the layers within the cable particularly when the layers have cold welded or are sensitive to fatigue failure; but not all damage is life threatening - the cold welds’ strength, the forced strain on the layers and the fatigue properties of the individual cold welded cable materials will determine the extent of the damage.  Through multiple cycles of cold welded cables the extent of the damage becomes progressively greater and can lead to a full break / breach in a cable.

Document

AMS_2012.pdf

221 Aside from cables, the risks may be obvious, but it is worth mentioning that similar metal to metal contact interfaces requiring motion within mechanisms are at risk.  If two surfaces are required to move relative to each other and there is no (non-metallic) barrier between them (lube, coating or otherwise) then cold welding can lead to failure (seizure / increased force to slide, move or separate; or low cycle fatigue) at that joint / interface.  Any critical surface contact and motion between critical interfaces should be tested in as flight like manner as possible (Test Like You Fly). Recall cold welding is exacerbated by clean surfaces, time, and contact pressure, so ensure testing incorporates / encompasses these critical parameters.  Dithering motion and vibration can exacerbate cold welds by abrading the surfaces and exposing the atomically clean metallic surfaces underneath. A space vacuum environment also contributes to maintaining clean surfaces (no oxidation, sulfides, water vapor, etc) – Test like you fly considerations should be validated.  Pausing, idle periods and non-operation between operation / motion cycles can exacerbate cold welds by allowing the two adjacent surfaces time to molecularly bond to one another and form stronger cold welds. Sequence of initialization, deployments, and storage periods should be given consideration during a qualification of critical contact surfaces.  A higher contact pressure and softer metallic interface contribute to a larger contact surface area and more energy at the interface to encourage the metals to share molecules, i.e., the higher the likelihood of cold welding. Reference 5 goes into further detail about the contribution of surface micro hardness of the constituents as it relates to cold welding.  Material Lot Changes: Hardness, contamination/cleanliness, alloy/material difference, subtle construction differences/ machine and assembly setups all make a difference and should be considered when making a decision on whether cold welding is a concern in your system. Complete a life test on every lot of material for critical surfaces whether or not you have completed a life test in one version or design. Conclusion Moving cables are sensitive to changes in material lot and test parameters including vacuum, strain / motion, speed, and other test like you fly assumptions much like other components within moving mechanical assemblies. This was the conclusion after a multiple year investigation completed on coaxial cable for an aerospace Moving Mechanical Assembly. Fatigue has been widely discussed as the primary failure mode of flexing wire; however, other phenomena need to be considered when designing a cable to move over life. Cold welding, a condition in which welds occurs between similar metals in the absence of heat, can occur on cable layers if precautions are not put in place when choosing a cable design to use for a moving application. For coax cable it has been shown that cold welding between the outer braid and the outer conductor can lead to degradation in RF performance. Three key points were discussed during the course of this paper: 1. Life Test Programs are not usually driven by lot/date code changes, however, life testing for coax cable from each production lot is recommended 2. Coax Cable Designs to be used in life critical MMA’s should incorporate a non-metallic intermediate layer to prevent multi-layer cold welding between any metal to metal layers. Additional attention should be paid to the lot construction of the individual metal constituents

Document

AMS_2012.pdf

222 including the center conductor. Ensure the center conductor is an alloyed copper (higher strength) and that the silver coating on the strands has a well controlled thickness (thicker likely is better based on the test observations). 3. Cold Welding happens with clean metal to metal contact. These lessons learned are applicable to other Moving Mechanical Assembly (MMA) components where sliding or rolling interfaces are metal to metal. Testing should simulate flight like conditions, sequences, and profiles to the maximum extent possible on any MMA especially considering the entire variable set that could potentially affect performance. Acknowledgments (1) Lockheed Martin Space Systems (LMSSC) Data, Assessments from Team, and Findings: a. Special credit goes to the RF Engineers for their extensive coax cable knowledge and testing b. Analysis Group for their analysis expertise in macro and micro structures, metal stress/strain, fatigue and dynamic simulation work c. Failure Analysis Lab for steady hands and patience dissecting tens and hundreds of cable specimens (DPA work) and documenting it beautifully. Additional acknowledgements on the LM Team in Denver: Quality Lab for scanning and rendering all the CT scan cable images, and many more folks on the team that put in countless hours and thought ensuring our results were something we could rely on. Thanks to the whole Lockheed Martin, LM Customer, Aerospace Corporation, and Subcontractor Team for providing their time, resources, test samples, and units from which we gathered a lot of data and knowledge. (2) Also special mention of Bob/Robert Pan from Aerospace Corporation for being a facilitator of information transfer ensuring the contractor / customer and Aerospace team were in constant communication and providing additional data with Aerospace Corporation support through repeating tests for confirmation, refinement, and interpretation of results. References (3) 2D In-situ X-Ray Image (Figure 7) courtesy of The Aerospace Corporation, Dr. Eric C. Johnson, Shant Kenderian, and Robert Pan, Space Material Laboratory. Published Papers on Cold Welding: (4) Coaxial Cable Failure in a Spacecraft Mechanism, by Michael Chiu, Proceedings of the 34th Aerospace Mechanisms Symposium, Goddard Space Flight Center, May 10-12, 2000 where the solution for cold welding in cables was to replace design with a design that had a PTFE barrier between the wire braid and foil (5) Material Property Effects on Coaxial Cable Mechanical Failure, by R. B. Pan, J.B. Chang, C.C. Wan, Y. R. Takeuchi, R. McVey, and I. Chen Proceeding of the 36th Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002 where it shows DPA results and postulates a failure mechanism that relates low yield strength associated with the low hardness values to the cold welding of similar material at the interface, rapid initiation of multiple crack sites and fast crack propagation. (6) Assessment of Cold Welding Between Separable Contact Surfaces Due to Impact and Fretting Under Vacuum; by A. Merstallinger, M. Sales, E. Semerad, Austrian Institute of Technology, and B.D. Dunn, ESA/ESTEC European Space Agency; STM-279, November 2009.

Document

AMS_2012.pdf

223 DLR’s Dynamic Actuator Modules fo r Robotic Space Applications A. Wedler, M. Chalon, K. Landzettel, M. Görner, E. Krämer, R. Gruber, A. Beyer, H-J. Sedlmayr, B. Willberg, W. Bertleff, J. Reill, M. Grebenstein, M. Schedl, A. Albu-Schäffer, G. Hirzinger * Abstract Based on our long-term goal to develop “robonauts” for space and the experiences our research department accumulated in space robotics, this article describes recent design and development results at DLR’s Robotics and Mechatronics Center. Herein, we focus on lightweight robot arms, articulated hands and highly integrated actuation modules for space applications. We show how the development process started with fully sensorized, highly dynamic joint modules with state feedback control and led to seven degree of freedom (DOF) torque-controlled robot arms that enable innovative Cartesian impedance controllers. Further, we present two space robotics examples that are based on such modular actuation units. The first one is the ROKVISS experiment which has been launched towards the ISS in 2004 and returned to earth in 2011. The second is the Dextrous Robot Hand, DEXHAND, which we developed on contract with ESA. These two projects underline the strong interest to transfer our knowledge and experience from terrestrial robotic developments to innovative space technology. Introduction Currently, many operations in space, such as maintenance or new module assembly, have to be carried out by astronauts and expose them to a hazardous environment. These missions are associated with high risks and costs. While robotic systems are not yet ready to replace humans in space, they may provide an excellent support for astronauts. Supported by telemanipulation concepts, robots could be used for many of the Extra Vehicular Activities (EVA). Herein, robotic perception and partial autonomy could strongly simplify tasks for the astronauts. Further, methods like force feedback allow operators haptic interaction at distant locations. Besides near earth orbital activities, we believe that autonomous and semi-autonomous robotic systems will take an important role in planetary exploration scenarios. In order to bring teleoperated systems to the International Space Station (ISS), the European Space Agency (ESA) is currently investigating different scenarios. We think that humanoid robots, like our terrestrial technology demonstrator “Justin”, (Figure 1) could fulfill ESA’s development goals within this context. Since the ISS is designed for humans, sending up robots with hands and arms is an obvious option that also NASA follows with its Robonaut [1]. In the case of a telerobotic scenario, humanoid robots simplify the mapping from the operator to the robotic system. * Robotics and Mechatronics Center (RMC), German Aerospace Center (DLR), Wessling, Germany Proceedings of the 41st Aerospace Mechanisms Symposium, Jet Propulsion Laboratory, May 16-18, 2012

Document

AMS_2012.pdf

224 Figure 1: Teleoperater using a head-mounted display to get 3D visualization and lightweight Robots (LWRs) as haptic input de vice with force feedback and "Jus tin" in telerobotic operation This results in an intuitive operation that minimizes learning time as well as user fatigue and improves execution speed. Upon those arguments, ESA decided to develop an EVA capable hand/arm system. Based on the experiences of our lab with space robotics projects like ROTEX and ROKVISS, ESA assigned the development of the space-qualifiable hand to the DLR Robotics and Mechatronics Center (RMC). Since space robotics applications require highly integrated, lightweight systems with low power consumption and high performance, DLR's robotic actuation technology and our longtime experience with a space robotic system may provide interesting solutions for future missions. Throughout this article we present our robotic drive train technology and its development process as well as two space related application examples. The next section) of this paper discusses the design of the DLR drive train unit including the in-house developed permanent magnet synchronous motors. Furthermore, it describes how a terrestrial seven DOF arm is built based on these actuation units. A later section presents ROKVISS, our robotics component verification experiment that was operated at the outer surface of the ISS from 2004 to 2010. The interesting fact of this experiment is that we used a mixture of space qualified and off-the-shelf components and achieved very reliable long-term operation in space. Receiving the robot back allows us to collect valuable data about the effects of the space environment on the system. Furthermore, this paper describes the development of the EVA qualifiable Dexterous Robotic Hand, DEXHAND, for the ESA. All presented results are based on the Critical Design Review (CDR) status, giving an overview of the technical decisions with respect to the mechanical design. DLR Drive Train Developing actuators for dynamic and sensitive robotic systems is a challenging task. The broad range of applications and the intended interaction with the environment require characteristics that differ from common actuation scenarios in industry. The same robot, for example, has to deliver high torques at low speed in one application while it has to provide high dynamics in another. Since robotic actuators are usually attached to the moving structure, they also need to be small and lightweight. In order to reduce positioning errors and to avoid disturbances during reversing motions all gears should be free of backlash. Thus, the requirements for robotic drive trains are manifold. The most important ones are summarized in the following list: • high torques at low speed or stand still • low weight • small sizes • high dynamics • high efficiency / low losses

Document

AMS_2012.pdf

225 • backlash free gearing • sensor feedback (position and force-torque) In the end of the 1990th no commercially available actuators could sufficiently fulfill those needs. Thus, at our institute, permanent magnet synchronous motors were developed that match the requirements. Being combined with very low backlash Harmonic Drive (HD) gears, they constitute high performance robotic actuation units. These units provide high dynamics and high link side torques due to low inertia parts and the high reduction ratios of the HD gears. Further, their low number of components reduces the probability of failure. In the following, we present the development of our lightweight robotic arms with a focus on our modular actuation units. We begin at the motor stage and progress via the highly integrated actuation modules towards the complete robot. The DLR brushless Motors The key components of the DLR drive train concept are the permanent magnet synchronous motors developed at our institute. The so-called “Innen-Läufer-Motor” (ILM: German for “inside rotating motor”) is well suited for highly dynamic tasks involving frequently reversing motions. For maximum performance, all of our motors are sized to comply with available HD gear dimensions. Figure 2: Concurrent engineering for the actuator design and the RoboDrive variations During the two year development process we strictly followed a concurrent engineering and optimization process (see Figure 2) to achieve high dynamics combined with low losses and low weight. The complete design is based on extensive multi-physics simulations, considering electromagnetic, mechanical and thermal effects. In the final design short copper paths, very dense and optimized windings, strong neodymium rotor magnets and low inertia rotating parts are the main source of the high dynamics. Furthermore, the ILMs consist of a high number of pole pairs (ILM 25 and ILM 38 have 7 pole pairs all others have 10), which allow very accurate positioning and reduce motor ripples. Using wires with a large cross-sectional area allows the motor to handle large currents with low losses and thus, to produce very large torques at low angular velocities. To give a performance example, the torque/current constant of the ILM 25 (stator diameter of 25 mm) is 0.008 N-m/A. The motor uses approximately 12.5 A (without saturation effect) to provide a torque of 100 mN-m. The corresponding HD gears with 1:100 reduction ratio (HFUC8) for example has a nominal torque of 2.4 N-m, a repeatable torque of 4.8 N-m and a momentary peak torque of 9 N-m. The limiting factors for the motor torques are thermal constraints related to the passive cooling. In case of active cooling the torque output of our motors could be even increased. At the time of development the motors achieved a 50% reduction of weight and losses with respect to commercially available units with comparable torque

Document

AMS_2012.pdf

226 ratings. Currently, our motors are becoming the state of the art for high performance robotic drives and are commercialized by the company RoboDrive [2]. The DLR Actuator Units As already mentioned, our motors are developed to comply with the ratings of Harmonic Drive gears. Together with those they form the core of our highly integrated actuation units. Aside those core parts, the units include power, data acquisition, motor control and communication electronics in a very dense package. The amount of wires has been drastically reduced to two power (one for electronics and one for the motors) two communication lines (SERCOS is a ring) and one emergency stop. Due to the large hollow shaft of the motors these lines can be easily guided through a robot joint and thus remain inside the robot. Furthermore, the actuation units comprise a large set of sensors enabling sophisticated impedance and admittance controllers with underlying high performance torque or position control loops. These sensors are motor side position sensors for commutation and relative position measurement as well as link side absolute position and joint torque sensors (see Figure 3). Aside the high performance, low weight and low losses, two properties of our actuation units are very interesting for space applications. First, the high torque capacity of the motors allows applying short and very strong “kicks” that could free the unit in case of cold welding within the bearings or the HD stage. Of course, in this case the power electronics has to be designed to supply large currents for a short time. Figure 3: DLR drive train concept a.) ROKVISS unit - ILM70 - HFUC25, b.) Joint module of the LWR III The second advantage is that the low number of mechanical components reduces the probability of failure (see Figure 4). In the following, the mechanical components of one joint unit are listed: • stator • rotor with magnets • rotor bearing • HD wave generator (WG) • HD flex spline (FS) • HD circular spline (CS) • output bearing. Within these components the rotor bearing and the WG bearing are the most crucial elements since the associated friction has to be overcome by the motor torque directly.

Document

AMS_2012.pdf

227 Figure 4: DLR drive train concept a.) DEXHAND unit CAD, b.) DEXHAND unit photo, c.) ROKVISS unit photo In summary, the design of our actuation units allows highly dynamic and powerful operation as well as precise control in fine manipulation tasks. An example for the usage of this drive train concept within a space application is the joints of our ROKVISS experiment that is described in more detail later. The DLR modular Arm Developing, building and controlling lightweight robotic arms for a large variety of applications are a research focus at the DLR Robotics and Mechatronics Center (RMC). Starting with the Lightweight Robot I (LWR I) presented in 1994 via the LWR II of 1999 to the current model LWR III from 2003, we gained a lot of experience within this field. Due to the success within the research community and an initial market demand for such technology, in 2006 the KUKA GmbH started to commercialize the LWR III. A first small batch series was mainly provided to different research institutions. Figure 5: a.) LWR III, b.) LWR as a KUKA product, c. ) modular robot system “Justin” with 54 DOF Currently, around hundred units are sold and the car manufacturer Daimler sees high potential for a factory-wide deployment of the LWRs. For them, especially attractive is the possibility of switching in-between torque-based and position-based control concepts. This is very promising for the solution of challenging assembly tasks. Besides the innovative impedance and admittance control algorithms developed for our lightweight robots, its modular joint-link concept is one of the reasons for its success. Hereby, the underlying modular assembly system with only a few basic components concerning joint mechanics, electronics and links allows composing completely different configurations in a short time. Thus with three different single degree of freedom (DOF) robot joints and a two DOF wrist joint, different kinematic configurations can be easily assembled. For example, in addition to the symmetric version an asymmetric version (Figure 6) can be built. This is particularly interesting for space applications where the arm needs to be folded for stowing.

Document

AMS_2012.pdf

228 Again the design process for the joint and link modules followed a concurrent engineering process including kinematic and dynamic simulations as well as an FEM analysis. Supported by a link component library with variable parameters, virtual prototypes could be assembled, tested and optimized. One special result of this process is the ball-shaped two axis wrist joint (see Figure 7). This joint imitates the human wrist but achieves much higher mobility. Its short distance between the wrist pitch axis and the tool-center-point (TCP), helps to reduce the motion of the lower robot joints in case of changes of the TCP orientation. Furthermore, the wrist design enables two kinematic configurations, a common configuration with roll-pitch-roll axis and optional a roll-pitch-pitch configuration. To reconfigure the joint the short standard flange simply has to be replaced with an extended 90° bend flange. Another advantage of the wrist is that its cardan joint avoids singularities in an extended position and the reduced joint mass allows carrying larger payloads. Figure 6: a.) symmetric LWR modules ,b.) unsymmetric LWR modules The DLR lightweight robots are very well suited for a large variety of service robotics applications with their modular assembly, kinematic redundancy (7 DOF), the torque controlled joints, the low inertia and a load to weight ratio of nearly 1:1,. Some examples at our institute are the humanoid “Justin”, the three arm co-worker scenario, several single arm manipulation tasks as well as the bi-manual input station for robotic experiments. The concept of highly integrated modular actuation units is further improved for the new DLR hand arm system as well as the DEXHAND. With their low power consumption and the above mentioned features our drive trains and complete robots provide interesting capabilities for future space missions [3]. Figure 7: LWR cardanic hand module joint a.) ro ll-pitch-roll configuration, b.) roll-pitchpitch configuration c.) LWR joint modules

Document

AMS_2012.pdf

229 ROKVISS The German space robotic project ROKVISS (German:”Roboter Komponenten Verifikation auf der ISS”) has just been finished. The system has been returned to earth early 2011 and the hardware is now subject to detailed analysis and advanced tests. ROKVISS (see Figure 8) aims at the space qualification of DLR’s LWR modules (modified for a space environment) with a reduced setup of a 2 DOF robot arm on the ISS. With this experiment DLR could prove the concept can use partly components-off-the-shelf (COTS) within a very tight and highly integrated mechatronic device based on the modular joint concept discussed previously. After more than five years of successful operation and without any failure the innovative mechanical and electronic concept seems very promising for the use in space projects. Those projects are in the wide range of future robonaut applications as well as planetary robotic missions. Indeed what we are still missing in space are fast signal transmissions. One relay satellite could provide signal round trip delays of 0.5 seconds within a coverage times of around 40 minutes. This allows even haptic feedback, which is a major idea of our advanced telerobot control strategy using visual and haptic feedback [4]. ROKVISS was also a telerobotic demonstrator with real-time stereo-video transmission and tactile feedback. The telerobot success was the second major achievement of ROKVISS, besides the component verification. We believe that realistic telerobotics probably combined with partial or full autonomy will be needed in future space robotic applications [5-7]. Figure 8: ROKVISS photo of the experiment mounted on the ISS The following subsection deals with the mechanical design and the detailed structure of the ROKVISS robot describes the friction analysis and shows preliminary results. Mechanical System Design In Figure 8a the overall ROKVISS is shown, while Figure 8b shows the joints as described in the previous section. Figure 8c shows ROKVISS in its position outside the ISS at the Zvezda service module. The ROKVISS experiment consists of a small robot with two torque-controlled joints, mounted on a Universal Workplate (UWP). Each joint module consists of a power supply, a controller board, a power converter, the drive train and a torque sensor. Inside the robot there are 2 joint modules, a stereo camera, an illumination system, an earth observation camera and a power supply. A mechanical contour device is mounted on the UWP for verifying the robot's functions and performance (see Figures8a and 8c).

Document

AMS_2012.pdf

230 Figure 9: ROKVISS FM modules with color changes on the surface after return to earth These two robot joints have been extensively tested and the joint parameters have been identified by repetitively performing predefined robot tasks in an automatic mode, or based on direct operator interaction. The automatic mode is necessary due to the fact that communication constraints limit the direct link experiment time to windows of only up to seven minutes when the ISS passes over the tracking station German Space Operations Center ( GSOC). Figure 9 shows the returned two joint modules. After five years of operation on ISS differences on surface with its anodic treatment (LN9368 I 2101) of the two different aluminum alloys (Al 7075 I Al 5083) are obvious [8, 9]. Friction Analyses The preliminary results of the on-orbit identification show that the total friction for joint 1 in space increased by about 50% compared to the friction on ground, taken at 20°C, under normal atmospheric pressure (Figure 10). However, only a small further degradation of the parameters has been observed so far during the mission. This friction change will be analyzed in the upcoming tests to determine if the lubricant (Braycote 601) and/or the guidance condition influenced this effect. The temperature dependency of the parameters is close to the range of identification uncertainty (Figure 10) [10]. Conclusion for the Friction Identification The friction in the two joints increased initially, already at the first experiment in space compared to the values measured on ground, but remained rather constant afterwards. All the controllers were robust with respect to these changes. The friction changes of joint 1 and joint 2 are different. Friction in joint 1 has the same structure as on ground, but values increased by 50%. The friction in joint 2 shows no more significant viscous friction and the total increase is lower than 20%. Since the same lubricant was used for both joints (Braycote 601), further experiments are planned in order to explain the different behavior. To provide a better fit of measured and simulation data, a nonlinear (third order polynomial) function is required. However, due to the higher number of optimization parameters, their variance over different experiments is greater. A trajectory with a higher number of distinct velocities is needed in order to reduce the variance of the results. Stiffness Identification The main sources of elasticity in the joints are the flex splines of the HDs and the torque sensors. The elasticity is identified by contacting a rigid surface with the tip of the robot and by commanding a slowly changing force to the joints. Since the torque is measured after the gear-box, the stiffness can be easily identified with the available torque and position signal. The model torque is computed as a product of position increment and stiffness. The stiffness for joint 2 has a value around 4900 N-m/rad in space.

Document

AMS_2012.pdf

231 Figure 10: Friction analysis of the FM joints (a. Joint one, b. Joint two) a.) b.)05101520253035404550 02.11.04 24.03.05 12.05.05/1 27.05.05 08.06.05 16.06.05 22.06.05 07.07.05 22.07.05 08.09.05 12.10.05 30.11.05 21.12.05 22.12.05 11.05.06 16.08.06 17.08.06 04.10.06 17.10.06 19.10.06 25.10.06 17.01.07 18.01.07 23.01.07 13.02.07/1 13.02.07/2 14.02.07/1 14.02.07/2 19.02.07 20.02.07 17.04.07 19.04.07 23.05.07 24.10.07 05.12.07 06.12.07 23.01.08 22.04.08 23.04.08 28.04.08 17.06.08/1 17.06.08/2 25.06.08 22.07.08 24.07.08 30.07.08 20.08.08 27.08.08 18.09.08 16.09.09 03.12.09 02.02.10 30.03.10 13.05.10 07.09.10 20.09.11/1 20.09.11/2ROKVISS - friction identifikation joint 1Nm-20°C<T<=-14°C-14°C<T<= -8°C-8°C<T<= -2°C-2°C<T<= +4°C+4°C<T<=+10°C+10°C<T<=+16°C+16°C<T<=+22°C+22°C<T<=+28°C+28°C<T<=+34°C+34°C<T<=+40°Cviscos friction at 1°, 5° [deg/s] coulomb frictionload-dependent friction 05101520253035404550 02.11.04 24.03.05 12.05.05/1 12.05.05/2 27.05.05 08.06.05 16.06.05 07.07.05 22.07.05 08.09.05 12.10.05 14.10.05 09.11.05 21.12.05 22.12.05/1 22.12.05/2 17.08.06 04.10.06 17.10.06 19.10.06 25.10.06 17.01.07 18.01.07 23.01.07 13.02.07/1 13.02.07/2 14.02.07/1 14.02.07/2 19.02.07 20.02.07 17.04.07 19.04.07 23.05.07 24.10.07 05.12.07 06.12.07 23.01.08 22.04.08 23.04.08 28.04.08 17.06.08/1 17.06.08/2 15.06.08 22.07.08 24.07.08 30.07.08 20.08.08 27.08.08 18.09.08 16.09.09 03.12.09 02.02.10 30.03.10 13.05.10 07.09.10 20.09.11/1 20.09.11/2ROKVISS - friction identifikation joint 2Nm-20°C<T<=-14°C-14°C<T<= -8°C-8°C<T<= -2°C-2°C<T<= +4°C+4°C<T<=+10°C+10°C<T<=+16°C+16°C<T<=+22°C+22°C<T<=+28°C+28°C<T<=+34°C+34°C<T<=+40°Cviscos friction at 1°, 5°, 10°, 20°, 30° [deg/s] coulomb frictionload-dependent friction

Document

AMS_2012.pdf

232 No significant differences between the stiffness values on the ground and in space are observed. Still, a change of about 15% can be observed. However, this deviation is within the range of the load dependent stiffness variation and the controllers are designed to be robust with respect to such uncertainties. Conclusions Though the ROKVISS mission was planned for one year, already after 8 months of operation the main goals of the mission were achieved. The hardware proved to work reliably under space conditions. The dynamical parameters of the joints, although somewhat different than on ground, show a small variation over time and with temperature. The controller structures proved to be robust with respect to these variations and to the used COTS elements. Within the cooperation with the Russian Institute for Robotics and Cybernetics in St. Petersburg (RTC) the mission of ROKVISS has been extended step by step up to end 2010. Then the collective decision has been taken to bring the main parts of ROKVISS down to earth. Currently we are in close cooperation with the RTC to accomplish tests and analysis of these modules. Initially, complete segment tests are planned to identify the behavior and especially understand the correlation between conditions on ISS, thermal vacuum chamber in the test facility and the conditions in the laboratory environment. With this experience we may gain the competence to build “Justin” for space application. DEXHAND The development of a torque controlled multi-fingered hand is also a domain in which the DLR has a long history [11–14]. The DEXHAND is designed to be able to perform a set of generic space oriented tasks. For example, the removal of a multi layered insulation (MLI) cover or the manipulation of a handle. This section presents how the design concept and the architecture of the hand are selected based on the hand capability requirements. As discussed in the previous sections, the DEXHAND is in line with the DLR overall goal to develop a complex two handed system for space application such as EVA assistance or substitution of astronauts [15, 16]. The first part that follows presents the main requirements for the system and explains how the general concepts have been selected and the overall architecture is described. The second part “reveals” the mechanical structure, the tendon actuation system and the torque sensor implementation. The fourth part gives an overview of the control system, controller architecture and the software distribution. Requirements and Concept The design of the DEXHAND is driven by its required capabilities. Some constraints are purely technical: operating temperature, maximum fingertip forces, joint velocity, but others are functional, such as grasping and operating a pistol grip tool (a space version of an electric driller). The desired capabilities must be translated into technical requirements that result in a trade-off between system complexity, capabilities, reliability, volume, weight and cost. Finally top level functional requirements have been defined, such as the DEXHAND shall be able to grasp the following EVA tools and to support their operations: Pliers, Scissors, Hammer, Tether, Scoop, Cutter, Allen Wrench et cetera. Successful operation of the tools implies force closure of the grasp with respect to the preferably form closure which should be achieved. In the robotic community, hands are ranging from the simplest grippers to the most advanced biomimetic devices. The design space (i.e., the possible design solutions) of hands is extremely large, therefore, the first part of the project was to select a concept that would fit to the initial requirements. Examples of parameters that must be selected are: • number of fingers • number of degrees of freedom (DOF) per finger • number of actuated DOFs • placement of the fingers in the hand • shape of the fingertips

Document

AMS_2012.pdf

233 • size of the fingers • etc… Certainly, each finger DOF brings more capabilities but increases the number of parts. The use of multiple small actuators instead of one large actuator, increases the capabilities to distribute power losses, might provide redundancy, and usually provides a better form factor. However, the raw power density is reduced. The control complexity and the number of sensors must also be increased in order to take advantage of the available degrees of freedom. The principle parameters are selected based on the manipulation experience gathered with the DLR Hand II and DLR/HIT hand. A parameter refinement is done by simulating grasps with each object of the EVA tool list. For example, in order to perform trigger actuation of the pistol grip tool, while maintaining tool stability, it appears that at least three fingers are required. .For fine manipulation, the shape of the fingertips is playing a key role. Several shapes are compared with respect to rolling, maximum load and the ability to pick up small objects. The DEXHAND is using a variable curvature with a flat end fingertip shape. Finally, the DEXHAND development differs from the hand developments at DLR. This time the modularization borderline changes (see Section 1.2). For the LWR III the border of the modularization was at joint level. The Hand II (as well as DLR/HIT hand) modularization borderline is based on fingers. While the DEXHAND and as well the Hand Arm System (HaSy [14]) are based on motor modules as previously discussed. Overall Architecture The DEXHAND system is developed for use with a robotic arm (Dexarm) designed and realized by Selex Galileo. The hand has 12 actuated DOFs, distributed in 4 fingers with 3 degrees of freedom each. Figure 11 presents the latest state of the CAD model of one Finger and Figure 12 shows two photos of the DEXHAND prototype with housings and without. The actuation system is based on geared motors followed by a tendon transmission system (see first section). The motors are controlled using a combination of a DSP, FPGA and motor controllers. Joint torque measurements are available and realized with full bridge strain gauge sensors. Multiple temperature sensors are available to protect the system against overheating and freezing. The control system of the hand is able to run entirely inside the hand. The DEXHAND is required to communicate over a CAN bus with a common VxWorks communication controller. The communication to control the Dexarm is routed as well through a real-time VxWorks system. It will allow the hand and the arm controllers to be tightly synchronized in the future. In the DEXHAND, modular fingers are used in order to increase the system reliability. It also improves the cost efficiency of the project. However, based on a kinematic analysis and the experience from the DLR Hand II, a special finger is used for the thumb. As shown in [17], the thumb deserves a special treatment in order to increase the hand dexterity. For example, in order to properly oppose to the other fingers the thumb should have at least twice the maximum fingertip force. The DEXHAND fingers are design to actively produce a fingertip force of 25 N (for the stretched finger) while withstanding 100 N passively. Mechanical Design The transmission system is using polymer Dyneema tendons and harmonic drives in order to bring the motor torque to the joints. The concept keeps the extremities (the fingers) with radiation uncritical electronics. The shielding strategy consists in housing the whole electronic system in an aluminum shell with at least 2 mm thickness. This leads to a fully electro-magnetic interference (EMI) sealed hand body containing:- the drives, the power electronics and the communication electronics. The only exceptions are the torque sensors, based on strain gauges, and some temperature sensors, which have to be placed in the fingers. The design successfully encases all electronic systems in its protective housing.

Document

AMS_2012.pdf

234 Figure 11: Actuation principle of the DEXHAND fingers The cardanic metacarpophalangeal (MP) base joint is driven by two motors. Due to the coupling of the tendons in the MP joint the two motor torques can be used together for one DOF of the base joint. This aspect opens the possibility of using the same motors for the base joint and the proximal inter-phalangeal (PIP) joint. Indeed, due to the difference of lever length (pulley radius), the required torques are scaled dependent to each joint. The PIP joint has a fixed coupling with the distal inter-phalangeal joint (DIP) with a ratio of 1:1. In Figure 11 the section lengths, joint positions and definitions are shown. The coupling matrix P, which relates motor velocity θ with joint velocity q is: ሶሶ (1) =ܲ ଵ ௥೛ଶ0 ଶ0 ଷ൱ (2) Where r p is the motor pulley radii, r 1, r2 and r 3, are the joint pulley radii, and r 13 and r 23 are the pulley radius of the PIP tendons in the base. Given that the coupling matrix is not configuration dependent, the relationship can be integrated in: ݍܲ=ߠ 3) The motor unit for DEXHAND has been developed based on the DLR / RoboDrive (see section1) [2] ILM 25 motor including the gear of a harmonic drive HFUC8 with a transmission ratio of 100:1. The whole unit fits into a cylinder of 27-mm diameter and a length of 17.5 mm with a weight of 46 g (see section 1 Figure 4). The unit provides a continuous torque of 2.4 N-m with peaks up to 9 N-m which is the maximum peak torque of the gearing. In the DEXHAND, the motor has been electronically limited to 2 Nm for power reasons. Each actuated joint has a reference mark in the middle of its motion range. These reference marks are composed of a small magnet and a Hall-effect sensor located in the actuated joint. The joint torque measurement is implemented using a sensing body and full bridge strain gauges sensors (Figure 12). The torque sensors are all physically located in the proximal finger link. Therefore, the sensor for the PIP/DIP-joints measures the reaction torque of the coupling tendons.

Document

AMS_2012.pdf

235 Figure 12: a.) Finger without housing, b.) Strain gauge sensor body All sensors, except one of the reference sensors of the MP joint, are located and mounted in one mechanical part. This part also contains the pull relief of the wires, as well as the shield connection of the cable from the fingers to the electronics in the palm. This simplifies the assembly and the maintenance of the finger. Special care was taken in the design of the sensor body in order to prevent temperature drift. The force measurements obtained in a thermal chamber from −50° to+70° confirmed the measurement stability. The palm structure consists of 11 main segments. These segments are massive aluminum parts to improve heat transfer and increase the thermal inertia. The modules of the DEXHAND are four different ensembles representing the ring-, middle-, index- and thumb finger actuation units. A unit includes the tendon guidance from the motor pulleys to the MP and DIP joint. The palm surface mainly consists of the outer shell parts. Furthermore, all parts are designed without sharp edges. They are optimized for ideal thermal allocation and minimum resistivity (Figure 13 shows the hand without the palm grasp pads). Figure 13: a.) Section view of DEXHAND with its components, b.) PCB placement in the wrist The wrist houses all electronic parts. This includes the analog, the digital and the power circuit boards. The electronic boards are coupled to the palm assembly with connectors. The housing of the DEXHAND is composed of 2-mm aluminum shells. It is fully closed to provide good EM compatibility. During electro static discharge (ESD) tests with 4-kV impulse at the fingertip and palm structure no failure has been detected. The final DEXHAND design is presented in Figure 13a. Furthermore the compact and also fully integrated housing of the electronic inside the wrist is shown in Figure 13b. Software and Control As presented in the previous section, the hand comprises a DSP and a FPGA. The FPGA is used for low level, high speed functional blocks. In the DSP, functional blocks such as calibration tables, impedance control loop, communication, and safety systems are implemented. The controller itself is an impedance controller running at 1 kHz. The communication to the system is done via a CAN bus, on which no hard real-time data is allowed. The commands to the system are always asynchronous. In order to establish telerobotic scenario, a VxWorks

Document

AMS_2012.pdf

236 real-time system is used to establish the data exchange between the data glove data source and the DEXHAND. Moreover, the system is used to monitor the communication bus and to provide a TCP/IP communication interface from any standard operating system. Conclusion After more than 20 years of torque-controlled lightweight robot development the third generation becomes a useful product now, where we have tried to use all present day available simulation and computational technologies to approach the technical limits. Taking into account its weight of 13-14 kg, its typical power consumption of little more than 100 watts, its load capability of around 8 kg, its motion speed based on maximal joint speeds of 180 deg /sec, it is probably one of the lightest robots that have been built so far. It is the basis for DLR’s future space robot developments . The presented ROKVISS project has proven the capability of using the lightweight robot and motor unit concepts for space application. The experience acquired with the ROKVISS project influenced many decisions. We believe that upcoming tests and results from analyzing the returned hardware will even further improve the DLR space developments. Furthermore this paper presented an overview of the DEXHAND project with its 12 motor units, its tendon driven actuation principal and its overall mass of 3 kg. The fingers proved to be capable of applying 25 N at the fingertip, and can withstand 100 N impact loads. It should be noted that the biggest issue in terms of thermal control of DEXHAND are the digital electronic circuits, the power inverters and not the motors. Indeed, they have very little dissipation capabilities. The EM compatibility issue is mainly solved by using a thick enclosing aluminum shell, for the radiation issues only space qualified parts are used. The used commercial parts are qualified with TID tests. The control system runs entirely in the DEXHAND, complying with the very low communication bandwidth requirements. Acknowledgment We would like to thank the Bavarian Ministry that has made mechatronics a key topic of Bavaria’s high tech offensive with lightweight robotics and articulated hands as central demonstrators. The authors would like to thank the DEXHAND and ROKVISS team at DLR, as well as, the ESA for the opportunity to develop an EVA capable hand. The DEXHAND Project has been founded with the ESA Contract No. 21929/08/NL/EM. References [1] M. A. Diftler, R. O. Ambrose, S. M. Goza, K. Tyree, and E. Huber, “Robonaut Mobile Autonomy: Initial Experiments,” in IEEE International Conference on Robotics and Automation, Barcelona, Spain, April 2005, pp. 1437 – 1442. [2] RoboDrive (2009) Website of the RoboDrive GmbH Company. [Online]. Available: http://www.robodrive.de/ [3] G. Hirzinger, N. Sporer, A. Albu-Schäffer, M. Hahnle, R. Krenn, A. Pascucci, and M. Schedl, “Dlr’s torque-controlled light weight robot iii-are we reaching the technological limits now?” in Proc. IEEE Int. Conf. Robotics and Automation ICRA ’02, vol. 2, 2002, pp. 1710–1716. [4] G. Niemeyer, C. Preusche, and G. Hirzinger, Handbook of Robotics. Springer Verlag, 2008, vol. ISBN 978-3-540-23957-4, ch. Telerobotics.

Document

AMS_2012.pdf

237 [5] C. Preusche, D. Reintsema, K. Landzettel, M. Fischer, and G. Hirzinger, “DLR on the way towards telepresent on-orbit servicing,” in Proc. Mechatronics & Robotics 2004, 2004. [6] C. Preusche, D. Reintsema, K. Landzettel, and G. Hirzinger, “ROKVISS - towards telepresence control in advanced space missions,” in Proc. 3rd. International Conference on Humanoid Robots (Humanoids 2003), Munich and Karlsruhe, Oct. 2003. [7] B. Schäfer, K. Landzettel, A. Albu-Sch¨affer, and G. Hirzinger, “ROKVISS: Orbital testbed for tele-presence experiments, novel robotic components and dynamics models verification,” in Proc. 8th ESA Workshop on Advanced Space Technologies for Robotics and Automation (ASTRA), Noordwijk, The Netherlands, Nov. 2-4 2004. [8] K. Landzettel, B. Brunner, R. Lampariello, C. Preusche, D. Reintsema, and G. Hirzinger, “System prerequisites and operational modes for on orbit servicing,” in Proc. ISTS International Symposium on Space Technology and Science, Miyazaki, Japan, May 30-June 6 2004. [9] G. Hirzinger, K. Landzettel, and et al., “ROKVISS robotics component verification on ISS,” in Proc. of ’the 8th Int. Symposium on Artificial Intelligence, Robotics and Automation in Space - iSAIRAS, Munich, Germany, 2005. [10] A. Albu-Schäffer, W. Bertleff, B. Rebele, B. Schäfer, K. Landzettel, and G. Hirzinger, “Rokviss - robotics component verification on ISS current experimental results on parameter identification.” in ICRA. IEEE, 2006, pp. 3879–3885. [11] J. Butterfaß, G. Hirzinger, S. Knoch, and H. Liu, “DLR’s Multisensory Hand Part I: Hard- and software architecture,” Proceedings of the IEEE Int. Conf. on Robotics and Automation, 1998. [12] C. Borst, M. Fischer, S. Haidacher, H. Liu, and G. Hirzinger, “DLR hand II: experiments and experiences with an anthropomorphic hand,” in ICRA, 2003, pp. 702–707. [13] Z. Chen, N. Y.Lii, T. Wimboeck, S. Fan, M. Jin, C. H. Borst, and H. Liu, “Experimental study on impedance control for the five-fingered dexterous robot hand DLR-HIT II,” Proceedings - IEEE IROS, 2010. [14] M. Grebenstein and P. van der Smagt, “Antagonism for a highly anthropomorphic hand arm system,” Advanced Robotics, no. 22, pp. 39–55, 2008. [15] A. Wedler, M. Chalon, and et al., “DLR’s space qualifiable multi-fingered dexhand,” in Proc.:11th Symposium on Advanced Space Technologies in Robotics and Automation (ASTRA), vol. 11, ESA. ESA/ESTEC, Noordwijk, the Netherlands: ESA, 12 14 April 2011, p. Session 3a. [16] M. Chalon, A. Wedler, and et al., “Dexhand: A space qualified multi-fingered robotic hand,” in Proc. IEEE Int Robotics and Automation (ICRA) Conf, 2011, pp. 2204–2210. [17] M. Chalon, T. Wimböck, M. Grebenstein, and G. Hirzinger, “The thumb: Guidelines for a robotic design,” in IROS, 2010.

Document

AMS_2012.pdf

239 Mars Science Laboratory Rover Integrated Pump Assembly Bellows Jamming Failure Michael R. Johnson1*, Joel Johnson*, Gajanana Birur*, Pradeep Bhandari* and Paul Karlmann* Abstract The Mars Science Laboratory rover and spacecraft utilize two mechanically pumped fluid loops for heat transfer to and from the internal electronics assemblies and the Radioisotope Thermo-Electric Generator (RTG). The heat transfer fluid is Freon R-11 (CFC-11) which has a large coefficient of thermal expansion. The Freon within the heat transfer system must have a volume for safe expansion of the fluid as the system temperature rises. The device used for this function is a gas-over-liquid accumulator. The accumulator uses a metal bellows to separate the fluid and gas sections. During expansion and contraction of the fluid in the system, the bellows extends and retracts to provide the needed volume change. During final testing of a spare unit, the bellows would not extend the full distance required to provide the needed expansion volume. Increasing the fluid pressure did not loosen the jammed bellows either. No amount of stroking the bellows back and forth would get it to pass the jamming point. This type of failure, if it occurred during flight, would result in significant overpressure of the heat transfer system leading to a burst failure at some point in the system piping. A loss of the Freon fluid would soon result in a loss of the mission. The determination of the source of the jamming of the bellows was quite elusive, leading to an extensive series of tests and analyses. The testing and analyses did indicate the root cause of the failure, qualitatively. The results did not provide a set of dimensional limits for the existing hardware design that would guarantee proper operation of the accumulator. In the end, a new design was developed that relied on good engineering judgment combined with the test results to select a reliable enough solution that still met other physical constraints of the hardware, the schedule, and the rover system. Introduction The Mars Science Laboratory Mission - Overview The mission consists of four discrete sections or phases: 1. Launch Phase 2. Cruise Phase 3. Entry, Descent, and Landing Phase 4. Surface Operations Phase An exploded view of the assemblies that make up the spacecraft as it heads to Mars is shown in Figure 1. The Cruise Stage is jettisoned just prior to Entr y, Descent, and Landing (EDL). The Backshell and Heat Shield are necessary components during the Entry portion of EDL. The Descent Stage contains fuel and rocket engines to provide a powered descent to the surface with a deployed Rover for a soft touchdown. After the Rover is on the surface, the Descent Stage is cut loose and flies away from the Rover’s location. * Jet Propulsion Laboratory, California Institute of Technology, Pasadena, California Proceedings of the 41st Aerospace Mechanisms Symposium, Jet Propulsion Laboratory, May 16-18, 2012

Document

AMS_2012.pdf

240 Figure 1. Mars Science Laboratory Spacecraft Major Assemblies The Mars Science Laboratory Mission – Thermally The cruise portion of the mission consists of the rover and cruise stage traveling from Earth to Mars over a period of 8½ months. During this time, power is provided to the cruise stage and the rover by solar panels located on the cruise stage and an RTG attached to the rover chassis. The RTG is located inside the spacecraft entry body. Heat generated by the RTG, as well as waste heat from the electronics, is transferred to radiators on the cruise stage using a system of pipes containing Freon R-11. This heat management system is called the Heat Rejection System (HRS). There are two separate fluid systems, one that is part of the cruise stage (CHRS), and one that is part of the rover assembly (RHRS). More detail on these two systems is provided in references 1 and 2. The CHRS is no longer needed after the several month cruise period to Mars so it is vented and then jettisoned from the entry body with the Cruise Stage just prior to the Entry, Descent, and Landing phase. The separate fluid system within the rover must continue to operate throughout landing and the 690 Earth-day primary surface mission, as well as any extended mission time. The main function of the heat transfer system is to distribute heat around the rover assembly for the purpose of maintaining a smaller operating temperature range for the rover electronics than the environment would otherwise require. Heat is transferred from the RTG to the internal Rover Avionics Mounting Plate (RAMP) during cold periods. The fluid is directed to radiators when excess

Document

AMS_2012.pdf

241 heat must be exhausted. By utilizing the heat from the RTG and the fluid Heat Rejection System (HRS), the temperature range for the internal electronics assemblies is substantially reduced from -128°C to +55°C to a smaller range of -40°C to +55°C for the length of the mission. Thermal Hardware Description The thermal system for the rover and cruise stage are very similar. The hardware consists of an Integrated Pump Assembly (IPA), a gas-over-fluid accumulator, a significant length of fluid tubing, and heat transfer surfaces. The IPA consists of a motor, centrifugal pump, motor drive electronics, and directional valves for maintaining the coolant temperature. The valves direct more or less Freon to the radiators depending on the temperature of the coolant. A representation of the internal piping of the rover is shown in Figure 2 and the piping on the cruise stage is shown in Figure 3. Figure 2. Heat Transfer System of Piping in the Rover and Around the RTG (Rover Integrated Pump Assembly)

Document

AMS_2012.pdf

242 Figure 3. Heat Transfer System of Piping in the Cruise Stage and Relationship to the Rover System of Piping Components of the Integrated Pump Assembly (IPA) There are two separate IPA units, one on the Cruise Stage and the other on the Rover. The Cruise Stage IPA is called the CIPA and the Rover IPA is called the RIPA. A model of the RIPA is shown in Figure 4. The pumps provide the circulation power for the Freon R-11, the pressure transducers provide telemetry back to the system indicating the health of the fluid system, the directional valves divide the flow between the radiators and the RTG, and the accumulator provides expansion volume for the fluid in the closed system. The accumulator expansion volume is absolutely necessary because of the large thermal volumetric coefficient of expansion of the Freon of 20% per 100 degree Celsius change in temperature. Providing a volume for the Freon to expand into prevents the following failures:  Mechanical failure of the system tubing due to over-pressurization  Bubble generation from low pressure Figure 4. Rover Integrated Pump Assembly (RIPA) Cruise stage Integrated Pump Assembly (CIPA) is located behind this radiator Motor Drive Electronics Pressure Transducer Directional Valves Accumulator Pump Assembly with Brushless DC Motor & Pump

Document

AMS_2012.pdf

243 The accumulator is a gas-over-fluid design that uses a metal bellows for the barrier between the fluid and gas. The gas is dry Nitrogen and the fluid is the system coolant of Freon R-11. A cross-section of the accumulator is shown in Figure 5. Figure 5. Accumulator Assembly showing bellows in fully extended position (maximum fluid volume, maximum pressure of gas) The free end of the bellows assembly consists of a metallic sweeper to which a Teflon ® guide disk with a larger outer diameter is mounted to provide low friction sliding on the bore of the outer accumulator housing. The bellows is prevented from stroking too far by a step in the housing bore that corresponds to the maximum design fluid volume increase along with the highest gas pressure. As fluid is removed from the accumulator, the gas pressure compresses the bellows until other features of the assembly act as a compression hardstop. This is the state of minimum fluid volume and minimum gas pressure within the accumulator assembly. Failure Event The initial functional testing of the accumulator assembly involved the following steps:  Closing the accumulator assembly without welding (using a mechanical clamp) for a functional check  Filling the bellows with fluid and removing any air bubbles (input tube upright)  Pressurizing the gas side of the accumulator until the bellows is fully compressed  Pumping fluid into the accumulator to extend the bellows, until the pressure difference across the bellows reaches a limiting value (this was the point where the bellows was calculated to be at full stroke with the guide seated against the step in the housing bore)  Removing the pump connection  Allowing the bellows to drain into a graduated cylinder to measure the amount of fluid the accumulator could contain  Verifying that the fluid capacity of the accumulator is within specified tolerances  If the assembly passes the test, weld the bellow end assembly onto the housing for a final flight seal  Repeat the functional test above to verify no change occurred during welding

Document

AMS_2012.pdf

244 During final testing of one of the welded accumulator assemblies, the accumulator’s fluid capacity appeared to be less than half the required value. Upon repeating the test several times, the capacity of the accumulator did not change. No test setup anomalies were observed and the accumulator assembly did not have any discrepancies that could account for the reduced capacity. The accumulator assembly internals were then imaged using X-rays with the fluid in the accumulator at the limiting pressure difference. The picture in Figure 6 shows what was observed. Figure 6. X-ray of Accumulator Assembly showing bellows shape with the maximum allowable pressure difference across the bellows. The accumulator fluid capacity is significantly reduced from the required value. The end of the bellows assembly with the sweeper and Teflon® guide had managed to tip to an angle where the piston would no longer move down the bore and had jammed in place. The force on the piston in the direction of extending the bellows with the maximum allowable pressure difference is 370 N, indicating the jamming force was quite high. When the fluid was removed, the bellows compressed flat against the compression hardstop (the right end in Figure 6) as designed. Repeating the test and measuring the fluid capacity indicated that the failure point was consistent and very stiff – meaning the location along the bore where the piston jammed did not move detectably when cycled between no fluid and the maximum possible fluid, limited by the allowable pressure difference across the bellows. The X-ray was used to make measurements of the geometry of the failure location so it could be identified when the assembly was opened for inspection. Figure 7 shows some of the measurements made to determine the location and the possible source of the jamming. Figure 7. Measurements made using the X-ray of the Accumulator Assembly

Document

AMS_2012.pdf

245 It was clear there was some obstruction to the motion of the bellows guide that would hopefully be identified when the assembly was opened up and inspected. The next step was to carefully remove the bellows assembly from the housing to find the source of the jamming by cutting the weld off the end of the housing. Figure 8 shows what was found inside the housing assembly. The location of the residue seemed to indicate it was the source of the jamming, as indicated in Figure 9. Figure 8. Tape adhesive residue on the Figure 9. Location of tape adhesive residue housing wall, left behind from previous corr elates well to the tip-initiation point rework operation The tape residue was cleaned from the housing bore and the accumulator reassembled. The test was repeated with the full expectation that the problem had been solved. The result of the test indicated no change in the behavior of the assembly, including the measured fluid volumes. The following analysis on the tipping phenomenon was performed (see Figure 10) to understand the sensitivity of the design to the variables involved. A description of the mechanical behavior model is presented. Mechanical Behavior Model Description The free-body diagram variables include the weight of the sweeper (W s), the weight of the fluid-filled bellows (W b), the friction at the sweeper-to-housing interface (µ s), and the normal force generated by a squirming bellows (F b). Prior to the tipping of the end of the bellows, F b is zero. Once the end starts to tip, the bellows squirms upward and reacts against the housing wall. This additional normal force increases the frictional drag between the housing wall and the bellows surface, including the sweeper-to-housing interface. An additional aspect of the tipping end of the bellows is how the contact between the Guide Disk and the housing behaves. The contact starts at a point and, as the tipping angle increases, this contact spot increases to an elliptical area. As the tipping angle is increased more, the contact patch splits into two small contact patches that travel around the housing surface. As the contact points get higher on the housing surface, the normal force necessary to balance the bellows squirm force, F b ,increases substantially. As the normal force increases, so does the friction force, increasing the tipping moment on the end of the bellows. This runaway behavior demonstrates that the tipping of the end of the bellows will eventually jam, and added force will not help to release it. See Figure 11 for a pictorial description of this phenomenon. Figure 10. Free Body diagram of bellows assembly in housing

Document

AMS_2012.pdf

246 Figure 11a. Bellows straight, normal Figure 11b. Start of tipping, normal force on force on bottom is from weight only (F b = 0) bottom is a small contact patch (F b = 0) Fb F b Figure 11c. Tipping angle has increased and Figure 11d. As tip angle increases more, the contact patch has split into two discrete areas contact areas move up the side of the housing. Radial force must increase to balance the bellows squirm force, F b. The most significant variable in this tipping and then jamming behavior is the friction force. The housing bores were cleaned very well to minimize the possibility for contamination. The Teflon ® Guide Disk was relied upon for its low friction properties at the bellows-to-housing interface. No other Accumulator Assembly has exhibited this characteristic and it had not shown up in any of the environmental testing either. Design History of the Accumulator Assembly This accumulator design was not new. There was a Heat Rejection System on the Mars Pathfinder Lander in 1997, as well as one on each of the rovers, Spirit and Opportunity in the 2003 missions. The accumulator in the MSL mission was larger than any of the prior units because the volume of heat transfer fluid was much greater. The design was scaled up in size to accommodate the volumetric increase. The mission life requirement for MSL is eight times longer than the Spirit and Opportunity rovers. The larger size of the accumulator and bellows assemblies coupled with the longer operational life required a cycle life test be performed. The cycle life test resulted in a fatigue failure of the bellows assembly. The bellows was redesigned to accommodate the higher life requirement by reducing the thickness of the convolutions a small amount. The bellows displacement then resulted in a lower stress level in the material. This had the additional effect of reducing the bellows stiffness, both axially and laterally. The Next Version of the Failure Mechanism Hypothesis When the elimination of the adhesive residue on the housing wall did not affect the failure results, the additional factor of the reduced stiffness of the bellows was considered. This would make the design even more sensitive to bellows tipping under smaller frictional forces, since the bellows stiffness helps keep the end from tipping by resisting the applied moment from friction with the housing bore. The reduced stiffness of the MSL bellows combined with a surface finish in the housing bore that was at the coarse end of the tolerance and some tape adhesive residue all seemed to have pushed this unit over the edge and initiated the tipping of the end of the bellows. Once it had started, the bellows continued to tip and jam at the same location, even though the adhesive had been removed from the bore. While a stress analysis indicated that the tipped distortion of the bellows should not have resulted in any yielding, and

Document

AMS_2012.pdf

247 there was no visual indication of damage or permanent deformation, it seemed that the bellows had taken a set that led to the repeatable failure. Additional stroke tests were performed with different orientations to gravity that indicated that the bellows was slightly biased to fail in the original location. Design Change Options – Highly Limited The opportunities for changing the design were explored at this time and included a new bellows assembly with a greater stiffness, reducing the stroke requirement, providing improved guidance of the end of the bellows, and increasing the length of the housing to accommodate more volume for some bellows guide options. The options that were possible in the schedule time available included reducing the stroke requirement and adding a guide to the end of the bellows. A new bellows could not be designed and fabricated in time and increasing the housing length was not possible since the rover had structure that would interfere with a longer housing. The option of reducing the stroke requirement did not actually solve the fundamental problem and would have required the temperature range to be less than the current system design provided, making this option unacceptable. The remaining option of improving the design of the guide on the end of the bellows was pursued. The length of the outside of the accumulator housing could not be changed, but the inside of the housing had an option for moving the end-of-stroke hard stop 25 mm closer to the end of the housing. This did not affect the outer profile, making it acceptable in the tight volume the accumulator occupied within the rover chassis. While the housing could not be lengthened, the end-of-stroke hardstop inside of the housing could be moved closer to the end of the housing. This provided 25mm of additional interior length in which to implement the improved guide design. What Guide Material and Length-to-Diameter Ratio is Acceptable? The additional guide length that could be obtained without changing the outside length of the housing was 25 mm and the diameter of the bore of the accumulator was 111 mm. For a piston type guide, the standard wisdom is to have a length to diameter ratio of 0. 5 to 2.0, with most ratios in typical applications near 1.0. The best that could be obtained within the constraints of the MSL application was 0.225. Additionally, the best material choice would be one with the following characteristics:  Low friction with the stainless steel housing  Minimal to no particle generation over the life  A life cycle capability of 10,000 cycles  A coefficient of thermal expansion (CTE) that is close to the bellows and housing values The best choices for low friction involved polymers and the best CTE choices were metals. The choices from a particle generation and cycle life perspective crossed between polymers and metals. Lubrication of the metals was considered, but no lubricants had been qualified with the nitrogen gas fill valve. If the gas fill valve leaked at all, the system would fail due to loss of pressure on the Freon R-11. The leak rate of the fill valve had been verified with everything clean, and there was not enough time available to repeat the valve qualification with a lubricant, liquid or dry. This eliminated the use of an applied lubricant as a possible solution to the failure mechanism. The analysis of the self-energizing jamming phenomenon showed that friction and initial tip angle are critical to a successful design. A larger radial clearance between the guide and the housing bore would allow the guide to rotate more before contacting the housing wall. The larger the initial contact angle is, the smaller the required friction to create a self-energizing jam. This demonstrates that smaller radial clearances are preferred. In addition, the CTE difference between the guide and the housing had to be considered. These properties determined the smallest radial clearance that could be selected for a particular guide material choice.

Document

AMS_2012.pdf

248 In order to answer the question of minimum acceptable guide length and material choice, manual testing was performed using the test assembly shown in Figure 12. The plates at the end of the assembly determined the radial clearance to the housing bore and the plates in the middle region controlled the effective length of the test guide. Figure 12. Test Guide Assembly - Used to qualitatively determine the best length and radial clearance combination using Delrin ® disks The test guide assembly was inserted into the test housing and the operator used the mandrel diameter to stroke the assembly within the bore. As the test guide was moving, the operator applied a moment to the assembly with the thumb and index finger. As moment was applied, some combinations of length and radial clearance would solidly jam, some would tip over freely, and others would get tight in the bore but not be self-energizing. The results of the testing are listed in Table 1. Table 1. Results of Qualitative Manual Testing of Test Guide Assembly as a function of radial clearance and guide length using Delrin ® disks Clearance, Ø [mm] Length [mm] 0.025 0.127 0.254 0.381 0.508 0.635 6.35 FAIL - FLIP FLIP FLIP FLIP FLIP 12.7 FAIL FAIL - FAIL - FAIL- FAIL/FLIP FLIP 19.1 FAIL FAIL - FAIL FAIL FAIL FAIL- 25.4 FAIL FAIL - FAIL FAIL+ FAIL+ FAIL 31.8 FAIL FAIL FAIL+ FAIL + FAIL+ FAIL+ 38.1 PASS - PASS - PASS PASS PASS PASS - 44.5 - - - - - PASS FLIP = Piston could freely flip through 360 degrees, i.e. provided no tip restraint FAIL - = Jammed badly, i.e. became stuck with little or no effort FAIL = Would jam, but only when a moment was applied or was pushed at edge FAIL + = Would jam, but took a lot of effort. Would not recover on its own PASS - = Judged not to jam, but considered to be closer to jamming than others PASS = Could not be made to jam

Document

AMS_2012.pdf

249 The manual testing indicated that a 38.1 mm minimum guide length was required; no length/clearance combination less than 38.1 mm was successful. However, the maximum possible length of the guide that could be implemented with the housing and structural constraints was 25 mm. The test results demonstrated that a Delrin ® solution was not acceptable. Since the longest possible guide length was 25 mm, a new set of tests were implemented with different guide materials and two different guide designs. The first test guide design was a solid cylindrical guide that was 25 mm long on a stainless steel mandrel. The second design was the same cylindrical guide with a mandrel and a slotted guide, with the slot in the circumference of the guide, similar to automobile engine piston ring designs. The slot allowed the CTE mismatch between the guide and housing to be accommodated in the expansion/contraction of the slot, rather than in the radial clearance between the guide and housing. This was crucial for high CTE materials, since a solid guide design would have required the guide to actually be smaller in diameter than the bellows themselves. The slotted guide was also placed on a stainless steel mandrel to provide support for the bore of the guide. Figure 13 shows the virgin Teflon ® version of the guide. Figure 13. Solid and Slotted designs of the guide on Test Mandrel and inside the Test Housing with the largest possible radial clearance. Note that the slotted design springs outward, closing the radial clearance under all conditions. (the terms Solid and Slotted refer to the circumference of the guide) The different materials tested using circumferentially solid and slotted guides on a mandrel were  Torlon ® 5030  Vespel® SP-3  Delrin® 100 AF  Teflon® 25% glass filled  Virgin Teflon® The testing was done manually with the same pass/fail criteria as was used in the previous Delrin ® guide testing (see Table 1). The test guides were fabricated to diameters representing a nominal radial clearance at 25°C and the maximum radial clearance associated with the lowest temperature condition and the smallest manufactured outside diameter (Least Material Condition, or LMC). The results of this testing are listed in Table 2. Note that a solid Delrin ® version was not tested. This was the version tested previously and represented in Table 1. For the Torlon® and Vespel®, the slotted versions were not tested

Document

AMS_2012.pdf

250 because the CTE of these materials did not require them to be slotted and the solid versions failed. Both Teflon® versions were successful, except that the solid version with virgin Teflon® had too high a CTE and lost all guiding function at cold temperature and least material condition. Table 2. Results of Qualitative Manual Testing of Test Guides using circumferentially slotted and solid test items as a function of radial clearance [Nominal and Least Material Condition Cold (LMC Cold)], and material The test results demonstrated that both glass filled and virgin Teflon ® were acceptable candidates, though only a slotted design was acceptable with the virgin Teflon®. Additionally, the glass filled Teflon® showed signs of abrading the bore of the housing by capturing fine particles of housing material within and on the wear surfaces of the guide. This abrasion was deemed undesirable and the virgin Teflon ® slotted design was chosen to proceed forward. The final design of the guide and hard stop for the flight assembly is shown in Figure 14 along with a representation of the original design, emphasizing the changes made to mitigate the demonstrated failure mode. Figure 14. Original and Final designs of the end of the bellows assembly. The figure shows the bellows stroked to the maximum fluid condition and against the end-of-stroke hard stop for both design versions Material CTE Solid Slotted [ppm/C] nominal LMC Cold nominal LMC Cold Torlon 5030 glass fill 16 FAIL Vespel SP-3 MoS2 52 FAIL Delrin 100AF Teflon fill 100 FAIL Teflon 25% GF glass fill 100 PASS PASS PASS PASS Teflon, virgin 151 PASS FLIP PASS PASS Old Hard Stop New Hard Stop added guide length

Document

AMS_2012.pdf

251 Acceptable to Fly? The final design of the bellows end guide was selected based on intuitive “feel” and engineering judgment. These are not exactly verifiable quantities that make an assembly ready to fly on a mission. The next challenge was to demonstrate, using quantifiable methods under worst case conditions, that the new design would not exhibit the previous functional limitation or any new failures. This was particularly difficult since the landing and launch environments present significant loading to the assembly, but they were not representative of the long life mission loads while the bellows is actually stroking within the bore. However, if the large loading of launch or landing were to initiate a self-energizing condition, then they were appropriate and representative to use for testing. Upon studying the types of loading the unit would be subjected to during all of the phases of the mission, the worst case conditions consisted of combined radial and moment loading at the end of the bellows. The magnitude of the moment loading would be dependent on the stroke position of the bellows, which is a function of the temperature around the internals of the rover. In order to cover the conditions on the surface of Mars for the majority of the mission as well as the shorter term higher loading conditions, a combination of moment and radial loading conditions were formulated. It turned out that the maximum radial load could not be achieved along with the pr oper moment load due to the dimensional constraints of applying the loads to the bellows inside of the housing. Since the moment loading was determined to be the most detrimental to the operation of the accumulator, the moment loading was matched and the radial loading was allowed to be undersized. The test loads were generated by attaching steel plates to the end of the bellows assembly and cantilevering them far enough to produce the required moment loading and best possible radial load. This was not possible to do with a closed and sealed accumulator, so it was performed with the nitrogen gas end cap removed from the end of the accumulator housing. Figure 15 shows the test setup. The test loads were applied as shown above and the bellows stroked using internal pressure. The maximum allowable safe pressure difference across the bellows assembly was 38 kPa. The pressure difference across the convolutions that moved the guided end of the bellows in the housing was measured and recorded. There was no nitrogen gas pressure applied to bellows during the testing. This meant that the only force available to compress the bellows was its own internal spring force, which was very low. The results of this testing are shown in Table 3. Note that for the higher moment loads, the bellows did not retract on its own due to the friction force. This would not occur in a sealed and pressurized accumulator due to the presence of the nitrogen gas on the end of the bellows, forcing it to retract. Conclusions and Lessons Learned A bellows assembly without end guiding that must carry moment loads is a marginal design at best. A design without adequate end guiding relies on the internal moment stiffness of the bellows itself to prevent a significant rotation of the free end. If the radial clearance and friction conditions are beyond certain limits, then the rotation of the end of the bellows can quickly result in a self-energizing jamming condition that cannot be overcome. Figure 15. Stroke Testing of the final guide design under high loading conditions (load is equivalent to 3.2-g radial and 5.4 times the maximum flight moment)

Document

AMS_2012.pdf

252 With a piston guide of reasonable length, friction and the allowable radial clearance are extremely important design parameters because this mechanical configuration is very sensitive to these variables. The best option, where possible, is to use a lubricant to control the sliding friction. Otherwise, the lubrication function must be provided by careful material selections of the moving components. Polymers like Teflon ® are great, but their CTE is large compared to metals and the applied contact pressure between the sliding surfaces must be limited. The slotted guide presented here worked exceptionally well to accommodate the large CTE difference. Table 3. Results from the load testing of the final guide design (all bellows pressure values were well below the 38 kPa limit) References 1. Bhandari, P., Birur, G., Pauken, M., Paris, A., Novak, K., Prina, M., Ramirez, B., and Bame, D., “Mars Science Laboratory Thermal Control Architecture,” SAE 2005-01-2828, 35th International Conference on Environmental Systems, Rome, Italy, July 2005. 2. Bhandari, P., Birur, G.C., et al, “Mechanically Pumped Fluid Loop Heat Rejection & Recovery Systems For Thermal Control on Martian Surface – Case Study of The Mars Science Laboratory,” 36th International Conference on Environmental Systems, Norfolk, Virginia, July 2006. 3. Flexial Corporation provided experienced advice on the details of bellows analysis methods within the confined space of the accumulator housing. Flexial Corporation, 1483 Gould Drive, Cookeville, TN, 38502-3105, USA, www.flexial.com Acknowledgements This work was performed at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not constitute or imply its endorsement by the United States Government or the Jet Propulsion Laboratory, Pasadena, California. Copyright 2012 California Institute of Technology. Government sponsorship acknowledged. The authors wish to acknowledge the many individuals working on the MSL project, of which the thermal subsystem is a part of the greater whole. Dave Bame and AJ Mastropietro from the MSL Thermal team provided excellent support for the testing of the bellows at JPL. Pacific Design Technology (PDT) in Goleta, CA is acknowledged for their excellent work in developing the pump assemblies and providing significant support for this investigation. N u m b e r o f t e s t p l a t e s 0 1 42 22 93 7 Added mass [kg] 0.000 1.540 2.368 3.093 3.921 Total mass [kg] 0.680 2.220 3.048 3.773 4.601 Total applied load [N] 6.671 21.778 29.901 37.013 45.136 Cantilever length [m] N/A 0.069 0.065 0.072 0.096 Applied moment [Nm] 0.000 1.022 1.492 2.160 3.647 Equivalent Radial load [g's] 0.5 1.6 2.1 2.6 3.2 Equivalent Moment load (x Flight) 0.0 1.5 2.2 3.2 5.4 Bellows Pressure, kPa gage Guide started to slide, extending 0.34 2.4 5.2 9.3 7.9 Guide is halfway, extending 1.4 4.5 8.3 10.7 13.4 Guide is almost at hardstop, extending 3.1 7.2 10.3 13.4 16.9 Guide started to slide, contracting 2.4 0.34 † † † Guide is halfway, contracting 1.4 † † † † † the drag was high enough to prevent the bellows from contracting on its own

Document

AMS_2012.pdf

253 Development of the Descent Brake Mechanism for the Mars Science Laboratory David Dowen*, Jeff Moser** and Jeff Mobley** Abstract This paper will describe the design and testing of an electromechanical damper assembly for the Mars Science Laboratory as well as provide the current program status. Unique test equipment was developed for verification of flight requirements and is presented. Included in this paper are the problems that arose during design and testing that are peculiar to a device intended to dissipate energy. Also discussed are the lessons learned relating to assembly and test anomalies and the resulting corrective actions. Introduction The primary mission of the Mars Surveyor Lander (MSL) program is to deliver the 900-kg Curiosity rover to the surface of Mars in the year 2012. Its “Skycrane” landing system uses a free-flying rocket-decelerated descent stage to set Curiosity down on the surface, which hangs on a 7.5-m-long bridle cord, without ever landing the descent stage or fully powering down the engines. This architecture enables safe delivery of Curiosity onto rugged terrain. A key component of the Skycrane landing system is the device which allows the rover to be lowered by a distance of 7.0 m to fully extend its bridle. The Descent Brake allows this deployment to happen quickly and with controlled speed that decreases toward the end of deployment for a soft stop at the full bridle length. The amount of energy dissipated by the Descent Brake during the event is approximately 24 kJ. Descent speed control is achieved passively by deploying the bridle cord from a tapered spool. The tapered spool allows an ending descent speed that is less than the average speed. The radius from which the cord deploys is decreased as the cord is deployed. Initially the vertical velocity of the rover is high relative to the angular rate at the input shaft. As the bridle deploys, the diameter of the spool decreases and the angular rate, at the shaft, increases relative to the rover’s vertical rate. For this design, the torque is initially low at the Descent Brake, and then rapidly increases to slow the vertical rate of deployment, reference Figure 2. The design chosen to manage the drop converts the rover’s potential energy into heat with an electrical generator, driven by the bridles wrapped on the spool and dissipating the rover's energy into a bank of resistors. This technique has the advantage of being completely passive, and unlike previous friction technologies, offers highly consistent performance over a wide temperature range. 1 * Sierra Nevada Corporation, Space Systems Group, Louisville, CO ** Sierra Nevada Corporation, Space Systems Group, Durham, NC Proceedings of the 41st Aerospace Mechanisms Symposium, Jet Propulsion Laboratory, May 16-18, 2012 Figure 1 - Sky Crane Landing Sequence

Document

AMS_2012.pdf

254 Figure 3 - MSL Deployment Mechanism Descent Brake Requirements Damping Performance The initial nominal performance specification for the MSL Descent Brake required a nominal linear speed dependent drag of 10.6 ± 2.3 N•m•s/rad over a temperature range of -35°C to +35°C. Figure 4 shows the performance envelope allowed as it relates to shaft speed and torque. The performance of the Descent Brake design needed to remain in the performance envelope over all variations in manufacturing tolerances and environments. Descent Brake Spool Device Confluence Point Bridle Guide Umbilical Device 3X Bridle Exit Guide Figure 2 - Output from MSL Lowering Device Simulation1

Document

AMS_2012.pdf

255 Figure 4 - Descent Brake Performance Requirement Design Limits Several other requirements key to the design were specified as design limits. These requirements from the initial specification were:  Output shaft peak speed of 28 rad/s  Instantaneous output shaft peak torque of 340 N•m  Event duration of 7 seconds maximum  Power dissipation capability of 9.5 kW  Approximate envelope of Ø265 mm (Ø10.43 in) maximum outer diameter by 230 mm (9.05 in) maximum length (circular form factor) Descent Brake Design The initial step in the system design of the Descent Brake was to evaluate the driving requirements and establish design goals. The peak power requirement of 9.5 kW drove the design to require the majority of the power to be dissipated by a resistor bank. To ensure consistent performance with variation in temperature the resistor bank was designed around high reliability wire wound resistors with very low change in resistance over temperature (~±20 ppm/°C). In order to minimize the generator winding copper losses (I 2R losses) it was desired that the generator winding current be minimized. To minimize the generator winding current, the generator voltage was designed to be as high as practical. In order to avoid corona issues, the maximum generator design voltage amplitude of 150 V peak was chosen. To meet the desired performance requirement of nearly linear damping over the speed range, it was critical that the inductive reactance component of the generator winding impedance be minimized. The inductive reactance component of the generator impedance is a function of the generator winding

Document

AMS_2012.pdf

256 inductance, the number of generator poles selected and the shaft speed. The shaft speed was determined from the gear ratio selected. The gear ratio was selected to be as high as practical to balance the generator and gearbox sizes while still surviving the specified load conditions with margin. Late in the design process several of the design limits changed as the Curiosity rover design matured. The damping ratio was modified to 9.45 ± 0.15 N•m•s/rad. This change in damping ratio resulted in the peak momentary torque rising to 583 N•m and independently the peak speed to 36 rad/s. The Descent Brake design was evaluated for performance to these new requirements and fortunately was able to meet them by analysis except for the new maximum speed requirement. Operating at the new maximum speed requirement drove the rotor speed to over 1750 rad/s. This new rotor speed was 20% over the manufacturer’s recommendations for the rotor bearings. A test rotor with bearings was used to show the bearings were capable of withstanding operation at the maximum speed requirements with no issues. The rotor bearing set was tested to speeds that equate to 40 rad/s at the gearhead input shaft. The bearings were inspected at SNC and by the bearing manufacturer after completion of the testing with no detrimental affects noted. Descent Brake Components Generator From the established generator parameters, the generator design space was evaluated. Several generator designs of various geometry and number of poles were created. Each design was optimized to minimize inductive reactance while meeting all other parameter requirements. A Simulink model of the Descent Brake system was also created. The Descent Brake performance was evaluated for each generator design to determine the best generator configuration to meet the design requirements stated above using the system model. The generator configuration that ended up with the lowest inductive reactance was a 6 pole generator with a 1.1:1 length to diameter ratio. The chosen design showed very near linear operation of the generator damping over the required speed range. 6X Resistor Boards Jumper Board Output Spline Interface Gearhead Generator Figure 5 - MSL Descent Brake Assembly

Document

AMS_2012.pdf

257 Gearbox The gearhead is a high torque density, 3 stage planetary gearbox with an overall gear ratio of 48.8:1. The gearbox mass is 3.73 kg within an envelope of 98.55-mm diameter by 91.44-mm length (excluding flanges and shaft extension). Gearhead capacity was optimized using the gearbox design guidelines developed through the NASA Phase II SBIR Lightweight Gearbox Technology Program 2, including material selection, design features, and analysis techniques. The gearbox was designed for an operating high torque load case (322.5 N•m at 25 rad/sec), an operating high speed load case (298.8 N•m at 36 rad/sec), a momentary load case (583 N•m at 18 rad/sec) and a static peak torque case (700 N•m). The required life of the gearbox is only 8 cycles of 120 radians (960 radians total) for the operating load cases with the momentary load occurring no more than twice per cycle. With the relatively short life requirement, the gearbox was designed to provide positive margin for the specific loads and duration specified rather than being designed for endurance limit. Because of the high loads and speeds involved, significant attention was paid to imbalance of loads between the planets and the resultant net load applied to the supporting bearings. At the high speeds required, the centrifugal force on the planet bearings was factored into the analysis in addition to the resultant gear forces. Thermal impacts upon the lubricant during high speed operation were also considered. Figure 6 - 3 Phase Generator Stator Assembly Figure 7 - First, Second and Third Stage Carrier Assemblies

Document

AMS_2012.pdf

258 Resistor Bank/Structure The design of the Resistor Bank for the Descent Brake was a challenge in packaging. The Resistor Bank consists of three 1.5-ohm phase resistances wired in a wye configuration. Each phase consisted of two Printed Wiring Boards or PWBs containing 25 resistors on each board. The resistors were set in series/parallel configuration to result in the required phase resistance. This number of resistors was also utilized to ensure the maximum power dissipated in any resistor never exceeded 10 watts. An additional PWB was used to make all the connections between the Resistor Bank and Generator Assemblies. The resistors were arranged in a radial pattern on the donut shaped PWBs due to the circular form factor of the required envelope. The PWBs were mounted onto aluminum chassis components that were optimized for weight under the required structural loads. Individual resistors of different values could be chosen to tune the damping of the assemblies as required to meet the desired requirement in the range of approximately 7.0 N•m•s/rad to 12.0 N•m•s/rad. Figure 8 - Resistor Bank Assembly Test Program/Equipment Functional testing of the Descent Brake can be thought of as essentially dropping the equivalent of a Mini Cooper from the roof of a second story building on a spooled cable system. Although the requirements could be met with this type of test setup it was not ideal for many reasons. The functional testing needed to be performed in many different configurations including in a thermal vacuum chamber. The test method chosen was to use a brushless DC servo system to supply the required torque at the Descent Brake input shaft and a reaction torque cell to measure reacted torque at the mounting flange. The DC servo was sized to be able to meet all torque requirements without any gearing to keep sources of error at a minimum. This type of test setup allowed for a flexible test rig that could be adapted to many different scenarios. The DC servo system also allowed for flexibility in test profiles with only changes in programming. Figure 9 shows three of the test setups used for in-process, acceptance and qualification testing. The upper left photo shows the test configuration used for component level verification and resistor bank tuning. An additional torque cell is added between the generator rotor shaft and input to the gear head. This allowed for easy measurement of gear head efficiency and generator performance in the system. The upper right photo shows the ambient test configuration for a fully assembled Descent Brake. This

Document

AMS_2012.pdf

259 setup was extremely robust and allowed for consistent test results from the Descent Brake. The bottom picture shows the test configuration for thermal vacuum testing. This setup is essentially the same as for ambient except the addition of a fluid coupling to pass torque through the thermal vacuum chamber wall. Although the fluid coupling does have some drag associated with it, the magnitude is small compared with the torque input so had little affect on the overall test results. The tests performed on the flight Descent Brake included the following:  Structural/Stiffness test  Initial ambient functional test  Vibration test  Ambient functional test  Thermal cycling  Functional testing at thermal extremes  Final functional test The flight unit completed all acceptance testing without issue. The measured damping rates for the functional tests performed were 9.51 N•m•s/rad during initial ambient, 9.42 N•m•s/rad during thermal vacuum hot extreme, 9.51 N•m•s/rad during thermal vacuum cold extreme and 9.36 N•m•s/rad during final ambient functional testing. The torque vs. speed curves for these tests are shown in Figure 10 and shows the consistency of operation under varying environmental conditions. Component Verification Configuration Ambient Test Configuration Thermal Vacuum Test Configuration Figure 9 - Performance test setups

Document

AMS_2012.pdf

260 The Qualification Unit completed the same tests as the Flight Unit with the addition of shock, more extreme temperature limits during thermal vacuum and life testing. All tests were completed with no issues. After completion of the life test the qualification unit was disassembled and all parts inspected. No issues were found during the inspection. Lessons Learned Collaboration and Robust Design Margins One of the biggest challenges on the program was reacting to changes in requirements throughout the program and in particularly after completion of the design phase of the program. At the point the last requirement changes occurred there was a lot of schedule pressure to complete the program to support launch date at that time. Fortunately, the gearbox and resistor bank were designed to take advantage of the required envelope and had adequate design margins. We were able to perform analysis to the new requirements and show that the gearhead and resistor were still compliant. The generator rotor bearings were the only components we could not show compliant to the higher limits via analysis. This verification was handled through empirical testing and turned out to be a nonissue. The highly collaborative environment cultivated between JPL and SNC throughout the program was instrumental in working through the requirement changes effectively. This type of relationship is a necessity in order to effectively work through a program with dynamic requirements. Robust Test Set Besides the design of the deliverable hardware, the test set was also a challenge. The size of the Curiosity rover drove torque and speed test requirements outside our normal range. Using the large direct drive DC servo system worked extremely well. The combination of size and flexibility allowed us to perform many different tests in essentially the same test setup. Development Test Anomaly The generator rotor is composed of three main parts, a stainless steel rotor hub, permanent magnets and a thin metal rotor band. The permanent magnets are bonded to the stainless steel rotor hub and the rotor band is thermal fitted over the outside diameter of the permanent magnets. The function of the rotor band is to provide secondary mechanical retention and to protect the magnets within the air gap between the rotor and stator. Under nominal conditions there should normally be 0.25 mm (0.010 in) clearance, typical for this type of device, between the rotor band outside diameter and the stator inside diameter. During Figure 10 - Flight Acceptance Performance Curves

Document

AMS_2012.pdf

261 initial testing of the development generator, abnormalities in the generator torque were noticed during inprocess run-in. The generator was disassembled and it was found that the rotor band had been rubbing the generator stator inner diameter, reference Figure 11. In order to determine the cause of the rubbing, all the dimensions were verified and a rotor deflection analysis was performed. The result of the analysis was that everything appeared correct. The only abnormality noticed in the test data was that the no load generator torque measurements indicated higher than expected torque. No load generator torque measurements are an indication of the losses in the generator magnetic core and bearings. From the higher than expected generator torque, it was determined that approximately 50 W of power was being generated in the rotor band at the higher speeds due to varying magnetic field. In a normal generator design, the magnetic field in the rotor band does not vary to the extent that any significant losses are generated in the rotor band. However, the optimization done to the generator stator design, in order to minimize inductive reactance, created larger than normal magnetic field variation in the rotor band. The rotor band is a thin metal band which is primarily in direct contact with the rotor magnets. The rotor magnets are Samarium Cobalt, which has relatively low thermal conductivity. Due to the relatively low thermal conductivity of the magnets, the 50 W of power generated in the rotor band caused the rotor band temperature to increase. Due to this, the band expanded significantly which caused it to rub the generator stator inside diameter. Normal operation duration for the Descent Brake is seven seconds, which does not generate enough heat to cause the interference. This issue did not show up until the run-in was performed for 1 hour in each direction at 6 rad/s. Since the rotor band is not exposed to significant varying magnetic field in a normal generator design, it is typically made from a material that has appropriate mechanical properties for the application. In this case, the material initially chosen was beryllium copper. Power losses due to magnetic fields are a function of the electrical conductivity of a material. Beryllium copper has an electrical conductivity of approximately 20% that of pure copper. To solve the band power dissipation issue, the beryllium copper rotor band was replaced with a one made from Inconel. Inconel has an electrical conductivity of approximately 2% that of pure copper. The Inconel band resolved the power dissipation issue. Development Rotor with BeCu Band Development Rotor After Run-In Development Rotor with Inconel Band Development Stator After Run-In Figure 11 - Rotor Development Issue

Document

AMS_2012.pdf

262 Although the operation time of the Descent Brake is relatively short, understanding all the environments the device is going to be used in is critical to achieving a robust design. Commercial Components Another issue encountered in the manufacturing process was the use of commercial needle roller bearings. Roller bearings were implemented in the design to support the planet gears on the second and third stage carrier assemblies due to the high radial load and speed requirements at these locations. These bearings were procured from a commercial vendor using their standard materials. The 440C Stainless Steel rollers were fine, but the carbon steel cages had corrosion issues even though they were coated with a proprietary silver plating. Strict corrosion prevention process had to be implemented to ensure parts with corrosion did not end up in the assembly. Conclusions Ultimately, the program achieved successful completi on of validation testing of the Descent Brake design for the qualification unit. High level collaboration between JPL and SNC led to the successful assembly and test of the flight unit. The flight unit has been integrated into the rover and completed system level testing. This included a system level drop test in January, 2011. The total drop time during this test was within 0.1 second of the predicted time. The Curiosity rover is currently in transit to Mars, it was launched November 26, 2011, and is scheduled to land on Mars on August 6, 2012. Acknowledgments This work was funded by JPL under Subcontract 1293381. The authors wish to express appreciation to Ted Iskenderian, JPL Mechanisms Engineering Manager, for his work on the Descent Brake and their assistance in the production of this paper. BUD and Curiosity images courtesy of JPL. References 1. Gradzial, M.J. and Holgerson, K.J. “Mechanisms for Lowering Tethered Payloads: Lessons Learned from the Mars Exploration Program” IEEE Aerospace Conference Paper No. 1030, Big Sky, MT, March 2008 2. Mobley, J. “D21507 Final Report: Lightweight Gearbox Technology Program Phase II SBIR” SNC

Document

AMS_2012.pdf

263 Mars Science Laboratory Differential Rest raint: The Devil is in the Details Elizabeth Jordan* Abstract The Differential Restraint, a mechanism used on the Mars Science Laboratory (MSL) rover to maintain symmetry of the mobility system during the launch, cruise, and entry descent and landing phases of the MSL mission, completed nearly three full design cycles before a finalized successful design was achieved. This paper address the lessons learned through these design cycles, including three major design elements that can easily be overlooked during the design process, including, tolerance stack contribution to load path, the possibility of Martian dirt as a failure mode, and the effects of material properties at temperature extremes. Introduction The Differential Mechanism, a series of linked, passive pivots, is a component of the Mars Science Laboratory’s rocker-bogie mobility design. The differential acts as a motion reverser for the suspension system. The Differential Restraint was designed to prevent the rotation of the Center Differential Pivot (see Figure 1) of the Mars Science Laboratory mobility system during the launch, cruise, and Entry Descent and Landing (EDL) phases of the MSL mission. The Mars Science Laboratory is JPL’s next generation Mars Rover, known to the public as “Curiosity”. The mission launched on November 26, 2011 on an ATLAS V rocket and is scheduled to land on Mars on August 5 th, 2012. The rover mobility system incorporates the heritage rocker-bogie suspension design which was invented at JPL for the first Mars rover missions. The large size of the rover, approximately nine feet in length, requires the mobility system to be folded during the cruise stage of the mission. During the Skycrane phase, in which the rover is lowered to the ground via bridles and a descent stage (see Figure 2) the rocker arms of the mobility system are released from their stowed positions and allowed to rotate about passive pivot joints until the four rockers latch in their ready-for-touchdown positions. The Differential Restraint is the last of the mobility restraints to be released, via a pyro, approximately 2 seconds before rover touchdown. * Jet Propulsion Laboratory, California Institute of Technology , Pasadena CA Proceedings of the 41st Aerospace Mechanisms Symposium, Jet Propulsion Laboratory, May 16-18, 2012 © 2012 California Institute of Technology. Government sponsorship acknowledged.

Document

AMS_2012.pdf

264 Figure 1: Key Components of the MSL Mobility Differential Suspension System Figure 2: MSL Entry Descent and Landing Overview Diagram

Document

AMS_2012.pdf

265 Differential Restraint Design Development and Description The Differential Restraint is a passive mechanism which relies on the compression of a series of Belleville Springs to limit the degree of rotational motion allowed at the Rover Mobility Center Differential Pivot during the Cruise and EDL phases, as well as to absorb and dampen the energy associated with the violent mobility deploy and latch events. Since the rover mobility system is also the vehicle’s landing gear it is critical that the 6 wheels are in, what is termed, their “ready-for-touchdown” state, or as close to 6-wheels flat as possible when the vehicle impacts the Martian surface. The Differential Restraint enables this goal by limiting the rotation of the passive center pivot joint, known as the Center Differential Pivot (CDP), to a small angle, keeping the two sides of mobility, port and starboard, nearly balanced. Once the chassis dynamics, namely, chassis pitch and roll from the effects of the mobility deploy and latch events, have settled out the Differential Restraint is released, via a pin-puller pyro just prior to rover touchdown. Design History The driving load case for the Differential Restraint is derived from the near-simultaneous deployment (and end of travel latch events) of the mobility rockers during the Skycrane phase of EDL. In the event of actual simultaneous latch events of the port and starboard sides, large shear loads are expected at the CDP. Alternatively, if the deployment and latch events are staggered from port to starboard large moments are created about the CDP, which the Differential Restraint must absorb. The original Differential Restraint design was a rigidly pinned interface between the externally mounted mobility Differential and the top deck of the rover chassis which prevented nearly all motion, except for the small amount of slop between the pin puller pin and interfacing monoball (see Figure 3). Approximately one year after the mobility Critical Design Review the increasing maturity of the ADAMS dynamic model produced a drastic increase in the predicted moment and shear loads at the CDP. This had the effect of creating significant negative margins in the original piece parts of the Differential Restraint; including the AerMet pin of the 3/8” Pyro, the shear pins and bolted joint which interfaced to the Rover top deck, as well as the Rover top deck itself. In late 2008 it was concluded that a completely new restraint design was required. The new design was to be analyzed, built, tested and integrated before the (then) planned 2009 launch. In addition to the tight timeline constraints, the re-design effort was further complicated by the fact that the hardware interfaces were already fixed, the available space was incredibly limited, and the 3/8” pin puller was the only feasible option for pyro devices.

Document

AMS_2012.pdf

266 Figure 3: Original, Rigidly Pinned, Differential Restraint Design Concept Since the rover top deck interfaces were already built, and the pyro device was already qualified and delivered the new design had to find a means by which the loads that the Center Differential Pivot would impart on these critical interfaces could be significantly decreased. The new design relied on the principle of applying a (theoretical) torsion spring at the center of the Differential Pivot which would absorb the large impact energy. After running idealized dynamics simulations, a spring rate of approximately 200,000 N-m/rad was selected as the stiffness that would reduce the impact loads enough to prevent negative margins in mobility and chassis hardware but still restrain the mobility system from excessive motion during launch, cruise and the initial stages of EDL enough to be able to ensure that the mobility system was in its “ready-for-touchdown” state in time for landing. The new restraint was designed to accommodate the worst case combined loading event as predicted by a 2000-run Monte Carlo ADAMS dynamic model of the MSL EDL system. Implementing a torsion spring at the Center Differential Pivot (CDP) was impossible due to packaging and size issues so it was decided that the Differential Restraint would use a 3-bar crank-slider approach to turn rotation into linear actuation (see Figure 4). Combinations of large Belleville washers in series and parallel were implemented to achieve the extremely high spring rate that was required in the limited space available. A rod, known as the Spring Plunger, wa s used to actuate the two stacks of Bellevilles. One stack is compressed from positive moments while the opposite stack is compressed by negative moment loads at the CDP (see Figure 5). The Spring Plunger, made from Titanium, was threaded at one end to allow it to be joined to a Clevis, which in-turn was connected to a Linkage with a Monoball in each end (see Figure 6 and Figure 7).

Document

AMS_2012.pdf

267 Figure 4: Crank-Slider Diagram of Differential Restraint Figure 5: Belleville Spring Cross-Section Loaded and Unloaded States

Document

AMS_2012.pdf

268 Figure 6 : Cross Section of Differential Restraint Mechanism Figure 7: Differential Restraint Mechanism Nomenclature Overview Tolerance Stacks and Stress Analysis Test Failure With the delay of the MSL launch the new Differential Restraint hardware was built but the planned component level testing of the hardware was delayed in an effort to conserve funds. In July of 2009, the Differential Restraint was integrated to the mobility differential system (see Figure 1) on the Dynamic Test Model of the MSL rover, in preparation for the differential system static test. With the restraint in place, the test was designed to load the entire differential system and generate a large moment load about the Threaded Interface Clevis

Document

AMS_2012.pdf

269 Center Differential Pivot by pushing on an additional lever arm connected at the Main Differential Pivot Interface. At 90% of the full test load the measured load began to drop rapidly and an audible snap was heard. Upon investigation it became apparent that the Spring Plunger had catastrophically failed at the point of thread termination at the interface between the Spring Plunger and the Clevis (see Figure 8). Figure 8: Failed Differential Restraint (Left) Disassembled from Clevis (Right) Determining the Cause of Failure Since the component level tests of the Differential Restraint had been skipped a wide array of failure causes existed that needed to be examined. Linear sliding mechanisms are often avoided at JPL on principle. This fact made the theory that one of the sliding surfaces between the Spring Plunger, Internal Guides, and Bellevilles had jammed, thereby preventing actuation of the mechanism, a prime suspect. Another theory was that the Spring Plunger, which was made from Titanium, had been made by a fraudulent vendor, and that imperfections in the material caused the crack initiation. The lack of a thread relief callout on the Spring Plunger drawing was also considered as a possible cause of failure, as was the large bending load on the Spring Plunger. The leading failure theory, that the sliding surfaces had jammed and prevent motion, was disproven in two ways. First, a line of Braycote, the lubricant used inside the Belleville Spring Assembly, was left on the Spring Plunger from where it had entered the Belleville Spring Assembly. By measuring the distance between the grease marks left on the Spring Plunger, it was possible to approximate the distance traveled between the start of test and the time of failure to be 5.46 mm (0.215 inch). At the load applied at the time of failure the total compression of the Belleville Spring stack was predicted to be 5.59 mm (0.220 inch). The close proximity of these values was strong evidence that the Spring Plunger did in fact slide along the Internal Guides and compress the spring stack as intended. Furthermore, the shape of the load deflection curve measured during the test exactly matched the prediction (see Figure 9), including the inflection point where the stiffness of the Belleville Stacks drops by one-half due to only one stack being compressed after the initial motion during which both stacks are in the load path. If the mechanism had jammed the applied load versus displacement would have spiked significantly rather than following the predicted linear line before dropping off when the Spring Plunger yielded and eventually snapped. Failure Theories: Jammed sliding surfaces Fraudulent Titanium Improper thread termination on Spring Plunger Excessive bending load on Spring Plunger

Document

AMS_2012.pdf

270 Figure 9: Applied Load vs. Pull-arm Displacement measured during failed Static Test Upon inspection of the material certs for the Spring Plunger it was confirmed that the material was from Western Titanium. However, upon SEM inspection of the failed threads (see Figure 10) it was determined that the failure was ductile and there was no evidence that a material imperfection initiated the crack. Figure 10: 10kX magnification of Spring Plunger Fractured Surface Bending Loads The Spring Plunger was originally analyzed for a bending load which would be applied as the Center Differential Pivot rotated. Since the Differential Restraint mechanism was designed to limit the rotation of the Center Differential Pivot to ±2°, the total misalignment between the line of action of the Linkage and the Spring Plunger was calculated by assuming a worst case rotation of the CDP to be 5°. If the CDP were allowed to rotate 5 degrees the misalignment between the line of action of the Linkage and the Point of Failure Expected change in stiffness

Document

AMS_2012.pdf

271 Spring Plunger axis caused by this rotation would be 0.5°. With these assumptions the Spring Plunger was shown to have very small, but positive, margins of safety at the root of the threads. What was thought to be a very conservative analysis, (doubling the expected rotation of the CDP) was in fact not nearly conservative enough. Figure 11: Line of action misalignment due to rotation of Center Diff. Pivot In the post failure analysis the total angular misalignment between the Linkage force vector and the Spring Plunger was further scrutinized. It quickly became apparent that the misalignment caused by the rotation of the CDP was actually one of the smallest contributors to the overall misalignment. More significant misalignments were due to piece part misalignments due to imperfect shimming of the mechanism, the vertical slop of the Linkage in the mouth of the Clevis and Horizontal Swingarm Bracket, the slop of the Center Differential Pivot in its own bushings, and the elastic deformations of the Chassis Bracket and the Rover Chassis top deck. The maximum possible misalignment was calculated to be 2.2 degrees (see Figure 12), which was over four times what was initially accounted for in the Spring Plunger stress analysis. +X +Y

Document

AMS_2012.pdf

272 Figure 12: Differential Restraint Angular Misalignment Calculations Recovery From Failure Immediately following the failure of the Differential Restraint many senior engineers began taking a closer look at the restraint design. They highlighted the fact that the restraint poorly implemented the concept of a crank-slider because the point at which the crank grabbed the slider was not at the center of the slider but rather it was offset- the link between the Clevis and the Spring Plunger was at the end of the Spring Plunger rather than the middle of the Spring Plunger (see Figure 13). Many engineers advocated redesigning the restraint to properly implement the theory of the crank slider; however limited schedule and funds made a complete redesign impractical. Instead, the design concept and most of the original hardware was kept intact, but key components were changed from Ti-6Al-4V material to significantly higher strength steel. By remaking two components and modifying two additional ones it was possible to quickly and relatively cheaply make the restraint design work. ∆X ∆Z ∆RX ∆RZ [in.] [in] [ °] [°] 0.00458 0.02 0.02 0.0233 0.0094 0.020518 0.03211 0.022 0.013 0.0448 0.0031 0.0053 0.629 0.0822 0.0045 0.0290 0.0188 0.0442 0.629 0 0.0822 0.629 0 0.08220.07492.1 RSS X_Z 3.0 3.3 Clevis Elastic Deformation 3.4 Spring Plunger Elastic Deformation Angular Misalignment (1.0,2.2,3.0) 2.2154 0.71120.7112 Angular Misalignment (1.0,2.1,3.0) 2.06143.1 Chasis Bracket Elastic Deformation (8443 lbf) 1.2 HAS Shim 1.3 Chassis Bracket Shim 1.4 Monoball to HSA 0.02555 RSS (1.0,2.2,3.0) 0.08340.0084 3.2 Sum_X_Z 0.0335 RSS (1.0,2.1,3.0)Angular Misalignment (Include RSS where applicable) 3.2 Top Deck Elastic Deformation (8443 lbf)RSS1.0 1.0 RSS_X_Z 0.03811 3.1 Sum_X_Z1.5 Monoball to Clevis 2.02.1 CDP TranslationMisalignment Source 2.2 CDP Rotation about X/Y1.1 Sprng Stack Deflection (.215" )

Document

AMS_2012.pdf

273 Using the interaction formula: ௕ଷయൌ1 ݏݏ݁ݎݐݏ ݁ݐܽ݉݅ݐ݈ݑ/ ݈݀݁݅ݕ ݋ݐ ݏݏ݁ݎݐݏ ݁ݒ݅ݏݏ݁ݎ݌݉݋ܿ ݂݋ ݋݅ݐܽݎ ݏݏ݁ݎݐݏ ݁݉݅ݐ݈ݑ/݈݀݁݅ݕ ݋ݐ ݏݏ݁ݎݐݏ ݎ݄ܽ݁ݏ ݂݋ ݋݅ݐܽݎ ݏݏ݁ݎݐݏ ݁ݐܽ݉݅ݐ݈ݑ/ ݈݀݁݅ݕ ݋ݐ ݏݏ݁ݎݐݏ ܾ݃݊݅݀݊݁ ݂݋ ݋݅ݐܽݎ it was possible to predict the allowable side load, and hence the misalignment angle between the Spring Plunger and Clevis that a steel Spring Plunger could withstand, if the Flight Limit Load was applied axially. After calculating the misalignment angles that would produce yield and failure for a Spring Plunger made from steel it was decided that the values calculated were too close to the worst case possibilities predicted by the tolerance analysis to make simply changing the material a comfortable approach. Therefore, in addition to changing the material, the Spring Plunger diameter and thread diameter was increased. With the larger Spring Plunger diameter, and the material swap from Ti-6al-4v to steel, the new Spring Plunger would be able to withstand 4.24° and 4.29° misalignments before yielding and ultimately failing respectively. Additionally, a thread relief was added to the Spring Plunger thread termination (see Figure 14) to prevent excessive stress concentrations. Figure 13: Comparison of “proper” crank-slider design methodology vs. Differential Restraint Design Figure 14: Spring Plunger Thread Termination Callout Key Findings:  Tolerance stacks and shimming procedures affect load paths- Make sure to account for this in stress analysis  The location of pivot joints in kinematic mechanisms is extremely important- Failing to follow best practices early-on can cause unexpected difficulties down the road  Proper thread relief callouts on drawings is essential for limiting the effect of stress concentrations

Document

AMS_2012.pdf

274  Although not always the most elegant design approach- making a design “good enough” rather than ideal can be a more effective approach for saving time and money Lubrication and Vacuum Bake Since it was a concern that the properties of the lubricant used in the Differential Restraint, Braycote 601EF, could be altered by vacuum bake, the vacuum bake out, required for all MSL hardware to meet contamination and planetary protection requirements was performed prior to performing life and component level testing on the Belleville Spring st acks. Upon removi ng the Differential Restraint components from a 50 hour, 110°C vacuum bakeout an oily substance was noticed on the bottom surface of the mechanism. Upon chemical analysis of the substance it was determined that it was the lubricant Braycote 601EF. According to the Braycote 601EF product data sheet, the temperature range for Braycote 601EF is -80°C to 204°C, however the viscosity at these temperatures had been overlooked. At a temperature of 99°C Braycote 601EF has a viscosity of 45 cST, which is approximately midway between the room temperature viscosities of Honey (73.6 cST) and Olive Oil (24.1 cST). When viewed in this context it is not surprising that the Braycote flowed and leaked out of the assembly when baked at 110°C. The Differential Restraint, Belleville Spring Assemblies were weighed before and after entering vacuum bake. With the knowledge of the decrease in weight due to vacuum bake and the knowledge of the mass of Braycote originally in the assembly it was determined that the Engineering Model and Flight units lost between 10 and 12% of their original Braycote lubricant due to the leak. Despite the loss of lubricant the stiffnesses of the Belleville Spring stacks measured during thermal characterization testing were well within the range of predicted and acceptable values and it was deemed unnecessary to add replacement lubricant to the assemblies. By exposing the potential for Braycote to leak from the assembly the unfortunate leak that occurred during vacuum bake turned out to be a blessing in disguise, in that it focused attention on a failure mode that would have otherwise been missed. Due to the hardware’s proximity to critical lenses for imagers on the Rover, focus was shifted away from the concern over the Belleville spring rate which was vetted through testing, to how to prevent the Braycote from leaking during flight and possibly contaminating other sensitive hardware. The actuation of the Spring Plunger through the Belleville Spring Assembly and its ability to squeeze Braycote out of the Belleville Spring Assembly like toothpaste became a concern. This problem was solved by adding a cover over the end of the Belleville Spring Assembly to catch any Braycote squeeze out (see Figure 15). Kapton shields were epoxied over the witness holes in the Belleville Spring Assembly (see Figure 15) to close out those leak paths as well. With these new covers in place the venting ratios of the Belleville Spring assembly came under scrutiny. A balance was struck between closing vent paths and preventing the Kapton tape from being blown off during depressurization by poking pin holes in the Kapton covers. Key Findings:  Just because a material is advertised to work over a specified temperature range doesn’t mean it will work the same over the specified temperature range-make sure to understand all key material parameters over the expected temperature range to avoid surprises  Consider “Band-Aid” solutions carefully while fixing one problem to avoid creating another

Document

AMS_2012.pdf

275 Figure 15: Before (Left) and after addition of Braycote leak path closures (Right) Working in a Dirty World As part of the final preparations for launch and closeout of the Verification and Validation matrices, the testing performed on all of the Rover pyros was summarized in order to ensure the Rover had been properly and completely tested. Although the pin puller used in the Differential Restraint had been qualified and testing had been completed at the sub-system and system level tests over temperature one concern was raised that the pin puller had never been fired in a dirty environment. None of the rover release pyros had been tested in dirt because they are fired shortly after the rover enters the Mars atmosphere while the rover is still connected to the Descent Stage and approximately one-hundred feet above the surface. The Differential Restraint pin puller, however, is fired much later in the deploy sequence, when the rover is approximately 2 seconds from final touchdown. Not only is there dust expected in the atmosphere at the level of the firing of the Differential Restraint pin puller, but the Descent Stage rocket engines are also expected to kick up significant quantities of Mars dirt into the atmosphere. The tight clearances between the Horizontal Swingarm Bracket bore and the pin puller pin (nominal 0.0015” diametral clearance) created a concern that small particulates of sand/dust could get stuck between the two surfaces and jam the joint, thereby preventing retraction of the pin. Additionally, there was a concern that if debris were to enter the pin puller device itself the retraction of the pin could fail. Two tests were conducted to demonstrate the ability of the Differential Restraint Pin Puller to successfully retract in the presence of large quantities of dirt. Hardware which mimicked the Flight hardware clearances and tolerances was used to perform the test. This hardware had previously been used to qualify the pin puller for release while subjected to large lateral loads over temperature and varying percentages of NASA Standard Initiator charge. An acrylic chamber was built to sit around the Differential Restraint mock hardware and a GN2 line was setup with two inlets to the container to create a simulated dust storm. In total, 700 ml of dust simulant was added to the chamber, including 600 ml of cohesionless fine grain sand, which was used in traverse testing characterization for the MSL rover, 50 ml of JSC-1 lunar soil simulant and 50 ml of BP-1 dust (see Figure 16). GN2 was pumped into the chamber at 345 kPa (50 psi) for 30 seconds prior to firing the pin puller to create the simulated dust storm.

Document

AMS_2012.pdf

276 Figure 16: Dirty Pin Puller Test Setup: Pretest (Left), Posttest (Right) The first test involved a dual firing of 100% charged NSI while the second test involved only a single NSI charged to 100%. Both tests included the application of the flight limit lateral load applied to the pin puller pin through the use of a strain gaged bolt connected to the Linkage hardware simulator. Both tests resulted in successful retraction of the pin puller. Upon inspection of the pin puller pin, Linkage monoball, and Horizontal Swingarm Bracket bore after the test no unusual wear marks or scoring were found, indicating that the fine grains of dust and sand had not jammed or impeded the retraction of the pin. The monoball in the Linkage, did however, have a significant increase in drag due to the presence of dirt in the lubricant on the monoball increasing the rolling friction between the ball and race. The increased drag in the monoball is not significant, however, since it did not provide enough resistance to jam the pin and prevent it from retracting. Key Findings:  Consider operational environments during the design of hardware (ex. how particulate contaminants affect hardware clearances and failure modes)  Pin Pullers can successfully operate in the presence of fine grain particulates Summary The design of the Differential Restraint for the MSL mobility system involved striking a difficult balance between strict load and deflection requirements with the tight design space available. In total, three major design iterations were required to get the Differential Restraint from concept to the launch pad, and along the way some interesting lessons were learned. The first redesign effort was required when the loads model drastically increased the predicted maximum loads. The failure to consider the shimming procedure and misalignments allowed for by piece part clearances and tolerances was a costly mistake which led to the failure of the design in a major system level test causing the entire design to be subjected to major scrutiny and redesigned for a second time. The tolerances and clearances, when considered, allowed for over four times the assumed misalignment between the axis of the Spring Plunger and the Linkage which proportionally increased the bending stress in the Spring Plunger. The absence of a proper thread relief callout on the Spring Plunger drawing provided an easy target for scrutiny of the design, but was also easily corrected in the redesign efforts. The leaking of Braycote during vacuum bake should not have been a surprise, but the failure to assess the meaning of the viscosity of the lubricant at high temperatures led to the requirement of yet another design Band-Aid, the implementation of which would have neglected the effects of depressurization and rules of thumb for venting ratios except for a last minute catch of the newly created issue. The development of the Differential Restraint was not a linear or

Document

AMS_2012.pdf

277 elegant process, but important lessons about tolerance stacks, lubricant leak paths, and even the use of pyrotechnic devices in dirty environments were learned; the knowledge of which will hopefully inform the design of future mechanisms for space applications. Acknowledgements This research was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration.

Document

AMS_2012.pdf

279 Mars Science Laboratory’s Dust Removal Tool Kiel Davis*, Jason Herman*, Mike Maksymuk*, Jack Wilson*, Philip Chu*1, Kevin Burke**, Louise Jandura** and Kyle Brown** Abstract The Dust Removal Tool (DRT) is designed to expose the natural surfaces of Martian rocks obscured by layers of dust deposited by aeolian processes. The DRT, contained within a cylinder 154-mm long and 102-mm in diameter, has a mass of 925 grams. Using a single brushless DC motor, the DRT removes dust from an area 45 mm in diameter. During the dust removal process, a set of brushes articulate to maintain surface contact as they rotate at high speed. The DRT belongs to a special class of aerospace mechanisms designed to interact with unstructured extraterrestrial surface objects and environments. The wide range of rock surface characteristics along with severe resource constraints makes the DRT solution non-trivial. The mechanism features a high reduction single-stage planetary gear box and pivoting brushes that both offered lessons learned. The flight unit DRT was integrated with the MSL rover in early 2011 and is currently on track to begin surface operations at Mars’ Gale Crater in August 2012. Introduction The Dust Removal Tool (DRT) is a critical component of Mars Science Laboratory’s (MSL) Sample Acquisition, Sample Processing and Handling (SA/SPaH) subsystem. The aeolian-deposited reddish iron oxide dust that covers ever ything on the surface of Mars masks many characteristics of rocks and makes it difficult for scientists to identify optimal rock targets for further interrogation and possible sample acquisition. For instance, as shown in Figure 1, dust layers only microns deep can obscure many visual clues to a rock’s origin (e.g., color or emissivity , cracks and inclusions) from instruments like the MSL’s MastCam and Mars Hand Lens Imager [1]. MSL’s Alpha-particle X-ray spectrometer (APXS), a key source of information about a rock’s elemental composition, is effectively blinded by layers of dust as little as 5 microns deep [2]. It is therefore critically important to remove dust from the surface of rocks. Figure 1. Left: MER Spirit Pancam false-color image of Mazatzal on Sol 86 after numerous Rock Abrasion Tool brushing and grinding operations; Right: MER Spirit Microscopic Imager false-color image of Mazatzal on Sol 79 after a RAT brushing operation. Image credit: NASA/JPL-Caltech/Cornell * Honeybee Robotics Spacecraft Mechanisms Corp., New York, NY ** Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA Proceedings of the Aerospace Mechanisms Symposium, Jet Propulsion Laboratory, May 16-18, 2012

Document

AMS_2012.pdf

280 Planetary scientists may have first appreciated the dust’s complicating effect during the Mars Pathfinder mission where APXS data were contaminated by dust and meaningful inferences could only be drawn through a process that assumed an amount of dust that had to be subtracted from the spectra [1]. Unlike Mars Pathfinder, the Mars Exploration Mission rovers Spirit and Opportunity both carried a tool for doing away with the dust a rock’s surface. This tool, developed by Honeybee Robotics, is called the Rock Abrasion Tool (RAT) [4]. The RAT’s primary purpose is to remove the rock’s weathered rind by grinding an area 45 mm in diameter to a depth of 5 mm. By accomplishing its primary objective it naturally has to also remove surface dust. The RAT is equipped with brushes for sweeping away cuttings produced during the grinding process – see Figure 2. During the initial stages of RAT development, it was never envisioned or required that the RAT’s brushes be used to strictly brush away surface dust layers. But through experimentation before and during the MER mission, it was eventually found that the RAT’s high-speed brush, which protruded several millimeters beyond the grinding wheel’s reach, is particularly effective at removing even fine layers of dust from all manner of rocks including those with pitted, vesicular surface textures. The RAT can also accommodate a large degree of surface topography variation and robotic arm (IDD) positioning error due to its design architecture and 3 degrees of freedom. Prior to brushing or grinding, the IDD preloads the RAT against the surface of the rock via the RAT’s butterfly mechanism (see Figure 2). The RAT has three actuators including a Z-axis actuator which moves the grinding wheel and brushes linearly toward and away from the rock surface. Once preloaded, a RAT software algorithm employing all three actuators, detects the rock’s local surface position in the Z-axis reference frame. The RAT grind brush is then positioned with respect to the surface such that its bristles are engaged and the grinding wheel is not engaged. In this way, the arm positioning error is rendered more or less moot and successful brushing is achieved on a wide range of uneven surfaces. Figure 2. Left: MER Spirit RAT on Mars prior to its first operation on Adirondack rock; Right: MER Opportunity RAT on Mars after many operations. Image credit: NASA/JPL-Caltech/Cornell The original plan for MSL was to include a next generation RAT with all the same functionality as the MER RAT but designed to last longer and penetrate stronger rocks [5]. However, that tool was eventually descoped by the JPL flight project office in 2007 to meet project budget constraints. This left the MSL science team without a method for clearing the blinding dust from rock surfaces. So later in 2007 a new effort was initiated to develop a tool, the DRT, designed solely for the purposes of removing dust. Like other MSL mechanisms [6], the DRT is designed for very long life in very harsh conditions. But the DRT is unique and interesting for at least two reasons. First, as a device that directly, physically engages

Document

AMS_2012.pdf

281 an extraterrestrial object (rock) so as to manipulate this object’s characteristics in some way, the DRT belongs to a special class of aerospace mechanisms. In this context, it is interesting and informative to future missions to understand how the DRT design solves the difficult problem of adequately removing micron-scale particles from a rock’s uneven and often pitted surface. Second, from a pure mechanisms perspective, the DRT offers a valuable data point for planetary gear box designers along with other lessons. Due to extreme volume constraints, there was just enough room for a single stage planetary gear reduction necessary for meeting torque margin requirements. Practical limits of planetary stage reductions are considered 3:1 (lowest, planets become very small) to 10:1 (highest, sun becomes very small). A ratio closer to 5:1 is the most balanced with the highest performance rating and is therefore more common in space mechanisms designs. The DRT design incorporated a 10.4:1 single stage planetary gear reduction which pushes the limits of what is conventionally considered practical. This paper will provide an overview of the DRT including requirements, design, manufacturing and qualification testing. It will present some of the key trades made during the development process including brush articulation kinematics and bristle geometry to maximize the reachable workspace. A significant part of the paper will deal specifically with brush pivot torque margin problems, the planetary gear design and related test results. DRT Overview The DRT is mounted on the MSL robotic arm (RA) turret as shown in Figure 3. Once placed on a rock by the RA, the DRT’s primary functional requirement is to clear dust layers up to 2 mm deep from an area on the rock’s surface no less than 45 mm in diameter – this diameter is driven by the APXS field of view and RA positioning accuracy. Like the MER RAT, the DRT is expected to face all kinds of rocks with diverse surface textures and topographies. Unlike the MER RAT however, the DRT only has a single actuator and is not preloaded against the surface. The design must therefore accommodate robotic arm positioning error in addition to varying local rock surface topography. These new design and operational constraints were challenging and required a substantial departure from the heritage RAT brush design. Figure 3. Left: MSL rover; Right: SA/SPaH turret with DRT. Image credit NASA\ JPL-Caltech [6] To deploy the DRT, the RA first confirms the rock surface location using various contact sensors on the turret and then positions the DRT at a safe stand-off distance (not in contact with the rock). The DRT motor is then energized such that the bristles spin at relatively high speed and the RA then moves the DRT toward the rock surface to a position where the bristles are theoretically engaged. The DRT engagement position and angle may be off by as much as 10 mm and 15°, respectively, in any direction due to RA positioning error. The DRT motor current and encoder are monitored for a stall condition, but otherwise it is an open loop operation where the DRT motor is left energized for a period of time

Document

AMS_2012.pdf

282 (~60 seconds) to sweep the rock. The RA then pulls the DRT back away from the rock while the bristles are spinning at low speed so as not to inadvertently foul the cleared area with a sudden stop and dragging of the bristles (an undesirable effect observed during testing). The DRT motor is then de-energized while in free-space. In addition to the single actuator and RA operational constraints, the other major constraints included the allowable volume and mass for the DRT and the fact that the DRT had to employ a particular JPL-supplied motor model. The DRT was to fit in a cylindrical envelope approximately 134-mm long and 141 mm in diameter. The allowable mass was 950 grams which included the mass of the motor (350 grams). The DRT design was to assume that the motor could provide 28 mN-m of torque at 10,000 rpm. The DRT engineering challenge could then be boiled down to designing a set of brushes with the reach and compliance to meet the primary functional requirement (i.e., sweep a 45-mm diameter) across the range of surface topography and texture scenarios while not exceeding the tool volume, mass and motor limits. The resulting DRT fit within a cylindrical envelope 154-mm long and 102 mm in diameter and weighed 925 grams. The mechanism is comprised of a JPL-supplied motor integrated with a Honey bee custom-designed planetary gearbox which drove a sub-assembly called the brush block. The brush block consists of an asymmetric set of brushes each on spring loaded pivot (hinge) in order to accommodate the wide variation in surface geometry. A resistive strip heater is bonded to the outside of the planetary gearbox. A rotating post at the center of the brush block guards against overloading the tool’s brushes against the rock. Figure 4. Dust Removal Tool

Document

AMS_2012.pdf

283 Brush Block Subassembly The brush design process was heavily informed by a prototype testing program that lasted beyond the Critical Design Review. Rocks and dust analogs we re supplied to Honeybee by JPL. A test and measurement methodology was established which quantitatively assessed brush performance. Per its equipment specification, the DRT dust removal capability was to be demonstrated by clearing 70% of dust particles less than 500 microns by surface area from a natural surface. It was especially important to clear the center of the 45-mm diameter area. Images of the surface were captured via digital camera and analyzed in software to determine the size of the area cleared by the tool. Additionally, the DRT would be required to perform dust removal operations up to 150 times on Mars and therefore a 2x demonstration (300 operations) would need to be performed with the final brush life test model. So quantitative measurements of bristle wear and brush shape degradation were also made to project life. Initially the team considered several variants of the very simple “brush-on-a-stick” concept (left most image in Figure 5) but soon realized that bristle compliance alone was not enough to compensate for large surface height variations without massive increases in parasitic drag on the motor. Furthermore, relying on bristle compliance also worked the bristles much harder causing wear and flexing that shortened bristle life. So the notion of allowing the brush bristles to pivot about a spring loaded hinge was quickly adopted. The brushes can pivot up to 30° so that large surface height variations (10-20 mm) can be accommodated. By using a soft spring with a flat spring rate to keep the bristles loaded against the rock surface, the bristle contact force and resulting motor torque changes very little. Figure 5. Brush design evolution Figure 6. Matrix showing the major brush related design/ performance drivers 45 mm Diameter Circle> 70% Clean Center ClearingAvoid Bristle EntanglementVolume Envelope Restrictions< 50 N Reaction Load into Turret< 5 Micron Dust Brush Width Brush Length Wire Bend Brush Offset Distance from Center Brush Separation Distance Brush Symmetry Brush Pivot Axis Distance from Center Brush Angle Spring Preload Brush Speed Approach Algorithm Retract Algorithm Engagement Distance Brush DurationBrush Geometry Brush Holder Geometry Software and PlacementRequirements and Restrictions

Document

AMS_2012.pdf

284 Many parameters were discovered to have an effect on the performance of this dual pivoting brush design. These parameters and their relationship to design/performance requirements are illustrated in Figure 6. Following is a summary of findings associated with key brush parameters:  Brush offset distance from center – A large offset distance made it difficult or impossible to clear the center of the 45-mm diameter area; a small offset increased the chances of bristle entanglement around the center post or with bristles from the other brush.  Brush separation distance – Naturally if the brushes were too close they became entangled; a minimum separation distance (~8 mm) was established which yielded best performance (minimal entanglement).  Brush symmetry – It was found that with a symmetric brush design, it was impossible to clear the center of the 45-mm diameter area without severe bristle entanglement; an asymmetric brush design was adopted where one brush (Inner Brush) overlaps and clears the center and most of the area while the other (Outer Brush) assists with clearing the outer part of the area.  Engagement distance (brush angle) – The engagement distance is the distance between the DRT and the rock surface as measured from the tip of the center post to the rock surface; being too close (<10 mm, shallow brush angle) resulted in a poorly cleaned surface while being further away (>10 mm, steep brush angle) resulted in better clearing and better ability of the bristles to pluck dust out of vesicular pits and crevices.  Brush width – There seemed to be a critical brush width (~15 mm or 1/3 the required cleared area diameter) where thinner brushes produced very clear surfaces of smaller diameter and wider brushes produced less clear surface of larger diameter.  Wire bend – Straight bristles had a very difficult time removing fine particles from voids and crevices unless the engagement distance was large enough; forming a bend in the bristles enhanced their ability to “scrub” the surface even at shallow brush angles (close engagement).  Brush speed – Speeds less than 300 rpm did not impart enough energy to effectively remove particles, instead the brushes just pushed the particles around in a circle; at speeds greater than 300 rpm, the brushes are much more effective at clearing the particles – a big difference between 300 rpm and 500 rpm was observed while a negligible difference between 500 rpm and 1000 rpm was observed.  Approach algorithm – Not spinning the brush block while approaching the rock occasionally produced large reaction forces and at times no “center clearing”; spinning while approaching rock reduced axial force on DRT, resulted in symmetric pivoting of brushes and helped to clear the center of the 45-mm diameter area.  Retract algorithm – Spinning at high speeds (or not spinning at all) while retracting tends to pull dust and debris back into cleared area; spinning at lower speeds (100 rpm or less) tends not to drag or eject material into clean areas. Ultimately, all of the prototype testing resulted in the final flight brush design shown in the far right-hand image in Figure 5. The DRT flight brush bristles were made of a material similar to RAT brush bristles [4]. A brush manufacturer delivered straight brushes of a specified width and Honeybee formed the final brush geometry by bending the bristles around mandrels per a template and potting with a suitable flight-grade adhesive. The bristles were then trimmed to length. At two points in the process the bristle wire and brush assemblies underwent ultrasonic cleaning for contamination control (CC) and planetary protection (PP) reasons. Samples were sent to JPL for CC/PP analysis and approved for flight. The spring-loaded brush pivots (hinges), Figure 7, were designed to keep the brushes lightly loaded against the rock surface across large variations in surface height relative to the DRT. The pivots allow the brushes to rotate 30° from hard-stop to hard-stop. The whole brush block assembly is designed to withstand inadvertent loading by the RA up to 200 N. This is one reason the pivot shaft is supported by bushings as opposed to small ball bearings. Spring-energized Bal Seals protect the bushings against dust ingress. The Bal Seals were match fit with the shaft to reduce the parasitic drag to near zero at standard temperature and pressure – a similar procedure was used with success for the RAT grinding wheel shaft

Document

AMS_2012.pdf

285 and Phoenix Icy Soil Acquisition Device cutting bit shaft [4][7]. The shaft is spring-loaded against a hardstop by dual torsion springs. End cover-plates with integral spring arbors enclose the spring-bushing area. The pivot shaft-bushing-spring system is dry lubricated. Figure 7. DRT Brush pivot Figure 8. DRT Flight Model brush contact force vs. surface height Sizing the pivot’s springs was an exercise in threading the needle. On the one hand, the springs needed to be strong enough to keep the brushes in contact with the rock over the complete range of pivot motion (including at its relaxed hard stop). And on the other hand they needed to be soft enough such that the brush contact force at the fully compressed pivot position caused very little drag on the motor and the motor torque margin requirement would be met. It only took on the order of 3 N of bristle contact force to achieve the allowable motor current limit. So the springs were designed to deliver about 1.4 N of contact Applied Brush Force vs. Center Post Distance from Target Across Operational Temperature Range 0.00.51.01.52.0 0 5 10 15 20 25 Distance from Center Post to Target Surface (mm)Average Measured Force (N)-70°C +23°C +70°C

Document

AMS_2012.pdf

286 force at the fully compressed pivot position and the spring rate was selected to be as flat as possible such that the contact force was about 0.4 N at the lower hard-stop. The brush block subassembly was dynamometer tested across the operational temperature range (-70°C to +70°C). Instead of using a torque watch to directly measure the torque supplied by each pivot axis over the range of motion, a test was designed to measure the net contact force delivered by both brushes against a contact plate instrumented with a load cell. At each temperature set point, the instrumented contact plate was translated toward the DRT brush block as the brush block was spinning at low RPM – this mimicked the flight operational concept. The DRT and contact plate translation stage were configured horizontally such that gravity should not have been a factor. Force measurements were taken at 21 mm (1 mm beyond the brushes reach), 16 mm, 11 mm, 6 mm and 1 mm between the center post tip and the contact plate. Measurements were not taken in the reverse order to quantify friction drag on the pivot axis. The results of this test, shown in Figure 8, were nominal (i.e., the springs produced forces that were within the expected range). Brush Pivot Problem/Failure Report Following the dynamometer testing of the DRT Flight Model (FM) and Engineering Model (EM) units and the subsequent delivery of the FM, the DRT EM was performance and life tested across the operational temperature range. During these tests, the EM motor was energized to spin up the brush block while it was positioned above a rock surface in the thermal vacuum chamber. Once at speed, a linear translation stage (ground support equipment playing the part of the RA) moved the DRT to a position of engagement with the rock. After a period on the rock, the motor speed was decreased to a lower setting and the DRT was retracted off the rock. This test procedure was consistent with the manner in which the tool would be used during the mission. During the -70°C tests, it was observed the EM Inner Brush did not return all the way to its hard-stop following the low speed retraction from the rock. Instead it stopped a few degrees away from the hard-stop. However, the Inner Brush did return to its hard-stop position following the nominal 10-second 900 rpm run. The EM performance and life testing continued until the EM had successfully met its dust removal performance requirements after 2x life (greater than 300 brushing operations). The Problem Failure Report process was initiated to capture the anomalous Inner Brush behavior and subsequent root cause investigation. The concern was that the pivot (unassisted by centripetal force) was not meeting its torque margin requirement. Per the DRT equipment specification, all actuators were required to demonstrate a minimum margin of 100% on torque required for operation under worst-case conditions where Margin = (Actual/Required – 1)*100%. In this case, the EM Inner Brush pivot spring (the actuator) had apparently failed to supply enough torque to overcome friction and return the brush to its hard-stop. The requirement in this case is that the spring be able to supply twice the torque required to overcome the friction opposing a return to hard stop – this would be considered a margin of 100% or factor of safety (FOS) of 2. The first thing the team did was to revisit the pivot analytical model used during the design process as well as the as-built in-process test data and dynamometer test data. According to the manufacturing documentation the measured seal drag on the EM’s Inner Brush pivot was 50% higher than that of the EM’s Outer Brush and both the FM Inner and Outer Brush. This appeared to be the smoking gun. Unfortunately however, the team realized that the dynamometer test approach inadequately demonstrated the torque (or force) margin and it was impossible to draw any further conclusions about the pivot spring’s torque margin. Pursuing the seal drag theory, it was hypothesized that seal contraction at cold temperatures had caused the drag torque to spike on the EM Inner Brush seal-shaft combination because it was perhaps past the “knee in the curve” known to exist in these types of configurations. So measurements of seal drag versus diametral interference were made at room temperature using gage pins to create a model (see Figure 9) and the maximum theoretical seal contraction was calculated. According to the model, the EM Inner Brush would have a margin less than 100% (FOS < 2) but the

Document

AMS_2012.pdf

287 margin should still be comfortably positive (FOS > 1). It was clear that something was missing from the model. Figure 9. Seal Drag Model Next a complete disassembly of the EM Inner Brush pivot was performed and documented. Unfortunately, other than very minor evidence of components rubbing (dark marks), there was no clear culprit explaining the problem with the EM Inner Brush pivot. The EM brush block was reassembled and the unit was sent to JPL for further testing. At JPL, tests were performed at temperature on both the EM and FM to more directly measure the pivot’s drag and spring torque and thereby achieve a clear demonstration of torque margin. A general model of the pivot behavior is shown in Figure 10 and assumes drag torque is equal in both directions and of course opposes the direction of motion. So in one direction, the measured torque is the sum of the spring and drag torque. While in the other direction, the measured torque is the difference between the spring and drag torque. Figure 10. Pivot spring drag torque model The JPL test results are shown in the columns labeled “0 rpm” in Table 1. At 0 RPM, the brush block pivots demonstrated a margin of 100% or more (FOS ≥ 2) in only one instance for the EM and only one 012345678 -0.001 0.000 0.001 0.002 0.003 0.004 0.005 0.006Additional Diametral Interference (in)Seal Drag Torque (in-oz)EM Outer Brush & FM Inner/Outer Brush @ -70C (predict) EM Inner Brush @ -70C (predict) Knee in the curveEM Outer Brush & FM Inner/Outer Brush @ 23C (measured) EM Inner Brush @ 23C (measured)

Document

AMS_2012.pdf

288 instance for the FM. In many cases, negative torque margin was calculated. The JPL tests confirmed that the DRT pivot design did not meet the torque margin requirement. Table 1. DRT Pivot Torque Factor of Safety (FOS) Test Results (min. required FOS: 2.0) Next calculations were repeated to determine the torque margin when taking into account the centripetal forces acting on the brushes when the motor is spinning the brush block at 900 rpm – this was the speed used at the end of each performance test brushing operation after the DRT had been retracted away from the rock. These calculations are shown in the columns labeled “900 rpm” in Table 1. It was determined that operating the DRT at 900 rpm produces a calculated centripetal force that will reliably return the brush pivots to their hard stops. The minimum FOS for the FM in this case is 1.64 (Outer Brush Test #1 at -70°C). No further work was done to determine root cause and no hardware changes were called for as a flight operations rule that spins the DRT brush block at 900 rpm before each use was deemed an adequate solution by the review board. Motor & Gearhead Subassembly A challenge in the DRT design was to overcome the torque margin constraints due to the required use of a predefined motor. Based on the motor performance specifications, torque amplification was necessary and a custom single-stage planetary gearbox was designed (ref. Figure 11). The pinion of the motor was predefined as a long-addendum spur gear. This was likely done to avoid undercutting and increase the load capacity and life of the small pinion, necessitating the use of modified profile planet and ring gears. Actuator margin analysis (ref. Figure 12) was performed based on those performance parameters as well as thermal limitations provided by JPL based on heat-up analyses. Various parameters were also estimated including seal and bearing losses as well as operational torque. The latter was estimated based on data from brush development tests. 0 r p m9 0 0 r p m0 r p m9 0 0 r p m0 r p m9 0 0 r p m0 r p m9 0 0 r p m Test #1 2.00 6.96 1.24 4.64 1.51 5.59 1.33 4.69 Test #2 1.78 6.54 1.23 4.60 1.83 6.44 1.56 5.54 Test #1 1.30 5.03 1.58 4.85 1.27 4.79 1.87 4.85 Test #2 1.23 5.28 1.42 4.63 1.17 4.60 2.20 6.20 Test #1 0.67 3.34 0.61 2.27 0.96 3.61 1.05 3.28 Test #2 1.06 3.58 0.54 2.29 0.74 3.13 0.52 2.13 Test #1 0.76 2.22 0.47 1.50 0.43 1.87 0.51 1.64 Test #2 0.57 2.04 0.35 1.33 0.38 1.78 0.64 1.77Test #FM DRT EM DRT Inner Brush Outer Brush Inner Brush Outer Brush Temp (°C) +70 +20 -30 -70

Document

AMS_2012.pdf

289 Figure 11. Cross Section of Gearbox Assembly Analysis predicted that a 10.4:1 reduction ratio gave a torque margin of 140% and a thermal dissipation margin of 475% which exceeded the 125% requirement for a CDR-level design. Reducing the ratio to 6:1 decreases those margins to 41% and 99%, respectively. The gearbox single-stage ratio selection of 10.4:1 is somewhat unconventional when considering planetary gearbox rules of thumb which limit the highest practical reduction ratio to 10:1 due to decreased pinion size and increased sliding within the gear mesh (i.e. decreased efficiency and life). Figure 12. Actuator margin analysis methodology The specific sliding ratio (SSR) is defined as the sliding velocity divided by the rolling velocity at the gear mesh. The SSR varies along each gear mesh. A higher SSR means that a gear mesh design has relatively more sliding action compared to a lower SSR and would therefore have increased friction and decreased efficiency and life. Therefore it is desirable to design a reduction stage to minimize the SSR within the constraints of other driving requirements. The rate of change of the SSR along the line of action should also be minimized for similar reasons.

Document

AMS_2012.pdf

290 The DRT required a high single-stage ratio in order to maintain torque margin on the motor as previously discussed. This in combination with a pre-defined 15T pinion fixed the gear proportions. Figure 13 shows how the SSR at the sun-planet gear mesh for the 10.4:1 design changes during a single mesh cycle. Looking at the lowest and highest points of single tooth contact (LPSTC and HPSTC respectively), i.e., the critical stress jump points, the SSR does not exceed a magnitude of 1.0. The worst-case SSR value occurs on the dedendum of the planet during recess action with a magnitude greater than 3.0. Because the SSR is less than 1.0 at both the LPSTC and HPSTC, the DRT gear design was deemed acceptable. Other analyses included gear life analysis based on the American Gear Manufacturers Association (AGMA) standards 2001-D04 and 908-B89, backlash analysis, and momentary overload analysis. Backlash analysis included contributions from manufacturing tolerances as well as the operational thermal range to ensure that backlash existed under worst-case stack-up conditions. This was validated via accurate involute profile simulations. Momentary overload calculations included a combination of AGMA 2001-D04 and finite-element analysis (FEA) depending on the gear material selected. Figure 13. Sun-planet gear mesh specific sliding ratio To validate the above analyses and verify that the DRT gearbox efficiency was acceptable, the DRT was subjected to output axis dynamometer testing across the operational temperature range. Dynamometer data was acquired at temperature set points (+70°C, +23°C, -55°C, & -70°C) at multiple motor voltages (6, 10, 16 and 22 volts). Speed-torque and current-torque profiles were created. The motor torque constant (K t) values calculated from the dynamometer data, as shown in Figure 14, match up well with the K t values measured for the A300 motor prior to integration with the DRT gearbox. This data shows the capability to design efficient planetary gearboxes with relatively high reduction ratios given careful gear tooth profile selection and analysis. Figure 14. DRT-integrated A300 Kt values across the protoflight thermal range are shown on the left. A300 Kt values (not integrated with DRT are shown on the right. 6 Volts 10 Volts 16 Volts 22 Volts 16 Volts 21 Volts Temperature CCW CCW CCW CCW CCW CCW +70°C 18.0 19.4 20.0 19.2 20.0 20.7 +23°C 18.8 20.4 19.8 20.1 21.1 21.0 -55°C 19.9 20.2 20.5 20.3 N/A N/A -70°C 21.4 22.0 20.3 20.7 23.1 23.1DRT-integrated A300 Kt values A300 Kt values

Document

AMS_2012.pdf