Datasets:

ai-aerospace

/

ams_data_full_2000-2020

like 0

Table 2. Ball Bearing Material Properties 8 0.06- E 3 0.04- aJ V u I m u 0.02 - oisson's Ratio 0.27 hermal Conductivi Yovanovich (theory) G = 0.276 * P'" Results A Yovanovich Following the experimental matrix, tests were performed to establish the influence of rotational speed, axial load, and temperature on bearing thermal conductance. While results for oil-lubricated bearings prove most relevant to a spacecraft application, the influence of axial load on the thermal conductance of a static dry bearing is already well known analytically [l-21. Thus, the experimental results for a static dry bearing provide an excellent means of verifying the experimental technique. Further testing extends the results to static and dynamic (constant rotational speed) lubricated bearings, and reveals drastic shifts in the magnitude and trend of bearing thermal conductance relative to axial load and temperature. Drv (Non-Lubricated) Static Bearinas The analytical relationship between thermal conductance and axial load in a non-lubricated, non-rotating bearing is well known. Yovanovich [l -21 developed the best-known analytical method, deriving the equation for bearing thermal resistance. A comparison with this model was used to verify the experimental approach, before proceeding to the more complex cases of lubrication and motion. Figure 1 provides a comparison of experimental and theoretical bearing thermal conductance as a function of axial load at 20 "C. Both experiment and analysis agree closely in both magnitude and trend. The Figure shows that thermal conductance of the dry static bearing responds to axial load to the 1/3 power. 0.08 1 o Experiment Axial Load (N) Figure 1. Comparison of Analytical Calculation of Thermal Conductance with Experiment for a Static Dry 204-Size Hybrid Bearing at 20 "C 293

Document

AMS_2006.pdf

Yovanovich’s analysis provides an explanation for the observed trend. His model assumes the mechanism of heat transfer through a non-lubricated static bearing is pure conductance, from the hotter race, through the ball, and to the cooler race (Figure 2) and the driving source of resistance across the bearing arose from the thermal constriction region between the ball and the race contacts. This recognition presumes perfect contact (no asperities), and negligible resistance within the ball and both races. Q A Cooler Race Rball = Rin + Rout Gball = 1 /Rball Hertzian Contact Hotter Race t Rin Figure 2. Thermal Resistance Across One Bearing Ball As such, Yovanovich calculated the thermal resistance across a dry static bearing by modeling the ball and races as semi-infinite half-planes with the Hertzian contact area modeled as the thermal constriction region. The basic equations to calculate thermal resistance across each of the ball to race contacts are: inner race to ball thermal resistance ball to outer race thermal resistance total thermal resistance across the ball Where Y is a non-dimensional geometric factor defined as: \I12 where n = i or o 17 “-b“sin2 a: (4) The Hertzian contact ellipse, and associated major and minor axes (a and b), can be calculated by a number of existing programs based on classical Hertzian theory, such as BRGSlOC [12]. Here, the major and minor axes are related to the applied axial load to the 113 power. Thus, the total Hertzian contact area is related to the applied axial load to the 213 power. a.b 4 A , =n.- Hertzian Once the resistance across each ball is known, the conductance is determined by taking its inverse. Scaling this result by the number of balls in the bearing yields the total bearing conductance (Equation 6). Equation 6 assumes that each ball provides an equivalent thermal pathway as a result of pure axial load, a condition consistent with our experimental setup. As the influence of axial load cancels out in the non- 294

Document

AMS_2006.pdf

dimensional parameter Y, only the influence on the Hertzian contact ellipse major axis remains in the thermal conductance equations of 1-3. Thus, analytically, conductance is sensitive to axial load to the 1/3 power. Both the magnitude and trend predicted by the analysis compared well with the experimental measurements. 0.070 - 0.060 - @ E 0.050 - E 3 0.030 - a 8 0.040- a g 0.020 - Figure 3 shows the effect of the average bearing temperature (average between the inner and outer races) on thermal conductance of a dry (non-lubricated) 204-size hybrid bearing for three different axial loads. The graph shows that the conductance of the dry bearing is not strongly responsive to temperatures, especially when compared to the effect of axial load. This is attributed to the weak dependencies of the relevant material properties (thermal conductivity and Young's modulus) of the steel races and ceramic balls to temperature. 8-4 Axial Loads +129 N 1 u 0.01 0 e84 N +40 N 0.000 I I 1 I I 0 10 20 30 40 Temperature (C) Figure 3. Effect of Average Temperature on Thermal Conductance of a Static Dry 204-Size Hybrid Bearing Oil Lubricated Static Bearinas Yovanovich's model is a good approximation for some applications, such as dry lubricated bearings with little motion, but once lubrication is introduced into the system, bearing thermal conductance can change significantly. Figure 4 plots the static thermal conductance of a 204-size hybrid bearing in three lubrication states; dry, virgin oil lubricated, and oil lubricated after run-in. The exercised bearing was fully run-in at 6000 RPM, then brought back to 0 RPM for testing. Thermal conductance was measured for three axial loads, over a range of temperatures for each lubrication state. The significantly higher thermal conductance of the oil-lubricated bearings, in comparison with the dry, warrants attention. To explain the difference between dry and oil lubricated bearings, Figure 5 depicts one of the ball-to-race contact regions. The mechanism of heat transfer is still pure conductance, but the large increase suggests that the lubricant meniscus surrounding the ball contributes a significant heat transfer path by increasing the constriction area at the ball to race interface. Lubricant could also potentially reduce the thermal resistance due to asperity contact at the metal-to-metal contact region; however, the close agreement between the Yovanovich model and the experiment for dry bearings suggest that this represents a secondary effect. The large increase in thermal conductance implies that the heat path provided by the lubricant meniscus ultimately dominates, masking the influence of the Hertzian contact area. While the thermal conductance across dry bearings are driven by the size of the Hertzian contact ellipse, and thus proves sensitive to 295

Document

AMS_2006.pdf

axial load, the masking of the Hertzian contact area by the meniscus explains the minimal influence of axial loading upon the conductance of the oil-lubricated bearing. The dominance of the meniscus also explains the observation that oil lubricated bearings were found to be more sensitive to temperature, a result attributable to the higher temperature dependencies of the lubricant material properties. 0.25 0.20 1 @I 0.15 Q) $ 0.10 - a a u 0.05 - +129 N, Dry -a-84.5 N, Dry 4 129 N, Virgin, Lubricated +84.5 N, Virgin, Lubricated +40 N, Virgin, Lubricated + 129 N, Run-In, Lubricated +84.5 N, Run-In, Lubricated 4-40 N, Run-In, Lubricated 440 N, Dry 0.00 I I I I I I -1 0 0 10 20 30 40 Temperature (C) Figure 4. Effect of Average Temperature and Axial Load on Conductance of a Static 204-Size Hybrid Bearing, for Dry and Oil Lubricated Bearings Lubricant meniscus erkian contact area Figure 5. Lubricant Meniscus Adds to Thermal Pathway Between Ball and Race The conductance values of the virgin bearing are slightly higher than those of the fully run-in bearing, as apparent in Figure 4. This observation was attributed to lubricant loss during the run-in process, as the ball pushed excess oil out of its pathway and centrifugal forces displaced lubricant from the ball. This means that the degree of difference will be dependent upon the initial amount of lubrication and the maximum run-in speed of the bearing. For space applications, bearings are typically lubricated once, and this lubricant may deplete over the mission duration due to various reasons, such as lubricant migration or run-in. Knowing this, one may construct bounds on static or slow moving conductance values expected throughout the spacecraft mission life. Yovanovich establishes the lower extreme bound, and test measurement of a virgin bearing establishes the upper bound. Lubricant loss can occur throughout the life of the bearing for various reasons including run-in and lubricant migration, meaning the end-of-life conductance would be somewhere between these bounds. This observation, however, does not hold for dynamic bearings. 296

Document

AMS_2006.pdf

Oil Lubricated Dvnamic Bearinas Figures 6-8 show the influence of average bearing temperature on conductance at different rotational speeds. Each figure represents a different axial load applied to an oil-lubricated bearing that had been run-in at 6,000 RPM. 0.20 - @ 0.18 - E 0.16 - 1 0.22 I +ORPM -ff 3000 RPM -A- 4500 RPM -8- 6000 RPM 40N Axial Load 0.08 ! I I I I I -1 0 0 10 20 30 40 50 Average Bearing Temperature (C) Figure 6. Thermal Conductance of a 101-Size Steel Bearing for a 40-N Axial Load 0.22 I 0.20 - a 0.18 - 0.16 - 1 0.14 - iE 0 'F) E 6 0.12 - 0.10 - +O RPM 84.5N Axial Load 43- 3000 RPM k- 4500 RPM I '3000RPM P 0.08 f I I I I I -1 0 50 0 10 20 30 40 Average Bearing Temperature (C) Figure 7. Thermal Conductance of 101-Size Steel Bearing for 84.5-N Axial Load 297

Document

AMS_2006.pdf

0.22 I 0.20 - 0.18 - 0.16 - y 0.14 - u 0.12 - 0.10 - @ i a 0.15 - @ 0.14 - E 0.13 - +ORPM 129N Axial Load -Ef 3000 RPM 4500 RPM 1000 RPM 3 8 0.11 - 3 0.12 - m u 0.10 - 0.09 - 0.08 t I I I d 0.08 1 I I I I I -1 0 0 10 20 30 Average Bearing Temperaturn (C) 40 50 Figure 8. Thermal Conductance of 101-Size Steel Bearing for 129-N Axial Load Several observations arise from these results. First, a distinct difference exists between static and dynamic bearings. Once in motion, the thermal conductance increased with temperature in a linear manner over the data range explored. Moreover, the slope and magnitude increased with axial load. To separate the response to each individual variable, Figure 9, shows the influence of axial load on the bearing at 6000 RPM and at a temperature of 20 "C. Unlike static lubricated bearings that are insensitive to axial loads, the dynamic lubricated bearing responds in a linear manner over the test range. P Figure 9. Effect of Axial Load on a Lubricated 101-Size Steel Bearing at 20 "C 298

Document

AMS_2006.pdf

Figure 10 re-plots the data to investigate the influence of rotational speed at a constant temperature (20°C). There is a distinct difference between 0 RPM and motion, which becomes more pronounced with increasing axial loads. Furthermore, contrary to intuition, at a given temperature and axial load the thermal conductance proved insensitive to rotational speeds below the run-in speed. The difference between static and dynamic bearings reflects a difference in the predominant heat transfer mechanism. 0.16 I 0 40N Axial Load 0 84.5N Axial Load A 0.15 1 5 0.13 0 0 A 0 0 1 I Static 0.09 1 Test Data Dynamic Test Data 0 1000 2000 3000 4000 5000 6000 Speed (RPM) Figure 10. Effect of Speed on a Lubricated 101-Size Steel Bearing at 20 "C For moving bearings, conductance through the bearing is not the primary mechanism of heat transfer. Figure 11 illustrates a ball with a lubricant film and a meniscus at the ball to race contact. The dominant mechanism of heat transfer is most likely mass transport, where the lubricant at the meniscus of the hotter race picks up heat, transports it with the ball as it rotates, then deposits heat at the cooler race. The film thickness on the ball becomes a dominant player as it determines the amount of heat transport that occurs. The ball film thickness should not be confused with the elastohydrodynamic (EHD) film thickness, which in the context of our argument has no influence on heat transport. Lubricant film I)- thickness on ball - EHD film thickness Meniscus Figure 11. Lubricant Distribution on Bearing Ball As the minimum rotational speed tested was 3000 RPM, it stands to reason that there may be a transition region between 0 RPM and 3000 RPM, where speed does influence bearing conductance. Looking through existing literature, Stevens and Todd from ESTL [3] explored the influence of speed on bearing thermal conductance at low speeds. An example of their findings, depicted in Figure 12, indicate an initial decline in thermal conductance, followed by an increase at higher speeds. 299

Document

AMS_2006.pdf

0.1 0.08 @ 5 a 0.06 8 8 a u 0.04 0.02 + 1OOrevs 0 15,OOOrevs (Speed Increasing) A 100,OOOrevs (Speed Decreasing) + 100,OOOrevs (Speed Increasing) 100,OOOrevs (Speed Decreasing) .. .. .: : ... 1 0.1 1 10 100 1000 Figure 12. Effect of Low Speeds on a 42-mm OD Bearing Lubricated with 11.4 mg of BP135 Oil with 40-N Load; Study by Stevens and Todd [3] Of additional relevance, Stevens and Todd [3] conducted tests multiple times with consecutively increasing or decreasing speeds. The result was a reduction in conductance occurring with additional revolutions. This is an effect that we also observed when comparing virgin and fully run-in bearings, again most likely due to the effect of excess lubricant being displaced over time. Our observations, and the literature results, indicate that this effect eventually diminishes and tests become repeatable when the bearing reaches a fully run-in state. Discussion The objective of this research was to acquire a fundamental understanding of bearing thermal conductance, to establish which variables influence that property, and to determine the dominant mechanisms of heat transfer. Research results indicate that bearing thermal conductance was influenced by a number of interdependent variables, underpinned by the lubrication and the dynamic state of the bearing. Figure 13 provides a global example by plotting the influence of axial load on a 101-size steel bearing at 20 "C. The large differences in trend and magnitude between the dry and lubricated, static and dynamic bearings illustrate that the thermal analyst needs to recognize the assumptions underlying experimental data or analytical models of bearing thermal conductance. Metal-to-metal contact, through the Hertzian contact ellipses, provided the heat transfer mechanism for dry static bearings. Experimentally, dry static bearings were found to be sensitive to axial load to the 113 power, as predicted by the Yovanovich analytical model, and relatively insensitive to temperature. Yovanovich's method establishes a lower bound for any bearing. The presence of lubrication can drastically change the bearing thermal conductance. The menisci contributed an additional thermal pathway that overshadowed the conductance through the metal-to-metal Hertzian contact area. In contrast to dry bearings, the influence of axial load was negligible for static lubricated bearings, but the effect of temperature proved significant. 300

Document

AMS_2006.pdf

Mass transport of oil was considered the dominant mode of heat transfer in the lubricated dynamic bearing. Both axial load and temperature all affected this mechanism of heat transfer, and ultimately the bearing thermal conductance. 0.1 6 0.1 5 0.14 0.1 3 0.1 2 0.1 1 0.09 9 s Y u E a a 0.08 0.07 0.06 0.05 0.04 0.03 0.02 0 Dynamic, Lubricated 4 Static, Lubricated -0- Static, Non-Lubricated 50 100 Axial Load (N) 150 Figure 13. Effect of Axial Load on Thermal Conductance of a 101-Size Steel Bearing at 20 "C 30 1

Document

AMS_2006.pdf

Conclusion This paper has shown that lubrication or motion affects bearing thermal conductance in both magnitude and sensitivity to operational conditions, such as temperature and axial load. In practice, engineers typically use heritage information from applications using similar bearings within a comparable system to obtain an estimate of bearing thermal conductance. But in some cases heritage information is altogether lacking due to a new or a one-of-a-kind system. Engineers have often used analyses for a first-order approximation of bearing thermal conductance, but the basic analytical equations apply to a non- lubricated, static bearing. Information available on other bearing types is sometimes used as well. However, drastic difference in both thermal conductance magnitude and sensitivity to variables such as axial load and temperature is dependent on bearing lubrication and state of motion. As a consequence, engineers need to evaluate where bearing conductance data came from, what conditions it represents, and whether those conditions reflect the application at hand before using the data to predict temperatures of a rotational device in space. Acknowledgements This work was funded by The Aerospace Corporation Independent Research and Development (RAD) program. The authors will also like to extend our deepest gratitude to our management, in particular, Dr. Michael R. Hilton. It was his outstanding support that made this research possible. References 1. 2. 3. 4. 5. 6. 7. 8. 9. Yovanovich, M. M., “Analytical and Experimental Investigation on the Thermal Resistance of Angular Contact Instrument Bearings,” Instrumentation Laboratory, E-221 5 (1 967), Massachusetts Institute of Technology, Cambridge, Massachusetts, 85p. Yovanovich, M. M., “Thermal Constriction Resistance Between Contacting Metallic Paraboloids: Application to Instrument Bearings,” Proceedings of Heat Transfer and Spacecraft Thermal Control, (1971), M.I.T, Cambridge, MA, pp. 337-358. K. T. Stevens, M. J. Todd, “Thermal Conductance Across Ball Bearings in Vacuum,” Report Number: ESA-ESTL-25, (1 977), National Centre of Tribology, Risley (England), 51 p. A. A. M. Delil, J. F. Heemskerk, and J. P. B. Vreeburg, “Design Report on the ESRO Test Rig to Measure the Thermal Conductance and Friction Torque of Rotating Bearings in Vacuum,” Report Number: NLR-TR-74069-U, (1 974), 9Op. J. F. Heemskerk, A. A. M. Delil, J. P. B. Vreeburg, “A Test Rig to Measure the Thermal Conductance and Friction Torque of Bearings in Vacuum,” European Space Tribology Symposium, (1 975), Frascati, Italy, 18p. Rowntree, R. A., Todd, M. J., “Thermal Conductance and Torque of Thin Section Four-Point Contact Ball Bearings in Vacuum,” Report Number: ESA-ESTL-54, National Centre of Tribology, Risley, England, (1 983), 57p. Rowntree, R. A., “Stiffness, Torque, and Thermal Conductance of Thin-Section Four-Point Contact Ball Bearings for Use in Spacecraft Mechanisms,” Proceedings of the Ist European Space Mechanisms & Tribology Symposium, (1 983), Noordwijk, Netherland, pp. 91 -100. Anderson, M. J., Roberts, E. W., and Rowntree, R. A., “The Thermal Conductance of Solid-Lubricated Bearings at Cryogenic Temperatures in Vacuum,” Proceedings of the 30th Aerospace Mechanisms Symposium, (1996), European Space Tribology Lab., Cheshire, England, pp. 31 -45. Nakajima, K., “Thermal Contact Resistance Between Balls and Rings of a Bearing Under Axial, Radial, and Combined Loads,” Journal of Thermophysics and Heat Transfer, 9, (1 995), pp. 88-95. 10. P. D. Fleischauer and M. R. Hilton, “Assessment of the Tribological Requirements of Advanced Spacecraft Mechanisms,” Materials Research Sociefy Symposium Proceedings, (1 989), pp. 9-20. 11. Y. R. Takeuchi, J. T. Dickey, S. M. Demsky, K. K. Lue, J. J. Kirsch, P. P. Frantz, “A Methodology in Measuring Thermal Properties of Bearings in Motion,” Proceedings of the 75th Annual Thermal and Fluids Analysis Workshop, (2004), JPL, Pasadena, CA, 25p. 12. A. Leveille, BRGSI OC, (a bearing analysis code), The Aerospace Corporation, El Segundo, CA. 302

Document

AMS_2006.pdf

Mars Exploration Rover Potentiometer Problems, Failures and Lessons Learned Mark Balzer* Abstract During qualification testing of three types of non-wire-wound precision potentiometers for the Mars Exploration Rover, a variety of problems and failures were encountered. This paper will describe some of the more interesting problems, detail their investigations and present their final solutions. The failures were found to be caused by design errors, manufacturing errors, improper handling, test errors, and carelessness. A trend of decreasing total resistance was noted, and a resistance histogram was used to identify an outlier. A gang fixture is described for simultaneously testing multiple pots, and real time X-ray imaging was used extensively to assist in the failure analyses. Lessons learned are provided. Introduction On the Mars Exploration Rover (MER), many angular positions required remote monitoring. An incomplete list includes the angular positions of four Steering Actuators (WSA), the azimuth, elevation, elbow, wrist and turret joints on the Instrument Deployment Device (IDD), and the azimuth and elevation axes of the High Gain Antenna Gimbal (HGAG). On the MER Lander, knowledge of the angular position of the three Lander Petals was also required. The MER avionics monitored these angular positions by reading the digital, incremental, rotary, magnetic encoder on each of the fourteen electric motor shaft inputs to these geared mechanisms. In case of encoder failure, or if avionics power and thus stored angular position data was lost, all fourteen angular positions could be determined and tracked via backup precision analog potentiometers (pots) configured to measure the output angles of the WSA, the IDD joints, the HGAG axes and the Lander Petals. Pots were also used in the MER suspension. The left and right rocker-bogie pivots each had one pot to measure the relative angle between each bogie and its rocker arm, while a third pot in the MER differential measured the relative angle between the left and right rocker arms. In conjunction with a three-axis accelerometer providing the Rover's orientation with respect to a gravity vector, relative angle measurements from these pots allowed the Rover's complete kinematic state to be determined. Though the suspension pots were the sole source for this data, it was deemed non-critical for a nominal mission. Pots are miniature electromechanical mechanisms and as such must pass a rigorous series of tests to be considered flightworthy. This paper will describe some of the more interesting issues that arose during qualification testing of the MER pots, detail their investigations and present their final solutions. Lessons learned along the way will be pointed out. Description of Pots The Lander Petal, Bogie Pivot and Differential pots shared a common design based on MIL-PRF-39023; single-turn rotary pots in a metal cup with a bearing-mounted input shaft. The HGAG and WSA pots were each custom, single-turn rotary units designed to be mounted within their respective actuators, and thus had no bearings of their own. All pots used on MER were of the precision rotary nonwire-wound (conductive plastic) type. Since the pots described in this paper (Figure 1) were provided by a single vendor, they shared basic design features such as: 1) An aluminum alloy "cup" which served as the frame of the pot. * Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA Proceedings of the 3dh Aerospace Mechanisms Symposium, Langley Research Center, May 17- 19,2006 303

Document

AMS_2006.pdf

Figure 1. MIL-PRF-39023-style pot, HGAG pot, and WSA pot, from left to right 2) A conductive plastic disk bonded to the cup. This disk contains two annular tracks: a <360" resistive track (the resistance element that the wipers rub) and a 360" conductive track (or slip ring) co-molded into the Diallyl Phthalate disk substrate. 3) External electrical connection via three brass terminals potted into the cup on the MIL-PRF-39023- style pots, or three 28 AWG pigtails provided on the WSA and HGAG pots. 4) Internal electrical connections made by welding, soldering or bonding with conductive epoxy, and employing solid, flat ribbon wire conductors connecting: a) the end termination junctions of the resistance element to the "CW" and "CCW" input terminals or input pigtails, and b) the junction on the conductive track (slip ring) to the "W" wiper terminal or wiper pigtail. 5) Series connected wipers mounted to, but insulated from, the rotating shaft. Sweeping over both tracks, they pick up the voltage from the resistive track and send it through the conductive track (slip ring) to the wiper terminal. Wipers were comprised of multiple, thin, cantilevered contact fingers made of precious metal alloys with contact forces controlled to approximately 209. During calibration, each resistance element was manually trimmed to a specified linearity tolerance by scratching the surface of the conductive plastic disk with a scribe, and/or by placing dots of conductive silver paint on its surface. All resistance elements were 6250 * 20% Ohm and rated for 1 watt; all wiper circuits were rated to carry 10 mA. The MER avionics energized the resistance elements with 5 VDC for a power dissipation of <5 mW and drew only 50 nanoAmps through the pot wiper circuits. Qualification Testing To expedite matters, the MER project designated one mechanical engineer to be cognizant of all pot technical issues and qualification testing. For validation purposes non-flight Engineering Model (EM) pots were ordered and received well before the flight pots. To reduce cost and lead time, the EM pots were subjected to a reduced set of acceptance tests by the manufacturer. When the EM pots arrived at JPL they were distributed to the end users without careful inspection or electrical testing. Most of the qualification testing on the flight pots was done by the manufacturer per a JPL-approved Acceptance Test Procedure (ATP). However, since the manufacturer did not have thermal-vacuum test chambers capable of the required -120" to +llOo C range, the thermal-vacuum qualification tests were performed at JPL. When the flight pots were received at JPL, each was given a room temperature functional test. First the total resistance of each pot was measured. Then 5 VDC was applied across the input terminals of each MIL-PRF-39023-style pot to energize the resistance element. While its bearing- supported shaft was manually rotated, the output voltage of each pot was plotted on a strip chart recorder. 304

Document

AMS_2006.pdf

Since the HGAG and WSA pots had no bearings, each one was placed into a functional test fixture and 5 VDC was applied across its input pigtails to energize the resistance element. While its rotor was manually rotated, the wiper or output voltage of each pot was plotted on a strip chart recorder. All pots were then placed in a vacuum chamber. Once evacuated the pots were subjected to three thermal cycles of -120” C to +110” C. Finally, each flight pot was given a second room temperature functional test. During the two functional tests, problems were noted with the following flight pots: MIL-PRF-39023-Style S/N’s 1035 and 1041, HGAG Pot S/N 1017 and WSA Pot S/N 1039 (Note: all S/N’s were randomly assigned by JPL). JPL Problem / Failure Reports (PFR) were initiated to describe the test failures. Shortly after the PFR’s were initiated the cognizant engineer for the MER pots left JPL and I assumed her role. Though I was experienced in the construction, cleaning and use of non-flight pots, I had no knowledge of the MER pot effort prior to my taking it over. The remaining three sections of this paper discuss each of the failure investigations and present lessons learned. MI L-PR F-39023-Style Pots The Initial Problem Functional testing of the MIL-PRF-39023-style pots before and after the three thermal cycles revealed: a) S/N 1035’s total resistance varied from approximately 5130 to 5720 Ohms as the shaft was turned, when nominally the value should have remained constant. In addition, there was an open circuit through the wiper terminal for all shaft positions, and b) S/N 1041 showed an open circuit through the wiper terminal for shaft positions that should have corresponded to voltage ratios between 0.0 and 0.1. After duplicating the original functional test setup and now taking the proper precautions (see Lessons 1 - 6), the problems were verified, the out-of-spec electrical performance was characterized and a “scratchy, detent” feel was noted as the shafts were rotated. Figure 2. X-ray images of the SIN 1041 pot at increasing magnification show wiper damage FeinFocus real-time X-ray images of the pots revealed wiper damage (Figure 2). Subsequent disassembly of S/N’s 1035 and 1041 revealed that the resistance element wiper, slip-ring wiper and the conductive plastic disk were all damaged due to high current and/or arcing (Figure 3). The wipers showed arc scarring: pitting coupled with discoloration and warping due to overheating (Figure 3). The end of one S/N 1035 wiper finger was burned off and embedded in its conductive plastic disk. The disks were damaged in two places: a) heavy arcing damage to the triangular shaped junction between the “CCW” end of the resistance element and the mound of conductive epoxy connecting the “CCW” terminal’s flat ribbon wire, and b) light arcing damage to the slip-ring portion of the disk where one of the two slip-ring wipers was located. The carbonized shell of a “bubble” was observed across the wiper tracks on S/N 1035’s conductive plastic disk where vaporized disk substrate “puffed up” the overheated top layers. There was a trough across the wiper tracks in S/N 1041’s conductive plastic disk where the substrate was vaporized away. 305

Document

AMS_2006.pdf

Figure 3. Arc damage to fingers on wipers, and to conductive plastic disks on S/N 1035 and 1041 Immediately after documenting the extensive damage in these pots, I began investigating probable causes. As I was not present during the original testing, I had to perform some detective work. I discovered that the manufacturer's rating for the current through the wiper circuit, though never specified in any of the MER pot specs, was only 10 mA. Whenever a wiper was positioned near the "CCW" end of the resistance element, a low resistance ( cc 500 Ohms ) circuit existed between the "CCW" and "W" terminals (Figure 4). If 5 VDC was applied between the "CCW" and "W" terminals when the wiper was positioned near the "CCW" end of the resistance element, a current well in excess of the manufacturer's rating would have flowed through the wiper circuit, damaging the pot. ccw TERM1 N AL (t W TERMINAL C W TERM1 NAL Figure 4. The wiper position shown forms a low resistance circuit between CCW and W terminals 306

Document

AMS_2006.pdf

The procedure that covered the functional testing of these pots read: Set multimeter to 10 kOhm - 20 kOhm range ... Perform the following steps for each potentiometer: 1. 2. 3. 4. 5. Remove each potentiometer from its shipping container. Record the resistance from the CW to the CCW terminals in the table at the end of this procedure. Resistance should be between 5 kOhm to 7.5 kOhm. Remove lead from CW terminal and reattach lead onto W terminal. Using a chart recorder, power supply set to 5V, and multi-meter, rotate the rotor of each potentiometer through its entire electrical range by hand. Attempt to rotate the rotor at a constant rate. Look for discontinuities or noise in the signal. Record the potentiometer part and serial number on each strip chart, and append the strip charts to this procedure. Replace each potentiometer into its shipping container. No electrical schematic was given and no polarities were mentioned. Note that step 3 left the pot's "CCW" and "W" terminals connected to test leads and implied rather than mandated the switching of the multimeter from the "Ohms" to "DCV" range. Step 4 connected a power supply but did not specify how, and though the power supply could deliver 3 amps, there was no instruction to set a current limit. Then rotating the pot ensured that the wiper would pass over the "CCW" end of the resistance element. The same set-up was used to test all pots. Had the power supply been wired across every pot's "CCW" and "W" terminals, all pots would have been damaged identically. However, the first seven pots tested using this set-up passed. SIN 1035 was the eighth pot tested but the first one damaged. S/N 1041 was the twenty-sixth pot tested, but only the second one damaged. Therefore, it can be concluded that nominally the power supply and instrumentation were wired correctly and that the power supply was not intentionally connected across every pot's "CCW" and "W" terminals. Close examination of the internal damage indicated that the wipers were stationary when the problem occurred. Note that Pomona Electronics brand "Minigrabber" test clips were used to connect to the pot terminals (Figure 5). When depressed, a beryllium-copper hook projects from the end of these test clips. As the procedure does not mention turning off the power supply outputs, the pots were likely hooked up to live (powered) test clips. The terminals on the MIL-PRF-39023 style pots are only 6mm apart. When attaching the powered clip meant for the "CW" terminal, it would have been very easy to bump the "W" terminal with the metal hook, or even clip it to the "W" terminal outright. If the wiper happened to be positioned near the "CCW" end of the resistance element, a high current could have flowed through the wiper circuit and caused the damage observed. Figure 5. The live, exposed hook on the test clip is a danger to pots with closely spaced terminals 307

Document

AMS_2006.pdf

How near is "near"? Ohm's Law says that at 5 VDC, 10 mA will flow through a resistance of 500 Ohms. If the resistance between the "CCW" and "W" terminals varies from 0 Ohms to the full 6250 Ohms over the 320 degree electrical range, then a current in excess of the wiper circuit's 10 mA rating would flow if the wiper was within 320 * (500 / 6250) = 25 degrees of the "CCW" end of the resistance element. If the shafts were in purely random orientations when connected, one would expect 25 / 360 = 7% of the pots to be damaged. Of the 31 MIL-PRF-39023-style pots connected, two (or 6.5%) were damaged. Based on the preceding failure analysis, this problem was attributed to test error due to a procedure fault. Pot S/N's 1035 and 1041 were scrapped. The pot specification and drawings were revised to include the 10 mA rating for the wiper circuit. The following lessons were learned: LESSON 1. Be aware of low resistance circuits which exist between the "CCW" and "W" terminals when the wiper is positioned near the "CCW" end of the resistance element, as well as between the "CW" and "W" terminals when the wiper is positioned near the "CW" end of the resistance element. Warn of these cofiditions in the test procedure. LESSON 2. In the pot specification and on the pot drawing, clearly state the current rating for the wiper circuit. LESSON 3. Based on a fraction of this rating, set current limits on all power supplies used in pot testing. LESSON 4. Provide an electrical schematic for connecting wires, and write detailed procedure steps. LESSON 5. Use insulated test clips to connect to terminals that are in close proximity to each other. Alternately, solder hook-up wires to the terminals and connect leads to the wires. LESSON 6. Turn off the power supply outputs when connecting test leads - do not hook live (powered) test leads to flight hardware. LESSON 7. Before they are allowed to play with electricity on the job, make sure all mechanical engineers and technicians can pass the "flashlight test" (hand them a 0-cell, a single piece of wire and a flashlight bulb and ask them to enlighten you). The RiDDk? Effect It would have been great to end the investigation there. However, every flight pot that underwent functional testing per the quoted procedure was potentially damaged and had to-be proven flightworthy all over again. It was not possible to tell from the X-ray images if the conductive plastic disk was damaged, therefore the only sure way to prove that a pot was undamaged was to remove its lid and conduct a visual inspection using magnification. Each lid was retained by three equally spaced radial set screws threaded into the wall of the cup which were tightened against the bottom of a circumferential groove machined in the lid. Loosening these setscrews should have allowed the lid to be removed. But disassembly of S/N's 1035 and 1041 showed that it wasn't quite so easy: a) one of the shaft bearings was mounted in the lid. Removing the lid thus represented major disassembly, unloading the preloaded wipers and possibly invalidating all the qualification testing performed to date, and b) the set screws used to retain the lid had no secondary locking feature, so the original cognizant engineer required that the manufacturer apply Solithane to the set screws to bond them in. The Solithane used by the manufacturer had wicked into the circumferential groove and effectively bonded the lids to the cups. Removing the lids would contaminate the wipers, tracks and bearings with Solithane debris, requiring complete disassembly for cleaning and re-lubricating. Time constraints necessitated an alternative though admittedly less rigorous approach. It was decided that the remaining flight pots would all be screened for any of the characteristics observed in the failed S/N's 1035 and 1041 , namely: a) out-of-spec electrical performance by electrical functional testing at room temperature, at the operating temperature extremes of -75" C and +70" C, and again at room temperature, and b) any "scratchy, detent" feel as each shaft was rotated. 308

Document

AMS_2006.pdf

A complete functional test required rotating each pot while monitoring its wiper terminal voltage. This test was quickly and easily done at room temperature, but having to test 28 pots at -75" C and +70° C in a minimum amount of time required a different approach. A gang fixture was designed and constructed with a stationary plate that supported up to 36 MIL-PRF-39023-style pots (Figure 6). A driveshaft turned a central crank arm which rotated a second plate eccentrically. The eccentric rotation of the second plate was converted into rotation of the individual pots by a crank arm on each pot shaft. Figure 6. Two views of gang fixture with 28 pots installed, before being placed in chamber This gang fixture was set up in a dry nitrogen purged thermal test chamber employing electrical heating and liquid nitrogen cooling. The driveshaft passed through an opening in the chamber wall to a gearmotor. With this set-up, electrical performance was rapidly verified at the temperature extremes in both the CW and CCW directions. The following lesson was learned: LESSON 8. Don't put all your eggs in one basket. Trying out a new test procedure on all the irreplaceable flight pots in one batch is risky. It is a great deal of work to recertify potentially compromised pots for flight. Instead prove out the test procedure with a few disposable EM pots. Complete the testing and data analysis, then revise the procedure with lessons learned before applying it to any flight pots. The Ripple Effect, Part II The next step was to check for any "scratchy, detent" feel as each pot shaft was rotated. This was a very delicate task as the torque required to rotate each ball-bearing supported shaft was at most a couple of millinewton-meters. With no sensitive torquemeters available, I turned off the laminar flow benches in the clean room to eliminate their noise and vibration, turned off most of the lights to get rid of the powerline buzz, and ripped the thumb and index finger off my right glove. One by one I took each pot in my hands and rotated the shaft back and forth with my bare fingers, eyes closed and concentrating. Soon I was writing another PFR to document the very weak detent that was felt as the shaft of: a) SIN 1027 was rotated by hand through the -190 degree CW shaft orientation, and b) SIN'S 1021 and 1036 were rotated by hand through the -240 degree CW shaft orientation In each case the shaft orientation where the detent was felt did not change, even after many shaft rotations; clearly it was not caused by a bad ball in the bearing or a bad inner race. The detent on SIN 1027 occurred at the -190 degree CW shaft orientation which placed the wiper in the electrical dead zone of the pot so damage from high current should not have been a problem. The detents on SIN 1021 and 1036 occur at the -240 degree CW shaft orientation which placed the wiper far enough away from the "CCW" end of the resistance element that damage from high current should not have been possible. Functional testing at room temperature, -75" C and +70° C had already verified the required electrical performance for these three pots. X-ray images revealed no macroscopic wiper damage, though it was not possible to tell from the X-ray images if the conductive plastic disk was damaged. How could I be

Document

AMS_2006.pdf

feeling what was supposed to be a symptom of arc damage, when arc damage was absent and/or impossible? And why did these detents feel much weaker than those on S/N's 1035 and 1041? This new problem had me stumped until the shafts were rotated to the detent positions and the pots were placed back in the FeinFocus X-ray machine. Close examination of the X-ray images showed that at the -190 degree shaft position on S/N 1027, the wiper bracket on the shaft passed very close to the flat ribbon wire connecting the "CW" terminal to the resistance element. On most pots, the flat ribbon wires were bent towards the case immediately upon exiting the hole in the conductive plastic disk. However, on SIN 1027, the "CW" flat ribbon wire was not bent towards the case, leaving it closer to the rotating parts than usual. Careful examination of X-ray images of S/N 1027 show that the wiper bracket rubbed the insulating Teflon sleeve on this flat ribbon wire. This rubbing interference was consistent with all observations listed above, represented a new discovery, and was declared to be the cause of the detent. The flat ribbon wires were welded to the ends of the terminals inside the pot case. However, from one pot to another there was no uniformity to the angle or direction that each flat ribbon wire pointed as it left its terminal (Figure 3). On S/N 1027, the "CW" flat ribbon wire left the "CW" terminal in a direction pointing away from its termination in the conductive plastic disk. Therefore, the flat ribbon wire had to be bent back upon itself in order to be inserted in its hole. Doubling the flat ribbon wire back on itself used up some of its finite length, and was likely why the "CW" flat ribbon wire wasn't bent towards the case when it exited its hole; there simply wasn't enough length available. r I L-i b Figure 7. X-ray images of pots show cause of detent: interfering Teflon sleeve on flat ribbon wire X-ray images of SIN'S 1021 and 1036 (Figure 7) also showed that their wiper brackets rubbed the insulating Teflon sleeves on their flat ribbon wires. This rubbing interference was consistent with all observations listed above, and was declared to be the source of their detents as well. Note that these three pots were never disassembled for direct visual verification because: a) the cover (lid) was glued on and prying open this lid would create debris which could contaminate the bearings and resistance element, and b) removing the lid with its integral bearing allows the shaft to tilt in the single bearing in the cup, preventing accurate checks of the clearances between the wiper bracket and Teflon sleeving, and c) since the detent-causing interference could be clearly seen from the X-ray images, it was not necessary to de-lid the pots. The three pots had been subjected to 5400 revs during testing by the manufacturer and then received another 300 revs at JPL during testing on the gang fixture. The Teflon sleeve had been rubbed by the metal wiper bracket 5700 times, deflecting the flat ribbon wire each time. The number of rubs that cause failure of the flat ribbon wire was unknown, so no prediction could be made as to how much life remained. 31 0

Document

AMS_2006.pdf

Based on the preceding failure analysis, this problem was attributed to manufacturing error due to a faulty production process. S/N's 1021, 1027 and 1036 were labeled "Limited Use (Non-Flight Part)" and were not used for flight. Except for S/N 1021, 1027 and 1036, all the other pots passed this screening process. Absence of any detent feel means only that the conductive plastic disk/wiper/bearings were not damaged and that there was some clearance between the rotating and stationary parts at room temperature. It was not a measure of how much clearance existed, nor was it a guarantee that clearance existed over the entire operating temperature range. However, it would have been impractical to check for detents at other than room temperature since detecting them requires a very sensitive touch in a quiet room with no vibration. This risk was accepted and the remaining pots were approved for flight use. Lessons learned include: LESSON 9. Sometimes you have to take off the gloves. In a quiet, vibration-free room, extremely subtle mechanical effects can be felt reliably only through bare skin contact. LESSON 10. An X-ray inspection machine with real-time imaging capabilities is an extremely valuable tool for performing non-destructive evaluation on assemblies that are sealed or for which disassembly would affect internal clearances or invalidate qualification testing already conducted. LESSON 11. Interferences within the MIL-PRF-39023-style pots are possible. The pot manufacturer should: a) individually specify and inspect the angle or direction that the flat ribbon wire points as it leaves each of the three terminals. The angle should be chosen to provide a direct route to each flat ribbon wire's respective hole in the conductive plastic disk. b) specify and inspect that each flat ribbon wire must be bent towards the case immediately upon exiting the hole in the conductive plastic disk. This would ensure that the flat ribbon wires will be against the inner wall of the case and maximize the clearance between the rotating and stationary parts. A cutaway cover or functional gage could be designed that would permit inspection and verification of a minimum acceptable clearance. c) investigate a shorter wiper bracket or an alternative "lower profile" method of insulating the flat ribbon wire, to provide more clearance between the rotating and stationary parts. Set Screwed While concentrating all my tactile attention on the flight pot shafts rotating between my fingers, I noticed that many were not smooth. Close examination revealed deep impressions with raised edges that made each precision 3.1 70 +0.000/-0.005-mm-diameter shaft reminiscent of a bastard-cut round file. Some EM pots showed this damage as well. During testing per the ATP, the manufacturer must have used a breaker bar to repeatedly tighten R, 52 hardened steel cup-point set screws into the Rb 82 annealed 303 stainless steel pot shafts. To correct this damage, the affected pots needed to have their shafts deburred with an India oilstone before I could deliver them to their end users. Deburring generates metal chips and abrasive fines, so to prevent contamination each pot was bagged and taped to expose just its shaft for deburring. After deburring was complete, the shafts were cleaned and the bags removed. The following lesson was learned: LESSON 12. If for testing purposes EM pots are ordered and received long before the flight pots, then immediately upon receipt the EM pots should be carefully inspected and the manufacturer should be contacted about quality issues like precision shafts marred by set screws. Before fabricating the flight pots the manufacturer should be instructed to tighten setscrews against the shaft's machined flat and not its precision outer diameter, to properly control setscrew torques, to use brass or nylon tipped setscrews, and/or to switch to non-setscrew style couplings on their test equipment. Observed Trends Continuing the investigation, the total resistance measurements captured my attention next. A plot of the total resistance of the MIL-PRF-39023-style pots showed it to be decreasing with either time and/or thermal cycles. The plot in Figure 8 shows the history of these total resistance measurements with each line representing a different serial number pot. The spec for these pots was 6250 Ohms f 20%, or 5000 to 7500 Ohms. There was minimal drift measured by the pot manufacturer before and after their thermal 31 1

Document

AMS_2006.pdf

testing (measurements #1 and 2), but significant drift was measured after each JPL thermal test (measurements #3 - 6). Discussion with the original cognizant engineer responsible for measurements #3 and 4 revealed that the ohmmeter was never zeroed, which would have put a constant offset error on all values measured. If that offset had been approximately -170 Ohms, there would be a uniform downward trend across nearly all measurements. I became concerned that if this trend continued, the daily thermal cycles experienced by the Rovers might cause the total resistance of the pots to drift out of spec. 0 1 2 3 4 5 6 7 Measurement Number Figure 8. Total Resistance measurement history When notified of this observation the pot manufacturer said that it was typical for the total resistance of a conductive plastic pot to change after thermal cycling. The mechanism was explained as follows: conductive plastic is made of carbon particles in an insulating matrix, co-cured at 160" C (and rated for operation up to 125" C). Thermal cycles change the insulating matrix on a microscopic level, allowing better conductivity between the carbon particles. Macroscopically these changes appear as a drop in the total resistance. The manufacturer said that as the number of thermal cycles builds, the rate at which the resistance drops will decrease and the resistance will eventually stabilize. The scratches and drops of silver paint used to linearize the pot have an effect on the rate at which the resistance stabilizes. The manufacturer assured me that the resistance changes would be uniform throughout the length of the resistance element and that the pot's linearity would not be affected by changes in the total resistance. Interestingly enough, the pot manufacturer also said that just leaving the pot on a shelf for a time after thermal cycling will cause the pot to recover some, but not all, of the resistance loss which occurs during thermal cycling. They called this a "memory" effect. LESSON 13. To gain insight, plot every measured quantity and look for trends. If a measurement is worth recording, it is worth plotting. LESSON 14. Calibrate test equipment before using it on flight hardware. The data is only as good as the calibrations; i.e., always "zero" an ohmmeter before measuring any resistance. LESSON 15. Manufacturing considerations and physical mechanisms are the reason a precision pot has the relatively loose tolerance of * 20% on its total resistance. Add Vent Holes. Then Glue Them Shut MIL-PRF-39023-style pot cups trap air so the MER Environmental Requirements Engineer requested a venting analysis. I did some digging and found that during a design review early in the procurement process, the original cognizant engineer had requested that a vent hole be added, so the manufacturer drilled one in the cup between the set screws that hold the lid on. Unfortunately, the original cognizant engineer had also requested that a Solithane mixture be applied to the set screws to lock them in place. When this Solithane mixture wicked into the circumferential groove and effectively bonded the lids to the cups, it also blocked the vent path leading to the drilled vent hole. The only vent path left was the one through the shaft bearing which is undesirable because vented air often carries contamination that can become lodged in a bearing. 312

Document

AMS_2006.pdf

LESSON 16. When a design uses one feature for multiple purposes (like the circumferential groove in the lid which was used both to retain the lid and to provide a labyrinthine path for venting), make sure all the effects of even the smallest design change are well understood before approving it. Turnina Gold Into Base Metals Prior to delivering the pots to the end users I noticed that the original cognizant engineer ordered the MIL- PRF-39023 pots manufactured to a drawing which specifies that its solder terminals are made from brass which has first been silver plated, and then very thinly gold plated. JPL’s internal requirements state that gold shall not be used as a surface finish for soldering, so I instructed the end users of the MIL-PRF-39023-style pots to “degold the terminals before soldering. LESSON 17. Soldering to gold can form brittle intermetallic compounds that can crack under thermal cycling. Gold plated conductors must have their gold plating removed by immersion in a designated solder pot, then tinned using the same processes in a different solder pot. Alternatively, gold plated leads must be tinned twice with solder wire, wicking off the solder in the first tinning to remove the gold, then tinning again. Trust Nobodv With Your Fliaht Hardware About this time MER’S Deputy Mission Assurance Manager requested a private meeting to better understand the pot details described in the PFR’s. Afterward this manager asked to borrow several of the pots that had been written up, and I foolishly obliged. When I asked for my pots back, S/N 1036 could not be found: Mission Assurance had lost my flight hardware! LESSON 18. When cognizant of flight hardware, especially small hardware, never let it out of your sight, and never, ever lend it to management. HGAG Pots Functional testing of the HGAG pots before and after the three thermal cycles revealed that on S/N 1017 the total resistance varied from approximately 5440 to 6190 Ohm as the rotor was rotated. The total resistances of the other HGAG pots tested did not vary with rotation angle. Nominally the total resistance should remain constant. After duplicating the original functional test setup and now taking the proper precautions (see Lessons 1 - 6), the problem was verified. Next the build documentation was examined to determine if it held any clues to this anomaly. The pot manufacturer’s ATP describes the resistance check, but does not mention rotating the rotor, nor does it specify that the total resistance must not vary with rotor angle. The spec only called for a total resistance of 6250 Ohms * 20%. The in-process data sheet filled out by the pot manufacturer listed the total resistance as 6046 Ohms. After environmental testing, the in-process data sheet listed the total resistance as 6121 Ohms. 6121 Ohms represented a 75-Ohm increase which was not in keeping with the rest of the pots in that lot which showed total resistance changes of no more than 5 Ohms. Random rotor angles during total resistance measurement explained the 75 Ohm change, and indicated that this problem may have been inherent at assembly. S/N 1017 was still a flight pot, so I was in no rush to tear it apart. Besides, FeinFocus X-ray images of SIN 101 7 revealed no internal damage. Instead I elected to do more testing. The pot was powered with 1.0 VDC across the input wires and the wiper voltage was measured and recorded as the rotor was rotated through 360 degrees in 5 degree increments. The wiper voltage was indeed the required linear function of rotor angle. Next the total resistance was measured and recorded as the rotor was rotated through 360 degrees in 5 dagree increments. The total resistance was a smoothly decreasing function of rotor angle (Figure 9). 31 3

Document

AMS_2006.pdf

Figure 9. X-ray image of undamaged wiper and slip ring brushes in HGAG pot After plotting and examining the data, I hypothesized that a high resistance leakage path between the “CCW” end termination of the resistance element and the slip ring could explain the anomaly. A schematic of a pot with an internal leakage resistance was drawn (Figure 10) and an equivalent resistance equation was written as a function of shaft angle. This mathematical model was iterated with different values of leakage resistance until the model predictions matched the measured total resistance data. Note that the 51.9 kOhm leakage resistance found in this way was 8.3 times the pot’s nominal total resistance. -I 0 50 100 150 200 250 300 350 Rotor Angle, CCW cw Figure 10. Measurements match model of HGAG pot with internal leakage resistance RL =51 kOhm The prediction of the model fits the data extremely well, supporting the hypothesis. The extremely linear output of this pot supports the conclusion that this leakage resistance was present all along but never discovered (because per a strict reading of the ATP instructions, no one was looking for it). Any non- linearity caused by the leakage resistance was likely removed by adding scratches and/or drops of silver paint when the pot’s linearity was adjusted by the manufacturer. True confirmation of the hypothesis requires direct measurement and so the pot was disassembled by removing the rotor from the stator. In this condition the leakage resistance could be accurately measured because the slip ring and the resistance element were no longer bridged by the brushes. With the pot disassembled, the leakage resistance measured 51.4 kOhms to the “CCW” lead, and 57.6 kOhms to the “CW” lead. Note that 57.6 - 51.4 = 6.2 kOhm, the nominal total resistance. The resistance element showed a very large number of linearizing marks, with all the silver paint dots grouped in the middle of the element and all the scratches at the ends near the junctions; exactly what one would expect if the pot had been adjusted to get rid of a manufactured-in leakage resistance. It is reasonable to ask what caused the leakage resistance. There was a concern that it could have formed from conductive debris worn off the surface of the resistance element or slip ring by a rough wiper, 31 4

Document

AMS_2006.pdf

and scattered between them. However that proposed mechanism was eliminated when upon disassembly the resistance element and slip ring showed no signs of wear, scraping or dusting. The conductive plastic disk was made of carbon particles in an insulating matrix, co-cured at 160" C. It is believed that if some conductive carbon particles in the junction area were displaced from the resistance element and co-cured in place, a high resistance leakage path between the element and the slip ring would have resulted. While the discovery of the varying total resistance was significant, S/N 1017 has always been within tolerance. Testing verified its linearity, and though it varied, the total resistance was within the 6250 * 20% (5000 - 7500) Ohm allowable range. Of course, this leakage resistance does waste a bit of power. Based on the preceding failure analysis, this problem was attributed to manufacturing error due to a faulty production process. Pot S/N 1017 was labeled "Limited Use (Non-Flight Part)" and was not used for flight. LESSON 19. Being within tolerance does not mean a pot is problem free. LESSON 20. Unique behavior is always worth investigating until it is understood, to make sure that the problem does not affect other pots in the lot. WSA Pots During functional testing of the WSA pots before the three thermal cycles, the total resistance of WSA Potentiometer SIN 1039 measured 7400 Ohms. After the three thermal cycles, the total resistance dropped to 6900 Ohms. Compared to the other WSA pots in the lot, this change was a full order of magnitude larger than expected. After duplicating the original functional test setup and now taking the proper precautions (see Lessons 1 - 6), the problem was verified. Next the build documentation was examined to see if it held any clues to this anomaly. The spec called for a total resistance of 6250 Ohms * 20%. The in-process data sheet filled out by the pot manufacturer listed the total resistance as 7391 Ohms. After environmental testing, the in-process data sheet listed the total resistance as 7404 Ohm. While S/N 1039 was in spec, all the other WSA potentiometers supplied to JPL had much lower total resistances. A histogram of WSA total resistance values (Figure 11) clearly shows SIN 1039 to be an outlier. Total Resistance, Ohms Figure 11. Histogram of Total Resistance values showing the lone outlier, S/N 1039 SIN 1039 was still a flight pot, so I was in no rush to tear it apart. Besides, FeinFocus X-ray images of S/N 1039 revealed no internal damage. Instead I elected to do more testing. The pot was powered with 1.0 VDC across the input wires and the wiper voltage was measured and recorded as the rotor was rotated through 360 degrees in 5 degree 31 5

Document

AMS_2006.pdf

increments. The wiper voltage was indeed the required linear function of rotor angle. Next the total resistance was measured and found to be 6890 Ohm. Note that this was within the 6250 f 20% (5000 - 7500) Ohm tolerance allowed for this potentiometer. When notified of this anomaly the manufacturer said that it was typical for the total resistance of a conductive plastic pot to change after thermal cycling. The mechanism was explained as follows: conductive plastic is made of carbon particles in an insulating matrix, co-cured at 160” C (and rated for operation up to 125” C). Thermal cycles change the insulating matrix on a microscopic level, allowing better conductivity between the carbon particles. Macroscopically these changes appear as a drop in the total resistance. The manufacturer said that as the number of thermal cycles builds, the rate at which the resistance drops will decrease and the resistance will eventually stabilize. The scratches and drops of silver paint used to linearize the pot have an effect on the rate at which the resistance stabilizes. The manufacturer assured me that the resistance changes would be uniform throughout the length of the resistance element and that the pot’s linearity would not be affected by changes in the total resistance. The eight flight WSA pots went through additional thermal cycling as part of the steering actuator qualification test program. After those thermal cycles, the total resistances were found to have changed by +73 to -283 Ohms. Another data point comes from the MIL-PRF-39023-style pots: between manufacture and use, the total resistance of each one dropped from 200 to 400 Ohms. From the beginning, WSA Potentiometer S/N 1039 was an outlier, with a total resistance 750 Ohms higher than its nearest neighbor. When S/N 1039 was 5 weeks old, it took advantage of a thermal cycle to drop its total resistance by 500 Ohms to a value more in line with, but still higher than, the rest of the batch. While this drift was significant, 1039 has always been within tolerance, and testing proves its linearity. Based on the preceding failure analysis, it was determined that there was no problem with the pot. However, just to be safe, pot S/N 101 7 was labeled “Limited Use (Non-Flight Part)” and was not used for flight. LESSON 21. Drifting values and excessive variation within a lot may indicate trouble, even if the individual pots are within spec. Compare the pots to each other as well as to the requirements. Conclusions As of this writing, the “Spirit” and ‘‘Opportunity’’ Mars Exploration Rovers have been operating on the Martian surface for over two years. In spite of the difficulties encountered during flight qualification of the MIL-PRF-39023-style pots, the HGAG pots and the WSA pots, all pots on the Rover continue to perform reliably under very harsh conditions. It is hoped that the lessons provided in this paper will help other potentiometer users enjoy similar successes. References 1. Iskenderian, T. “Lessons Learned from Selecting and Testing Spaceflight Potentiometers.” Aerospace Mechanisms Symposium #28, (1 994), pp. 339-358. 2. Defense Logistics Agency, “MIL-PRF-39023B Resistors, Variable, Nonwire-wound, Precision, General Specification For.” Feb 1999. 3. “VRCI-P-1 OOA, Industry Standard for Wirewound and Nonwirewound Precision Potentiometers, Terms and Definitions, Inspection and Test Procedures.” Variable Electronic Components Institute, Vista, CA, http://www.veci-vrci.com/, 1 988 31 6

Document

AMS_2006.pdf

Mechanism Development, Testing, and Lessons Learned for the Advanced Resistive Exercise Device Christopher D. ~amoreaux* and Mark E. Landeck* Abstract The Advanced Resistive Exercise Device (ARED) (Figure 1) has been developed at NASA Johnson Space Center, for the International Space Station (ISS) program. ARED is a multi-exercise, high-load resistive exercise device, designed for long duration, human space missions. ARED will enable astronauts to effectively maintain their muscle strength and bone mass in the micro-gravity environment more effectively than any other existing devices. ARED's resistance is provided via two, 20.3 cm (8 in) diameter vacuum cylinders, which provide a nearly constant resistance source. ARED also has a means to simulate the inertia that is felt during a 1-G exercise routine via the flywheel subassembly, which is directly tied to the motion of the ARED cylinders. ARED is scheduled to fly on flight ULF 2 to the ISS and will be located in Node 1. Presently, ARED is in the middle of its qualification and acceptance test program. An extensive testing program and engineering evaluation has increased the reliability of ARED by bringing potential design issues to light before flight production. Some of those design issues, resolutions, and design details will be discussed in this paper. Figure 1. Test subject performing a squat on ARED NASA Johnson Space Center, Houston, TX Proceedings of the 3@ Aerospace Mechanisms Symposium, Langley Research Center, May 17-19,2006

Document

AMS_2006.pdf

Introduction and ARED Background This paper will discuss the design, development, and testing of the Advanced Resistive Exercise Device for the International Space Station program. ARED is scheduled to fly on shuttle flight ULF2 and will be located in Node 1 of ISS. ARED is a multi-format resistive exercise machine specifically designed for a zero-gravity environment. Capable of 30 different exercises, ARED will be used daily by the astronauts aboard the ISS to counter the loss of muscle and bone mass associated with long duration human space missions. ARED is being developed at NASA’s Johnson Space Center in the Biomedical System Division. The team consists of both civil servant and contractor engineers from many different organizations at JSC. ARED has been developed specifically to improve the on-orbit resistive exercise capability, reliability, and availability. The resistive force is generated by two, 200-mm (8-in) diameter dynamic vacuum cylinders. A vacuum exists on one side of the piston and atmospheric pressure exists on the other. The piston has a stroke of 30.5 cm (12 in) inside the cylinder. The cylinders are capable of delivering a nearly constant load. In combination with various mechanisms, ARED provides a range of 0 - 272 kgf (0 - 600 Ibf) to the exerciser, which is important effective zero-g exercise. In addition, the flywheel mechanism simulates the inertial force component of lifting free weights in a 1 -G environment. ARED thus provides a more complete weightlifting experience than any previous on-orbit resistance exercise device. It will allow astronauts to perform a wider variety of exercises at higher loads, higher speed, and longer stroke which, in turn, will enable them to maintain their health more effectively in a zero-gravity environment. The need for ARED, arose out of reliability and performance concerns with previous resistance devices for ISS. ARED is being designed, tested, and certified for a 15-year service life, which is much longer than any previously designed device. This long service life is needed to support long duration space missions, which will require a robust and reliable weightlifting machine. ARED’s predecessor, IRED, has a 0 - 136 kgf (0 - 300 Ibf) load range whereas ARED will enable the crew to exercise up to 272 kgf (600 Ibf). ARED will allow for a wide range of both bar exercises (squat, dead lift, heel raise, etc) and cable exercises (hip abductors, one-arm curls, etc). The load and stroke capability for bar exercises is 0 - 272 kgf (0 - 600 Ibf) and a 76.2-cm (30-in) stroke. The load and stroke capability for cable exercises is 0 - 68 kgf (0 - 150 Ibf) and a 183-cm (72-in) stroke. ARED’s vacuum cylinders have a nearly constant loading profile, which is more medically advantageous than the varying loading profile that is provided by springs and rubber based exercise devices. In addition to providing more constant load than IRED, ARED attempts to simulate the inertia that is felt during free-weight, 1 -G exercise by employing a flywheel that is directly tied to the motion of the cylinders. The differences are shown in Table 1. Table 1. Comparison chart of ARED vs. IRED 31 8

Document

AMS_2006.pdf

Framc I Main Arm/ Figure 2. ARED - Man in the Loop Testing Unit - Profile and Front View General ARED Mechanism Overview ARED has seven main subsystems (Figure 2); vacuum cylinders, flywheels, frame, platform, arm base, cable-pulley, and the main armllift bar. A motion schematic for a squat on ARED is shown in Figure 3. The vacuum cylinders are the main generators of resistive force for ARED. The details of the cylinders will be discussed fully in the next section. The piston rods are attached to the arm base assembly. Motion I$ Figure 3. Schematic of ARED showing motion and simple free body diagram

Document

AMS_2006.pdf

The main purpose of the arm base assembly (Figure 4) is to provide a mechanism to adjust the load setting. A ball screw in the arm base provides this function. By turning the ball screw, the attachment point of the piston rods moves along the ball screw. This changes the moment arm between the pivot point and the cylinder load application point and allow for load adjustment at 1-lb increment. The arm base assembly transfers the load from the cylinders to both the bar and cable exercise hardware. The dual nature of the arm base minimizes the reconfiguration required to switch between the two types of exercises, resulting in a more efficient exercise routine. Details of the arm base will be discussed later. I Figure 4. ARED Arm Base Assembly (Top Cover Removed) The flywheel assembly is mounted to an end cap of the cylinder and directly meshes with the motion of the piston by means of a gear and gear rack. The details of the flywheel assembly will be discussed in the next section. The main arm and lift bar assembly (Figure 2) transfers the load from the arm base to the exerciser. The main arm assembly lifts the front end of the arm base using a contact surface. The lift bar portion of the assembly allows for bar adjustment from 25.4 cm (10 in) above the platform to 183 cm (72 in) above the platform. This wide range of adjustment allows for any bar exercise from a dead-lift to a squat and accommodates a range of human subject from 5Ih-percentile Asian female to 95Ih-percentile American male. As with free weights, the position of the bar can also be “racked using the upper stop mechanism. The upper stop mechanism allows the crew to start the squat and heel-raise exercises from a standing position. The cable-pulley assembly also transfers the load from the arm base to the subject. The cable arms push down on the rear portion of the arm base assembly (Figure 5). The cable-pulley assembly enables a variety of cable exercises such as one-arm cable row, hip abductions, and one-arm curls. It employs a series of pulleys, timing belts, and cables to achieve this function. One of the pulleys in the assembly is cammed to compensate for geometry changes during the stroke in order to create a constant load at the end of the exercise rope. The pulley ratios provide a maximum of 136 kgf (300 Ibf) and 183 cm (72 in) of stroke. 320

Document

AMS_2006.pdf

Cable Arm to Arm Base Interface -1 Figure 5. Detail showing the cable arms pushing on the back of the arm base assembly Figure 6. ARED in the folded configuration Front view shows the details of the cable pulley mechanism. The platform (Figure 7) assembly's main purpose is to provide an adequate exercise surface and to provide containment for some of the pulleys for the cable-pulley mechanism. Its secondary purpose is to house load cells and wiring for the instrumentation system. The platform and main arm also can be folded up (Figure 6) to aid in storage and crew translation on ISS. 32 1

Document

AMS_2006.pdf

Figure 7. Picture of the platform CAD model with the footplates removed. This shows the electronics inside the platform assembly. Vacuum Cylinder Design Details There are two vacuum cylinders (Figure 8) used on ARED. Each cylinder provides a constant 340-kgf (750-lbf) load. The design uses a standard cylinder/piston concept with endplates and four tie rods holding the assembly together. The original cylinder shell was made of 6061-T6 AI, anodized on the outer surface, and left bare on a 16 RMS ID surface. Braycote 601 lubricant was applied to the entire ID of the cylinder shell. Early vacuum cylinder tests indicated that the surface finish of the interior surface, measured in RMS, is the critical parameter in maintaining a vacuum, while the piston is moving. The current flight design calls for a 20.3- cm (841) ID with an 8 RMS or better surface finish, 4.8-mm (0.19-in) wall thickness, and is 38.1 cm (15 in) in length. To achieve the best possible surface finish, three different manufacturing methods were tried; ground and honed, electro polish, and hand polish. Results showed that the ground and honed process provided a more consistent and controlled surface finish while maintaining the required roundness. To date, the best surface finish achieved is less than 1 RMS on an aluminum 6061-T6 cylinders, manufactured at the Micro-machining Department at NASA’s Glenn Research Center. The current flight piston design is made from 6061-T651 AI using a Nylon Molygard wear strip and a self lubricating Carboxilated Nitrile U-ring seal from Parker Hannifin Corporation. The open end of the U-ring seal is oriented toward the pressure side of the piston. Braycote 601 lubricant is applied to the wearstrip and u-ring seal. Three different wearstriph-ring seal combinations were evaluated to achieve the most efficient vacuum under dynamic conditions. 1. A single wearstrip with a single u-ring seal. 2. A single wearstrip with 2 u-ring seals. 3. A single wearstrip with a spring-energized u-ring seal. Through a series of tests, it was determined that option 1 worked well as long as the manufacturing and assembly tolerances (Figure 8) were controlled. With such a large diameter piston/cylinder and long stroke, it was important to maintain perpendicularity and parallelism between the piston, piston shaft, and cylinder wall. A jig was designed to assist in assembling the cylinder and piston to the required alignment. 322

Document

AMS_2006.pdf

Figure 8. Piston/Cylinder Design The bottom end cap design, also 6061, includes a Linear Bearing packed with Rheolube 2000 grease. This bearing interfaces with the piston shaft made of 15-5 PH stainless steel and heat treated to a H1025 condition. The shaft has a circular cross section the length of the piston stroke that interfaces with the linear bearing. The remaining portion of the shaft is rectangular for attaching a gear rack used to drive the flywheel. The bottom cap design includes a Fluorocarbon rubber bumper and a 3.2-mm (0.125-in) thick polyester grade polyurethane foam filter. The foam filter was added to the bottom end cap design to restrict airborne debris from being sucked into the cylinder. The rubber bumper prevents piston damage in the event of bottoming out during assembly (Figure 8). The head cap, on the vacuum side of the piston, is 6061 AI and is sealed with the cylinder shell using a standard Butyle rubber O-ring seal. The head cap also includes a relief valve and a fluorocarbon rubber bumper. To evacuate the cylinder, the relief valve is opened and the piston is pushed to the top of the stroke and bottomed out against the rubber bumper. The valve is then closed (Figure 9), and a vacuum is established. I I Venting/Evacuation Figure 9. Assembled Cylinders showing top caps and valves for venting/evacuation 323

Document

AMS_2006.pdf

Design Details - Flywheel Assembly The flywheel assembly is the component of ARED that simulates the inertial effect of weightlifting in 1-g environment. When lifting free weights in 1-g, the total foot reaction force is dependent on the acceleration of weights on the exerciser's shoulders. While performing a squat, this inertial effect creates a spike in load at the bottom of the stroke and load relief at the top of the stroke. The exercise physiology community speculates that this inertia spike plays a major role in increasing bone density in 1-g and, by the same effect, slows the rate that bone density is lost in a zero-g environment. -- . - I' I Figure 10. Picture showing the over-center mechanisms, in the locked position, between the flywheel and cylinder assembly The flywheel assembly mounts to one end of the cylinder assembly. It attaches using a hinge and an over-center mechanism (Figure 10). The purpose of this attachment method is to allow for the flywheel mechanism to be engaged and disengaged depending on the exercise and the preference of the user. A U-link (Figure 11) deforms during actuation and allows for the mechanism to go over-center. The design of the U-link was challenging, because it had to elongate by 0.43 mm (0.017 in) and still be under the allowed stress limits. After using the first prototype, it was obvious that the tolerance stack-up between the hinge and the over-center mechanism made a significant difference in the over-center force. In order to better control the over-center force, an eccentric spline bushing (Figure 11) was used in the over- center handle. The hole in the bushing is 0.64 mm (0.025 in) off-set from the center of the bushing. The offset allows for small adjustment of the over-center force. 4014171 4 01 7717 4015334 Figure 11. Detail of spline bushing and displacement analysis of the over-center u-link. The heart of the flywheel assembly is a set of three gears and two inertial flywheels (Figure 12). The spur gear meshes with a spur gear rack on the piston rod which makes the motion of the flywheels directly tied to the motion of the cylinders of ARED. The flywheels provide the most load and rotate the fastest at the higher load settings of ARED due to the longer cylinder stroke. This flywheel design uses a fixed gear 324

Document

AMS_2006.pdf

ratio and a fixed mass for the flywheels. While this design is simple, it only allows for the inertia to be tuned to one combination of exercise subject mass, exercise subject deceleration, and free weight mass. The ARED flywheels do not account for the changes in the exerciser subject’s mass. They do, however, adjust to the exercise subject’s deceleration and free weight load. The flywheels add more load into the system at higher load settings, because the piston moves further per stroke, in the same amount of time. Also, a faster deceleration of the exercise subject causes the flywheel to decelerate faster as well. This increases the inertial load felt at the lift bar. In the beginning of the project, the flywheels were tuned to a 227-kgf (500-lbf) squat setting on ARED with a 2 second period and 76.2-cm (30-in) stroke. After ARED is on-orbit for a significant amount of time, the flywheels could be tuned to a more optimal load setting and stroke. This would need to be determined from statistical analysis of the exercise frequency, load, and stroke data from on-orbit use. A chart showing the calculated variation from the true 1 -G inertia is below (Figure 13). The assembly contains one spur gear and two helical gears. The spur gear has a 38-mm (1.5-in) pitch diameter, the large helical gear has a 127-mm (5.0-in) pitch diameter, and the small helical gear has a 38- mm (1 5in) pitch diameter. The spur gear is made out of 17-4 PH SS and the teeth are also ion nitrided to provide a longer life of the gear teeth and to prevent pitting. The helical gears experience lower loads and are not ion nitrided. The initial prototype of the flywheel assembly used two lubrication methods to determine which would perform better. One gear set employed standard lubrication (Rheolube 2000) while the other had a ceramic dry-film lubrication called Vitro-lube. Our initial life-cycle test showed that standard lubrication was more reliable. Early in our life-cycle test, the Vitro-lube began to flake off significantly. As a result, the team opted to go with standard lubrication. Spur Gear- Figure 12. Flywheel assembly with front cover removed and close up of gears and flywheels. The spur gear is not visible. A cross-section view through the spur gear teeth is also shown for clarity. 325

Document

AMS_2006.pdf

The flywheels are 14 cm (5.5 in) in diameter and 4 cm (1.6 in) thick, stainless steel. As mentioned previously, their inertial properties and the gear ratios were specifically chosen for the 227-kgf (500-lbf) load setting on ARED. The flywheels can see a maximum speed of 800 rpm. Therefore, when the exerciser reaches the bottom of the stroke, and stops the flywheels from spinning, the flywheels could deliver a maximum of 91 kgf (200 Ibf) extra at the lift bar. This would occur at the 272-kgf (600-lbf) load setting, 76.2-cm (30-in) stroke, and a 2 second period. It is highly unlikely that anyone will exercise at this load, speed, and stroke. However, these parameters were the maximum range of the project’s requirements. In order to protect the exerciser and the ARED hardware from overloading, the flywheels are directly in line with a friction disk slip-clutch. This slip-clutch is set to slip at 3.39 Nom (30 inolbf) of torque which is less than 10% more torque than the maximum load described in the previous paragraph. The slip-clutch eliminates concerns of excessive loading being imparted on the device. Both Percentile (162 Ibm) Person Petforming a Squat of 30 in Stroke with a 2 Second Period 600 IY tMallrtt might bat162 lbm AStOMUt Ptrtorming a 138 Ibf squat in 1.G I 162 Ibm Astonaut Performing a squat, in 0-0. on AREDvilhat600IMLordSctting 500 IM total free might bad: 162 Ibm ARonau Perfoming a 338 bl squat in 14 -20 LY) Ovrr Desirrd Ine I 9 c Desired Inertia A 300 Ibf total lree weight load: 162 Ibm ARonau Perhrning a 138 IM squat in 1.0 162 Ibm Aslonau Performing J quat. in 0-Gon AREDvirhat300IMLoadS.tling 0 200 Ibf totallrrrmightload:162IbmAaonnn Pwforming a 38 IM squat in l.0 I 5% (4 LY] Undrr Desired Inertia Desired Inertia 162 Ibm Astonaut Perlonning a squat. in C-G. on ARED vilh at 200 IM Load Setting 162 IM tMal free weight bat 162 Ibm AStOMut Perlorrnhg a 0 bl squat in 1-0 162 Ibm Astonaut Perlorming a squat, in 0.6 on ARED with a 162 kl Load Setting -- \I 162 Ibm AstOMut Peffom*1g a squat. in 0-G. on 2x13 LM) Undrr Desired Inertia ARED vilh u 500 IY Lord Stttiiy 4~) IY tor4 tree might bat 162 ~bm ~sroMut 162 Ibm A~OMW Performing a squat, in 0-G on . - e Performing a 238 Ibl swat in 1.0 3% [14 Lbf] Under 1 AREDwlhat4WIMLoadSeniiy - Desied Inertia I I 0 100 2w 300 400 500 Foot Reaction Load (Lbf) 600 700 800 900 mConstant ARED Load Component in C-G OARED Body Inertia Component in 0-G .Free Weights I-G Inertia Component .Constant Body Weight Component in 1-G oConstant Free Weight Component in 1-G .Flywheel Inertia Component in 0-G H Body Weight 1-G Inertia Component oload Short or Over Goal Figure 13. Chart comparing the load contributions during 1-G free weight squat exercise and squat exercise on ARED in a 0-G environment. This chart is for a 50th percentile astronaut performing a 76.2-cm (30-in) squat with a 2 second period at 45.4-kgf (100-lbf) increments. It shows that the inertial variation between ARED exercise and 1 -G exercise is relatively small. Design Details - Arm Base Slider Each cylinder shaft is attached to the arm base slider through a rod-end with a spherical bearing. The slider is attached to a ball screw, in the arm base, with a slider housing (Figure 14Figure 16Figure 17). The ball screw is used to adjust the exercise load of ARED. This is done by changing the moment arm between the pivot point of the arm base and where the cylinders attach. A scale and position indicator is provided on the arm base cover to aid in this load adjustment. The attachment to the ball screw uses a housing/slider design that follows curved tracks (Figure 17) on the sides of the arm base with cam rollers. 326

Document

AMS_2006.pdf

The track is curved to minimize force non-linearity as a function of load adjustment. The slider moves up and down inside the housing, which is attached to the ball screw, as the slider follows the tracks (Figure 17). The life cycle design requirement for ARED is 15 years. A preliminary test was run on an engineering unit of the arm base and cylinders at various load settings. During this test, at a 181-kgf (400-lbf) load setting, the slider housing broke into 2 pieces (Figure 15) at 31 8,000 total cycles. After close evaluation, it was determined that the stress analysis overlooked a critical load case. There was more moment on the housing than first calculated and a fatigue analysis had not been done. As a result of the improved analysis, the slider housing was thickened and strengthened with closeout plates. The original housing was made of 7075 AI with a Tufram surface coating to increase surface hardness, extend wear, and reduce friction. During the test, the slider did not slide very well along the Tufram surface. As a result, rollers were also added to the slider design to reduce friction and to implement a zero clearance fit between the slider and housing. With the new rollers added to the slider, steel wear plates had to be added to the inside of the slider housing (Figure 16), to increase the life of the assembly. I' I i II i Figure 14. Old SliderRlousing Design for Arm Base Figure 15. Old Slider/Housing design during life cycle test showing broken Housing 327

Document

AMS_2006.pdf

Slider w/ Rollers '1 Figure 16. New Slider/Housing Design showir,, Glider w/ rollers and Hnfilcirlg wear plates a L 4f ., I \ ~ . Curved Track -a .. - h . Closeout Plates Figure 17. New Slider/Housing design attached to ball screw showing closeout plates The team also performed an engineering life cycle test on the arm base/load adjustment components. This test rotated the ball screw and drove the rod ends back and forth along the length of the arm base. The purpose of this test was to prove the critical components in the arm base assembly. This test was a success and cycled the ball screw for 33,000 m (1.3 million in) of travel, before failure. Assuming a six person ISS crew and various uncertainty factors, this would be enough ballscrew life to last for -1 4 years 328

Document

AMS_2006.pdf

on-orbit. According to Nook, the manufacturer of the ball screw, the failure was a classical fatigue failure of a ball screw and ball nut assembly. Testing - Development As has been discussed, ARED is well into the testing phase of the project. Below is a summary of ARED’s testing program. Preliminarv Life Cvcle Units Two life cycle test rigs were built. Rig 1 was intended to exercise the cylinder/flywheel, the main arm, the arm base, and the pivot arm. Rig 2 was intended to exercise the ball screw and the arm base slider housing. Rig 1 was used to evaluate different design parameters, especially in the piston/cylinder interface. At the end of development phase, rig 1 has accumulated 320,000 cycles. Rig 2, as mentioned in the previous section, produced 1.3 M inches on travel on the ball screw. Both Rig 1 and Rig 2 will be modified and upgraded to serve as the qualification lifecycle test beds. Hi-Fidelitv Enuineerinu Unit Lessons learned from testing on Rig 1 and Rig 2 were incorporated into the design. To further increase confidence in the design, a Hi-Fidelity Engineering Unit was constructed, using near-flight-quality drawings. This Hi-Fidelity Engineering Unit was used for Man-In-The-Loop testing, with the intent of evaluating human-machine interface issues and ergonomic issues. Over a period of five months, forty- three human subjects performed 700 workout sessions on this unit, accumulating over 150,000 cycles. The unit was also used extensively by the team to evaluate various design modifications as well as to develop operational and maintenance procedures. In addition, astronauts with long-duration spaceflight experience on both Shuttle and ISS were invited to exercise on ARED to identify issues unique to zero- gravity such as the location of handrails and foot restraints. Certification of Flight and Qualification Units Having established high confidence through development testing on Rig 1 and Rig 2 and the Hi-Fidelity Engineering Unit, production of three flight-quality ARED units are underway and should be nearing completion at the time of this symposium. One unit (Dash 301) will be designated for delivery to the International Space Station, the second unit (Dash 302) will serve as the Lifecycle Qualification Unit, and the last unit (Dash 303) will be the General Qualification Unit. The ARED Qualification and Acceptance Test Plan stipulates that each of ARED’s components must be subjected to an Acceptance Vibration Test (AVT). This test is intended to be a workmanship screen. The Flight Unit, Dash 301 will then undergo a series of functional tests and packaged for launch. Dash 302, after a series of functional tests, will be installed on Rig 1 and Rig 2 and undergo continuous lifecycle testing. Current requirements state that ARED must successfully complete 1.5 M cycles on the ground in order to qualify the Flight unit for one year of on-orbit usage (6 crew members, exercising 1.5 hours per day, and 6 days per week). Results from lifecycle testing on Dash 302 will be used to determine maintenance and re-supply plan for the on-orbit Flight Unit. Dash 303, will be subjected to a launch vibration test (QVT) and an acoustic emission test. In between these tests, Dash 303 will undergo a series of functional tests. At the conclusion of the Qualification and Acceptance Test Program, Dash 303, along with the Hi-Fidelity Engineering Unit, will become crew training units and/or research test beds for ground-based physiological studies. 329

Document

AMS_2006.pdf

Conclusion The most important lesson that the team has learned in the course of the ARED project is that prototyping and development testing is paramount and indispensable for developing a robust, reliable, complex mechanical system. Good initial designs are important as well, but nothing can replace developmental testing as a means to flush out design, manufacturing, assembly, and human-machine issues. Although not popular, a prototype-centric development phase, in advance of flight production, will result in fewer design problems, fewer modifications, fewer design iterations and, in the long run, decreases cost and schedule. As a result of AREDs extensive prototyping and testing, both in the developmental phase and the certification phase, the team is extremely confident that ARED will perform well on the International Space Station and provide years of effective resistive exercise for the crew. Furthermore, as NASA prepares to implement the President’s Exploration Initiatives, ARED’s technology will be mature and proven in time to support long-duration expedition to the Moon and Mars. 330

Document

AMS_2006.pdf

Radarsat Range Adjustment Mechanism Design Xilin Zhang* and Sylvain Riendeau* Abstract RADARSAT (Figure 1) is a series of sophisticated earth observation satellites developed by Canada to monitor environmental changes and the planet’s natural resources. RADARSAT also provides useful information to both commercial and scientific users in the fields of agriculture, cartography, hydrology, forestry, oceanography, ice studies, and coastal monitoring. At the heart of each RADARSAT satellite, there is an advanced Synthetic Aperture Radar (SAR) payload. The SAR antenna is a 15-meter-long microwave instrument that sends pulsed signals to Earth and processes the received reflected pulses. Its on-ground test fixture has high-resolution alignment requirements. It is a challenge to any mechanical designer. The Radarsat-1 test fixture included an 18- meter-long near field range scanner, an offloaded 6 degrees-of-freedom fine adjustment system to handle and maneuver the 15-meter-long SAR antenna, an automatic closed-loop motor-driven adjustment system, and a 15-meter-long coarse adjustment table. A large R & D effort was required throughout the design, fabrication and final calibration of the equipment to ensure final success. Lessons learned from the RADARSAT-1 range alignment fixture paved the road for the RADARSAT-2 system. This type of 6 degrees-of-freedom adjustment mechanism could be very useful to all kinds of large space structures for on-ground alignment and test like solar arrays, antenna panels, etc. This paper covers the 6 degrees-of- freedom mechanism design, development, fabrication and off loading calibration method. The lessons learned during design, fabrication as well as integration and calibration are extremely useful to avoid future problems for all mechanism designers. _” 1 Figure 1. RADARSAT-2 Radarsat Mission Introduction boiar Array =- * MDA Space, Inc., Ste-Anne-De-Bellevue, Canada Proceedings of the 3dh Aerospace Mechanisms Symposium, Langley Research Center, May 17- 19,2006 331

Document

AMS_2006.pdf

Being the second largest country in the world, with a variety of landscapes and climatic conditions, Canada recognized the practical and economic benefits of using space for Earth observation early on. With global environmental monitoring and protection being a worldwide concern, Earth observation is a key priority of the Canadian Space Program, and RADARSAT was developed as Canada's flagship to pursue this priority. It is a Canadian-led project involving the Canadian federal government, the Canadian provinces, and the private sector. It provides useful information to both commercial and scientific users in such fields as disaster management, interferometry, agriculture, cartography, hydrology, forestry, oceanography, ice studies and coastal monitoring. Radarsat-2 capabilities are shown in Figure 2. RADARSAT-1 is a sophisticated Earth observation (EO) satellite. Launched in November 1995, it provides Canada and the world with an operational radar satellite system capable of timely delivery of large amounts of data. Equipped with a powerful Synthetic Aperture Radar (SARI instrument, it acquires images of the Earth day or night, in all weather and through cloud cover, smoke and haze. RADARSAT-1 has proven to be an invaluable source of Earth observation data. The satellite's images are used internationally to manage and monitor the Earth's resources and to monitor global climate change, as well as in many other commercial and scientific applications. RADARSAT-1 is ideally suited to support these tasks, thanks to its right and left looking modes, wide range of beams, SAR technology, frequent revisit periods, high-quality products and fast, efficient delivery. In November 2005, Radarsat 1 celebrated its 10th anniversary and is still functioning very well in orbit. Representing a significant evolution from RADARSAT-1, RADARSAT-2 will be the first commercial SAR satellite to offer multi-polarization - an important tool increasingly used to identify a wide variety of surface features and targets. As prime contractor for RADARSAT-2, MDA will develop, build, integrate and launch RADARSAT-2. Once operational, RADARSAT-2 will be wholly owned and operated by MDA. To be launched in 2006, RADARSAT-2 will be lighter, cheaper, more capable, and will ensure data continuity well into the new millennium. Its enhanced capabilities include additional beam modes, higher resolutions, multi-polarization, more frequent revisits, and an increased downlink margin enabling reception of data from lower-cost receiving antenna systems. RADARSAT-2 will carry a C-band remote sensing radar with a ground resolution ranging from a mere 3 to 100 meters. Other key features of RADARSAT-2 include the ability to select all beam modes in both left and right looking modes, high downlink power, secure data and telemetry, solid-state recorders, an on- board GPS receiver and the use of a high-precision attitude control system. Fully flexible polarization options and the ability to acquire images to the left and right of the satellite will double the accessibility swath. The three-meter resolution data generated by RADARSAT-2 will be the highest-resolution commercially available SAR data, offering enhanced detection of closely spaced objects, as well as enhanced definition of other objects. At the heart of each RADARSAT satellite, there is an advanced Synthetic Aperture Radar (SAR) payload. The SAR antenna is a 15-meter-long microwave instwment that sends pulsed signals to Earth and amplifies the received reflected pulses. Its on-ground test fixture has high-resolution alignment requirements. The design of the required Mechanical Ground Support Equipment (MGSE) is a challenge for any mechanism engineer. SAR Range Test Setup and Adjustment Mechanism Background The sophisticated SAR RF system requires a high-resolution adjustment mechanism for the range test setup. The four panels must be properly aligned to meet the stringent electrical performance requirements and to allow for the proper installation of the expendable support structure. The basic Mechanical Ground Support Equipment requirements for the range test are listed below: 332

Document

AMS_2006.pdf

Figure 2. Radarsat 2 Capabilities 1. The front radiating surface flatness of SAR 15-meter-long 1.5-meter-wide panel shall be within 0.7 mm (0.028 in) to meet the fine resolution RF requirement. 2. The front surface flatness of SAR shall be adjustable in 6 degrees-of-freedom. lt12.7 mm (0.5 in) translation in all X, Y, Z directions. 0.35" of rotation around X, Y, Z axes to optimize the RF performance. 3. After RF optimization, the SAR panel's positions are defined. The positions shall be able to be locked in position to allow ESS installation. 4. A heavy-duty working platform is required to locate SAR at the center of the scanner plan to allow a 1.5-meter over-scan all around the SAR. 5. SAR shall be able to move 700 mm in and out from the range scanner probe to enable the RF engineer to optimize the test distance. 6. Range Adjustment Mechanism shall allow full access for ESS, Spacecraft Y panel assembly and dummy sidewall installation after RF performance to be optimized. 7. Range platform shall be able to allow a minimum of 10 operators to be on the platform and to work at the same time. The range test setup design (Figure 3) includes the following major MGSE assemblies to meet the above requirements: 1. 18 m x 6 m near field scanner to perform all required RF tests. 2. A range platform with 700-mm travel in and out from the scanner probe. It enables RF test engineers to optimize the SAR panels to probe distance. 3. Four support trusses and a steel structure foundation are made to support the SAR panels' test setup at the center of the scanner. 4. Dummy spacecraft sidewalls for ESS integration as well as a Y panel support to bring all necessary payload electronics up for the range test. 333

Document

AMS_2006.pdf

5. A 6 degrees-of-freedom adjustment mechanism to allow SAR panels to be located according to their designed positions on range and to optimize the RF performance. I -1 INNER> SAR PANEL RANGE PLAT F OHM RANGE AOJ!ISTMENT SPACECRAFT Y- PANEL DUMMY SIDE WALL 1 r- Figure 3. Complete SAR Panel’s Range Test Setup This paper will concentrate on the 6 degrees-of-freedom mechanism design, development, fabrication and offloading calibration method. During the design, the fabrication, the final integration and the calibration phase, many problems occurred. The lessons learned are extremely useful to any mechanism and MGSE designer to avoid future problems. SAR Range Adjustment Mechanism Design The SAR antenna is made of four SAR panels (+X inner and outer panels and -X inner and outer panels). Four separate panels are required to stow the antenna in the launch configuration. The extendible support structure will deploy the SAR Panels and hold them in position in orbit. During the range test, the antenna was held in position in the deployed configuration by four sets of identical/mirror-imaged mechanisms. Those mechanisms are used to adjust each panel in the required position and then hold them in place. The mechanism working principles were sketched out during the preliminary design phase. Eight different design proposals were studied, each one coming from a different design approach. A trade off analysis was performed and the Preliminary Design Review approved the principle shown in Figure 4 and detailed 334

Document

AMS_2006.pdf

below. Each set of the adjustment mechanisms has an X-direction mechanism, a Y-direction mechanism and a 2-direction mechanism. The Z mechanisms are installed directly on the range platform. The X mechanisms are seated on two Thomson linear slides. A Thomson linear shaft connects each X mechanism to a 2 mechanism through two ball joints. The Y mechanism consists of two Thomson linear bearings. The Y-direction linear shaft is press fitted into the X-mechanism beam. The linear motorshicrometers are installed on the mechanisms in every direction. & LINEAR BEARING LEGEND OF FIGURE Figure 4 SAR Range Adjustment Mechanism Design Principle The panel linear translations in X, Y, 2 directions are very simple. The two linear motorshnicrometers in the same axis need to be driven in a synchronized fashion. The rotations are achieved by driving the two motors/microrneters on the same axis for a different distance or direction. For a simple example, when the linear motors/micrometers M2 and M3 are driven in a different direction, the panel will rotate around the 2 axis. If both a translation in the Y direction and a rotation around the Z axis are required, the linear actuators M2 and M3 can be driven in the same direction but for a different distance. The panel will move up or down and rotate around the 2 axis. During all the adjustment operations, the SAR panel is connected to the adjustment mechanism by three corners; the fourth corner (upper left in Figure 4) will be a safety-locking device. During the adjustment process, it has to be completely loose to give the panel the 6 degrees-of- freedom. Once the panel is in the required position, the fourth corner is locked in place with the safety-locking device. Once the design principles are clear, many fine details have to be carefully looked at. For example, the safety issues for the operators working around the Range Fixture, the flight hardware protection, etc. The 335

Document

AMS_2006.pdf

MGSE requirements 3 through 7 mentioned earlier are addressed by other generic mechanisms. Their interaction with the 6 degrees-of-freedom adjustment mechanism needs to be carefully looked at during the detailed design phase. The final design of the 6 degrees-of-freedom adjustment mechanism is shown in Figure 5. Single SAR L palell \ -' ' Figure 5. Single SAR Panel Range Adjustment Mechanism Figure 5, extracted from the CAD model, only shows the +X outer panel range adjustment mechanism. The -X inner panel mechanism is a duplicate of this design and the other two sets of mechanisms (-X outer panel and +X inner panel) are mirror images. SAR Range Adjustment Mechanism Motor-Driven Software For Radarsat-1 , the linear motors, with 8.2-kg (18-lb) maximum capacity and their control system are selected early in the program. Each SAR antenna panel adjustment mechanism is driven by 6 linear motors M1 - M6. The motor execution program controls the mechanism motors and give the 6-degrees- of-freedom adjustment to the antenna panels. 336

Document

AMS_2006.pdf

-t Motor Displacement Calculation CAD Graphic Exec File Motor Format Conversion CAD Graphic Display Adjustment Exec File Encoder Linear Actuator Figure 6. Panel Adjustment Software Block Diagram This system is used to align the panel in the plan parallel to the scanner plan or to adjust the panels to any pre-defined test position. Also, the discrete manual input to the motion control program can drive the panel during RF testing to the optimum RF locations for each panel. This special software was developed for this specific application. The panel adjustment software block diagram is shown in Figure 6 [l] and is followed by a detailed description of the main modules. Data input: Panel surface points, panel four corners points and pre-defined set up position. Optimization: Determine the best-fit antenna plane. Adjustment sequence and actuator displacement calculation: Calculate each of the required motor displacements from the adjustment sequence. CAD output: Display the adjustment sequence on CAD screen. Linear motors format conversion: Convert the required distance travel to the appropriate linear actuators format in number of encoder counts. Assign the appropriate speed to the synchronized motion. Motor/micrometer movement to achieve translation in X, Y, Z axes direction and rotation around XI Y, Z axes are listed in Table 1. For Radarsat-1, once the SAR panel is installed on the range, the first set of panel position data was taken and the data inputted in-to the program. The software calculates the movement distance of each motor required to drive the panel in the designed position. 337

Document

AMS_2006.pdf

Table 1. Six Degrees-Of-Freedom Linear MotorNicrometer Adjustment , Micromew- rranslation X I l rranslation Y lrranslation Z Rotation ex Rotation By + or - 1 F or - + or - + or - Mt5 *l. Small adjustment might be required, but it is not the same distance as the others. System Calibration and Verification Most large MGSE designers typically specify a pre-calibration activity to verify the system design and manufacturing before using it with flight hardware. Nobody wants to see expensive flight hardware hanging in the air waiting to be installed on a MGSE that doesn’t work. To avoid surprises during the flight hardware final installation phase, a mass dummy with flight mass, CG and flight representative interfaces is designed and fabricated specifically for the calibration and verification. The Radarsat-1 range fixture calibration went through many troubleshooting exercises; some of the problems will be discussed in the lessons-learned section. All encountered problems were resolved during the Radarsat-1 calibration and made the final integration a clean and problem free operation. These experiences paved the road for the Radarsat-2 range fixture design and calibration. For Radarsat-2, a new mass dummy was designed. It can be calibrated to represent all four Radarsat 2 panels. Weight location and calibration hoisting points are designed to be adjustable to meet all 4 different panels’ CG and mass. All linear motor/micrometer loadings are finally balanced to less than 5.5 kg (1 2 Ib). The Radarsat-2 mass dummy calibration is shown in Figure 7. During the calibration, each panel location was positioned relative to the others using the NFR 3-axes scanner in order to meet the design requirements. To calibrate the position of each panel, an X, Y, Z mechanical block was installed at the center of the platform as the original point. The travel of the scanner is used with a dial gage, which is installed on the near field scanner probe to precisely measure the position of each measurement point. The final calibration got the dummy panels within 0.38 mm (0.01 5 in) of their as-designed position. This simplifies the final SAR panel range set up installation and alignment processes. 338

Document

AMS_2006.pdf

1 Figure 7. Using Mass Dummy to Pre-Calibrate the Design Setup SAR Panel Final Range Alignment Result [2] The four flight SAR panels were installed on the near field range adjustment mechanism. They were positioned relative to each other and relative to the scanning plane using the NFR 3-axes. In order to facilitate the positioning of the 4 SAR panels in a later stage of integration, 16 alignment-aid interface points per SAR panel were machined on their back face. Laser tracker measurement equipment was used to determine the position of these points relative to a reference coordinate system. 339

Document

AMS_2006.pdf

* Large Scale Measurement Device > */ Pane's Base # ,/- Frame d . Base (41 Frame Laser Tracker + Laser Tracker Targets Figure 8. SAR Panel Location Measurement Setup The line of sight to these points in the test range is limited due to mechanisms and guardrails in the back of the panels. Different set-ups were required. Figure 8 shows a typical set-up schematic. The measured point positions were finally given in the mechanical build (mb) coordinate system that was predefined. Repeatability of the measurement was verified and the error from nominal position was calculated. The final Laser tracker measurement results of the three-dimensional positional errors are listed in Table 2. The error is defined as the difference between nominal and measured positions. The maximum error calculated is 0.61 mm (0.024 inch) in the X direction. The compound error is attributed to the tolerance of machining and assembly, the relative position of each panel to the reference frame and the measurement set-up accuracy. 340

Document

AMS_2006.pdf

Table 2. Panel Measurement Point Location Error (in thousandths) 12 -2 -10 13 4 -9 18 8 -8 44-8 -Xim &*a 18 5 -2 10 4 8 15 3 2 18 -9 3 22 2 17 18 -5 13 22 9 16 12 3 5 10 -1 -1 17 -5 0 213 2 3 27 -3 8 29 -14 -13 22 4 -17 +Xim -2 0 4 -2 3 4 4 4 -1 -5 12 3 -9 2 3 8 8 -1 -15 2 0 -1 1 3 -6 -13 -2 -7 -15 a0 4 -a 10 -7 -22 4 -9 -21 10 -8 -1 -16 a -24 5 -9 -17 -15 -24 -7 -17 -10 -12 -8 -23 -19 -21 -5 -17 -11 -14 -10 -10 6 3 4 -3 4 18 -1 8 0 6 -19 -12 -19 -19 -17 -14 -18 -17 -7 Lessons Learned Lessons Learned 1 Linear motors with a computer software auto-adjustment system offer a nice system to have if there are several hundred panels to be adjusted, aligned and tested. They can significantly reduce the labor costs (on mass production). For single SAR requirements, it is not cost effective. For Radarsat-1, the problems with the motors, driving system and software, as well as all related troubleshooting. It became the major cost for the overall Range Fixture. For RADARSAT-2, MDA decided to replace the system with a set of 6 (total 24) high-resolution manual micrometers, a dial gage O,O,O reading system and somebody on the floor with a good calculator. Good results were achieved at a fraction of the price. Lessons Learned 2 Offloading point selection is critical to get ideal 0 g and linear sliding bearing moment free conditions. The RADARSAT-1 range fixture offloading point was first designed to be at the center of the X-direction slides with a constant-force spring offloading system. It completely failed. The 8.2-kg (1 8-lb) loading capacity motor can not drive the panel to move up and down at all because of the internal friction of the constant force spring and the large moment generated by the panel weight. The offloading point was finally moved as close as possible to the panel CG and a balanced-weight system was used. The moment on the sliding bearing was then removed. Before the motor installation, the moving force required to drive the panels up and down were calibrated to between 5.5 kg and 7.5 kg. The analysis shown in Figure 9 gives suggestions for offloading point selection. 34 1

Document

AMS_2006.pdf

CONSTANT F1 FORCE SPRING Y DIRECTION MOTOR M2 Br M3 lBkg MAX. CAP AC I TY . L L- L2 M=Fl*Ll +F2*L2 BALANCE ,WE I GHT SAR SINGLE PANEL WEIGHT F2=250kg Figure 9. Off loading Point Selection and Incorrect Offloading Point. From above Figure 9, one can see that with balanced panel conditions: The moment M on the Y bearing shall be: C M=O M = F1 * L1 + F2 * L2 The Radarsat-1 constant force offloading line went through Y bearing L1 =o M= F2*L2 342

Document

AMS_2006.pdf

Moment M applied to the sliding bearing significantly increased the friction. It is required to balance the moment to 0 in order to make the selected 8.2-kg loading capacity motors work for this mechanism. Following the modifications, the balanced-weight loading point was moved as close as possible to the panel CG to remove the moment from the bearing. Then, the panel can be easily driven up and down within the motor’s loading capacity. Lessons Learned 3 The constant-force spring off loading system is not ideal for small-driving capability actuator systems because the internal friction inside the constant-force spring is significant. For the 250-kg offloading system used for Radarsat-1, the best results obtained during the calibration was a force between 28 kg and 30 kg required to move the panel up and down. It significantly exceeded the specified motor maximum capacity. Conclusions The 6-degrees-of-freedom Range Adjustment Mechanism principle was successfully used for Radarsat-1 and -2. The basic principle was debugged and improved from Radarsat-1 and successfully re-used for Radarsat-2. The mechanism is very cost effective and provides good adjustment and alignment results. The Radarsat-1 launched in 1995 for a 5 years design mission life has just celebrated its 10th anniversary and is still working very well in orbit. Radarsat-2 SAR panels were delivered to David Florida Laboratory in Ottawa for final integration and will be launched in 2006. This basic principle of the 6-degrees-of-freedom adjustment mechanism can be widely used for any on-ground large flight structure alignment and test. Offloading methods for this mechanism were presented along with the calibration of the system using a mass dummy to represent the flight hardware. Picture 1 shows the -X outer panel during the range test and Picture 2 shows the SAR system final range test. Acknowledgements The authors would like to thank their retired colleague Alex Csaki for his technical advice through out the project. Thanks also go to Nathalie Hadida and Catherine Yi Zhang for their sincere effort in editing this paper. References 1. Derek Louie “Panel Alignment software - SAR antenna” MDA SPACE/ SPAR 823870 (Feb.1992) 6-8. 2. Erick Charbonneau “Project 601 8N Radarsat-2 4-SAPA alignment report, Technical Report RML 009- 2005-1 55. 343

Document

AMS_2006.pdf

Picture 1. SAR -X Outer Panel Range Test Setup I. Picture 2. SAR Range Test Setup 344

Document

AMS_2006.pdf

Come-Along Tool Development for Telerobotic In-Space Servicing of the Hubble Space Telescope Jonathan Penn* Abstract The Come-Along Tool was created for the Hubble Space Telescope (HST) Robotic Servicing and De- Orbit Mission (HRSDM). During the mission, an unmanned spacecraft rendezvous with the HST and deploys a robotic arm, which manipulates the Come-Along and other robotic tools to perform servicing operations originally intended for astronauts. Developed at Goddard Space Flight Center (GSFC), the Come-Along Tool overcomes a difficult set of obstacles to open and close the HST Aft Shroud Doors by accomplishing tasks in separate robotically feasible steps. The final design incorporates lessons in functionality, operator visibility, mechanism reliability, and design for ease of robotic operation pertinent to telerobotic tool development for future missions. Introduction Beginning in 1993, NASA has conducted a series of four highly successful astronaut missions to service and repair the HST, however, after the loss of the Columbia Shuttle in 2003, the next shuttle mission to deliver critical repairs to the HST was deemed unsafe for astronauts and canceled. It was at this point that the HST Program Office at GSFC, commenced developing the HRSDM, creating a unique opportunity to develop telerobotic technology. In order to replace the HST’s ailing gyroscopes and batteries and add more powerful science instruments, robots would be used to eliminate the risk to human life. The HRSDM makes use of a Dexterous Robot to manipulate various robotic tools to make repairs and improvements to the HST. Operation is termed ‘telel-robotic because ground operators command the robot using telemetry. I I - Hubble Telescope Space J Dexterous Robot - ‘b Spacecraft Figure 1. Hobot Servicing Hubble Space Telescope * Swales Aerospace, Beltsville, MD Proceedings of the 3dh Aerospace Mechanisms Symposium, Langley Research Center, May 17- 19,2006 345

Document

AMS_2006.pdf

One of many robotic tools created for the HRSDM, the Come-Along Tool is used during the Cosmic Origins Spectrograph (COS) Task. The Come-Along Tool opens the HST Aft Shroud Doors so that the Corrective Optics Space Telescope Axial Replacement (COSTAR) can be removed and replaced by COS, improving Hubble’s sensitivity to ultraviolet wavelengths by a factor of ten. Operation of the Aft Shroud Doors is a challenging undertaking that pushes the limits of what can be accomplished by a telerobotic system. Intended for use during crewmember extra-vehicular activity (EVA), these doors have proven troublesome on previous missions. During the first mission to service the HST, astronauts spent over two hours attempting to close the doors and it was concluded that the door operation is a two-person task. Even with two robotic arms, it would not be easy to mimic the dexterity of two astronauts performing EVA, and for this task, only one robotic arm is available. The Aft Shroud Doors are difficult to operate because they have shear plates at their top and bottom that tend to jam during closing even when properly aligned. Other tasks involved with operating the doors, while easy for an astronaut, are difficult for a robotic system. Acquiring an unrestrained door was at first difficult to perform during ground testing of the robotic system because efforts to hook the handle of the door pushed the door further away. Pushing both Aft Shroud Doors closed at the same time is a two- handed astronaut operation and challenges arose designing a mechanism that could perform this using a single robotic arm. Hubble Space Telescope 1 Left Aft Shroud Door Right Aft Shroud Door - Door Handrails Figure 2. Aft Shroud Doors Through creative problem solving and ground testing, the Come-Along Tool design evolved to overcome the difficulties associated with operating the Aft Shroud Doors in a manner that optimizes reliability, visual indication, and ease of robotic operation. The Come Along Tool is able to perform its role because it has features and an operation sequence that require the robot to move in only one degree-of-freedom at a time. Reliability is attained through use of simple, robust mechanisms and redundancy that allows the tool to perform its task even if a single failure occurs. The addition of features within the field of view of the cameras mounted on the robot allow the ground operator to watch gears rotating, discern whether the door handles are under control, and measure the tension force applied to close the doors. The ability of the tool to perform its function is further improved by combining the functionality of mechanisms to reduce the number of robotic actions and ooeration time. The details are further described in this DaDer. I Figure 3. Come-Along Tool 346

Document

AMS_2006.pdf

Background Robotic Svstem The robotic system for the HRSDM is composed of a spacecraft, a robot, and various robotic tools. The spacecraft is unmanned and intercepts the HST using a grapple arm. Once the HST is captured, the Dexterous Robot, attached to the spacecraft via a boom, is deployed to begin servicing and upgrading the telescope. The Dexterous Robot, built by MacDonald Dettwiler Robotics for use on the International Space Station, has two identical arms, one of which serves as backup in case the first fails. At the end of each arm is an Orbital Tool Change-out Mechanism (OTCM) with jaws to clamp a “Micro Fixture” interface and a socket that advances onto and rotates a 7/1&inch hex head. The Dexterous Robot was designed to move large payloads on the International Space Station, not to perform the precision operations necessary to service the HST. Bridging the gap between the Dexterous Robot and the HST, Swales Aerospace developed over twenty robotic tools for the HRSDM, each customized to specific tasks necessary to service and upgrade the HST such as manipulating connectors, attaching to science instruments, and opening bay doors. Micro Fixture Figure 4. Dexterous Robot Orbital Tool Change-out Mechanism and Micro Fixture Worksite The Aft Shroud Doors on the ‘-V2’ face of the HST must be opened to access COSTAR. Restrained by four latches, the Aft Shroud Door set has gaskets along its perimeter and shear plates where the doors interface each other and the HST structure. The gaskets protect the optical equipment inside the telescope from light and contamination and become stiffer at cold temperatures. Providing attachment Doints for the Come-Alona Tool are EVA handrails on each door. Hut I loor i’ A Shear Plate Locations Marked ‘X‘ Right Door Right Door Handrail * Tool Attachment Points Figure 5. Worksite 347

Document

AMS_2006.pdf

Door operation is complicated by sets of mating shear plates at the top and bottom of each door and at the interface between the doors. These shear plates transfer structural loads from the Aft Shroud Frame and require the right door to close before the left to mesh properly. Operation of the '-V2' door set during the first mission to service the HST revealed that the shear plate on the bottom left door tends to jam even when the door is properly aligned. Temperature gradients distort the shape of the telescope's outer frame potentially exacerbating this problem. Svstem of Tools to Operate the Aft Shroud Doors In order to perform the COS Task, the Aft Shroud Doors are operated by three robotic tools: the 90 Degree Door Latch Tool, the Door Restraint Tool, and the Come-Along Tool. The Door Latch Tool is used to release the four door latches during opening, and is used to reengage the latches once the doors are closed. Helping to control the doors once released, the Door Restraint Tool is used to hold the doors open during operations inside of the Aft Shroud, and is reconfigured to hold the doors nearly closed. The Come- Along Tool, described in this paper, controls the opening of the Aft Shroud Doors and closes and restrains the doors in the shut position. Door Latch Tool Door Restraint Tool Figure 6. The Door Latch Tool and the Door Restraint Tool Come-Alona Tool Requirements The Come-Along Tool must satisfy mission level and task specific requirements. During door openina, the Come-Along must attach to the leh and right door handrails,' hold the doors shut while &e DooiLatcLTool is used to release the latches, and slowly release the doors such that the pressure from the gaskets does not throw the doors open. The Come-Along Tool must then release from the right door handrail so that the doors can be fully opened. When it is time to close the doors, the Come-Along Tool must reacquire the handrail of the right door which is now free to move. The Come-Along Tool, hooked on to the left and right door handrails, closes the doors by applying a tension force between the two door handrails causing the doors to rotate along their respective pivot points. The tool must then hold the doors in position until the door latches can be reengaged. Tension to Close Doors -- -* \& .-.4 Come-Along Tool Left Aft Shroud Door A/ Door Pivot Point Right Aft Shroud Door Door Pivot Point -e \- Figure 7. Bottom View of Operation to Close Doors The Come-Along Tool must also satisfy requirements for design for minimum risk and single-fault tolerance, meet weight restrictions, and provide the ability to overcome jams at the door shear plates. The tool must be able to perform all of its functions using only one half of the robot's force and moment 348

Document

AMS_2006.pdf

capability in an environment that cools to below negative 80°C. Satisfying these requirements with a robotic tool that is easy to use and that provides good visual cues provided a great engineering challenge. Concept Development The Come-Along Tool was developed from May of 2004 through July of 2005, culminating in an Engineering Test Unit. Tests were performed at GSFC using a door trainer constructed to flight drawings and a ground trainer Dexterous Robot. With each successive generation, features were added to incorporate lessons learned from previous testing. Door Trainer Right Door Left Door Ground Trainer Dexterous Robot Left Tool Section Left Door, Handrail Right Tool ’ Section Right Door Handrail Generation 1 In order to begin to understand door operation using the ground trainer Dexterous Robot, a mockup of the Come-Along Tool uses hooks and a modified tape measure reel. This mockup demonstrates the ability to attach to the Left and Right Door Handrails. Generation 2 This generation is more robust, containing c-clamp type jaws that actively clamp the Left and Right Handrails. The ground trainer Robot closes the doors by winding a cable reel that is coupled to a one-way ratchet, holding the doors shut. Figure 9. Generations 1 I & 2 Come-Along Tool Prototypes 349

Document

AMS_2006.pdf

Generation 3 With the addition of a DC motor, the third generation prototype can disengage the ratchet system and control door opening. The tool can capture the handrails when the doors are latched closed, but not in the more difficult situation in which the doors are unrestrained. Generation 4 In this generation, the DC motor is removed to simplify the tool. The handrail clamps are lengthened to spread the load of pitch moments induced by the ground trainer Robot to overcome door jams. Finger features are added to the clamps allowing the tool to capture the handrails of unrestrained doors. Generation 3 Generation 4 Figure 10. Generations 3 & 4 Come-Along Tool Prototypes Generation 5 The Engineering Test Unit incorporates many new design features and successfully performs all reiuired door operations. The cable system is removed, alleviating complications associated with tethers , and replaced with a threaded Lead Screw. This makes the tool more reliable by reducing the number of parts and providing a rigid link between tool sections. Operations of clamping the handrails, opening and closing the doors, and holding the doors in the closed position are combined into one to improve ease of robot operation. The robot grips the tool via Micro Fixture robotic interfaces. The most significant addition is the spring-loaded Lead Screw Strut, which provides visual indication of the tension force between the two doors. The added benefit of the spring-loaded Lead Screw Strut is that it continues applying a closing tension to the slightly ajar doors. This allows the robot to release the tool and perform contingency operations in case either the left or right Aft Shroud Door is jammed. Micro Fixtures Come Along Tool \ Lead Screw- Dexterous Robot Ground Trainer Figure 11. Generation 5 Come-Along Tool Engineering Test Unit 350

Document

AMS_2006.pdf

Design for Functionality The ability to perform certain tasks associated with operating the Aft Shroud Doors was achieved by separating operations into discrete steps that the ground trainer robot could accomplish by moving in only one degree of freedom at a time. During early phases of development of the Come-Along Tool it was difficult or impossible to acquire the handrail of an unrestrained door or overcome a door jam because complex robot movements were required that involved various combinations of translations and rotations. This approach to telerobotic operation is not effective because it assumes that a robot is able to coordinate movements with a level of dexterity approaching that of an astronaut. By breaking down movements into small simple steps, functions were made feasible. Acquire Unrestrained Door Handle The task of acquiring the handrail of an unrestrained door was at first difficult and later made easy by separating operational steps. During door closure, when the doors start open and are free to move, attempting to hook a tool section onto the handrail of a door is difficult because any inadvertent bumping of the handrail sends the door off into another position. The robot then has to reposition the tool for another attempt at handrail capture. This operation is made possible by adding a ‘finger’ feature to the tool and first pushing the door against the door seal to lock out its movement before sliding the clamp over the handrail. This new operation requires two steps that are pure translations and can be performed easily by the robot operator. Implementing this same strategy also was successful for more difficult tasks. Handrail Fingers ‘f w \ ’1 Right Door Right Door Handrail - Ground Trainer Dexterous Robot \ Right Tool Section I Figure 12. Come-. ..-ng Acquiring Door Handrail Overcome Door Jams The ability to overcome door jams was achieved by separating actions into simple steps. As mentioned before, it took two astronauts working in tandem more than two hours to close the same door set during the first mission to service the Hubble Space Telescope. Door jams were induced in the ground trainer door set testing and early attempts to lift, pull, and twist the door into place were unsuccessful. Ultimately, it was discovered that applying a pitch moment to a door overcomes jams. To allow greater moment input without damaging the door handrails, the tool’s handrail clamps are extended to spread the loading. Operation to overcome a door jam is made possible by completing the following steps: (1) pushing the door closed until it jams, (2) applying a pitch moment in the up or down direction to relieve the jam, and (3) completing door closure. 35 1

Document

AMS_2006.pdf

Pitch Moment applied by Dexterous Robot. 13.6 N-m f 10 ft-lb) Come-Along Left Tool Section 4 Left Handrail \ Figure 13. Pitch Moment Application Design for Visual Indication Ability to view the worksite is essential to completing sequential operational steps. Mounted to the robot, cameras are intended to alert the operator if a dangerous situation arises and provide views so that the robot operator has confidence that the tool is performing as intended. However, these cameras are only effective if the tool design makes good use of them. The addition of features within the field of view (FOV) of the cameras mounted on the robot allows the operator to watch gears rotating, discern whether the door handles are under control, and measure the tension force applied to close the doors. Primary OTCM meras Come-Along Tool + Cai -. __ \\ OTCM / Handrail Finger Features h + Right Tool 1 Section . ConicalShapes / Represent FOV from OTCM cameras Figure 14. Field of View r Handrail Capture To hook the Come-Along Tool onto a door handrail, the robot operator needs to know the tool’s distance from the handrail while making an approach. The gap between the tool and the handrail is difficult to judge when the finger feature, shown below, is in-line with the camera. To gain an offset view, the finger features are placed on the edges of each tool section at the periphery of the OTCM camera’s field of view. This lets the operator discern the changing gap between the fingers and the handrail during handrail capture. 352

Document

AMS_2006.pdf

Finger Feature / Finger 4 Fentiire Come- Alnnri Tnnl - Handrail - / Handra I I . Come- Alnnn Tnnl In-Line View: Gap Not Visible Offset View: Gap Visible Figure 15. Indication of Distance from Handrail Tiahtenina and Loosenina Doors Visual indication of whether mechanisms in the tool are operating is necessary during door opening ana closing. During operation, the robot operator sends a command to input torque to the tool's gear train, causing the tool to move along the lead screw which tightens or loosens the doors. Since the movement of the doors is slow, it is difficult to ascertain whether the command had the intended effect or when the operation is complete. Cutaways were added to the tool's housing so that views of rotating gears can be seen. By watching gear rotation, the operator can tell when the doors are moving and when they have stopped. I Miter Gears Cut Away Section Cut Away Section Figure 16. View of Rotating Gears Measurement of Tension Force Between Doors Applying tension between the doors is a sensitive operation; if the operator applies too small a force the doors will not close and if he applies too much he could damage the tool or the handrails. In order to provide the operator with force information, a spring-loaded gauge is added in line with the lead screw. The gap between the gauge housing and indication mark varies with the tension between the two Come- Along Tool Sections. 353

Document

AMS_2006.pdf

Gauge Housing Threaded Lead Compression Spring Screw \ \ Distance varies with tension L Indication Mark Figure 17. Tension Indicator Design for Ease of Operation The Come-Along Tool is made easy to operate by minimizing the number of times the Robot must disengage and reattach since each reattachment has an associated risk of error that could cause damage to the tool, robot, or telescope. This is accomplished by allowing the robot to control all aspects of door operation using a single mechanism and by including a backup of this same mechanism. Combine Operations The Come-Along Tool is optimized for ease of operation by allowing the robot operator to control the following three functions from a single attachment point: 1. Apply tension to close the doors 2. Release tension to open the doors 3. Hold the doors in place This is made possible by the design of the gear train. When attached to a tool section the robot powers the tool’s gear train using its ‘advance’ mechanism. A 7/1&inch hex socket extends from the robot engaging a mating male hex in the tool. Rotational motion from the robot travels through the tool’s gear train through two right angle bevel gear connections. Clockwise input rotates a nut about the lead screw causing the tool section to translate along the lead screw and apply a closing tension on the doors; a counter clockwise input causes the tool section to travel in the opposite direction along the lead screw allowing the doors to open. The non-back driving nature of the nut on lead screw connection is sufficient to lock the doors in place. This arrangement is easy to operate since the robot operator does not need to let go of or reattach to the tool during nominal operation. HexHead (Robotic Interface) Bevel Gear \ Stationary Miter Gear Bevel Pinion L a Pivoting Miter Drive Nut -kd Screw Figure 18. Gear Train Perform Operations from Either Tool Section Adding the same gear train mechanism on both tool sections further optimizes the Come-Along Tool for ease of operation. It was found during ground testing that the Robot must attach to a door to overcome that door’s jam. In the case in which the tool lets go of one tool section and moves to the other to relieve a 354

Document

AMS_2006.pdf

jam, it does not need to move back; it can continue adding tension to the lead screw and hold the doors in the closed position using the interface on its new attachment point. Right Come-Along Left Come-Alon Tool Section Figure 19. Redundant Tool Sections Design for Reliability Reliable functioning of the Come-Along Tool is necessary to ensure that door operation is completed; if the doors are not properly closed, the HST cannot continue its science mission. Since no human will be available to perform unanticipated workarounds it is especially important the robotic tool be dependable. The robustness of the Come-Along Tool’s design is due to its simple mechanisms, hard stops to prevent position overrun, and ability to continue its task if a mechanisms failure occurs. Mechanical Hard Stops The Come-Along Tool has mechanical hard stops to ensure that the tool sections do not rotate too far or run off the end of the Lead Screw. To prevent the Lead Screw from pitching out of position, the tool housing provides a stop. Collars on the edges of the Lead Screw prevent either tool section from running off. Lead Screw Hard Stop 3- Collar Stop on Tool Housing to prevent Lead Screw Pitching - Night Tool Section Figure 20. Hard Stops Simde Mechanisms In order to reduce complexity and thereby increase reliability, the Come-Along Tool incorporates simple mechanisms to hold the doors in place, provide visual indication of tension between the doors, and allow each tool section to rotate about the Lead Screw. During earlier generations of the Come-Along, a ratchet mechanism, used to lock the doors in the closed position, required a ratchet gear, pawl, pawl spring, and motor actuated pawl release. To simplify the tool, the ratchet mechanism is replaced with a nut to lead screw connection that does not allow the gear train to back drive. Avoiding use of electronics eases temperature requirements on the tool and eliminates the need for batteries. One of the challenges of operating the Aft Shroud Doors is that the angles of the door handrails vary with the doors rotation. To remain attached to the handrails as the doors open and close, each of the tool 355

Document

AMS_2006.pdf

sections must rotate with respect to the Lead Screw. To accomplish this in a simple manner, a property of miter gear connections is used: as long as the apex of the two mating miter gears pitch cones remain aligned, one miter gear can rotate about the other. The miter gear connection allows power transfer through a rotating joint. Pivot Housing. Main Housing a Right Tool Section (Nominal Position) Stationary Miter Gear Right Tool Section Pivot Point (Rotated Position) Figure 21. Rotation Relative to the Lead Screw Sinale Fault Tolerance One of the mission level requirements of the Come-Along Tool is that it be single fault tolerant, meaning that it shall continue to operate even if any single mechanical failure occurs. The tool accomplishes this by including a redundant tool section. Should a part on one tool section fail, the other tool section can be used to continue operating the Aft Shroud Doors. Conclusion Reliable operation of the Aft Shroud Doors with the Come-Along Tool is critical to performing the COS Task and achieving the scientific objectives of the HRSDM. The tool’s current design overcomes a difficult set of obstacles to make door operation possible. Through testing and development of the tool, we gained insight regarding Aft Shroud Door operation, reliability issues involved with Robotic Tool design, and the use of telerobotics. Robotic missions require a high level of reliability since there are no humans to perform unanticipated ‘work arounds’. After first struggling to operate more complicated prototypes, the Come-Along Tool design was optimized through use of simple and redundant mechanisms. Functions for closing the doors, clamping the handrails, and holding the doors in place were combined into one. If any part of one tool section fails, redundant features on the other tool section can be used to complete operation. Important lessons were also learned about telerobotic operation. Using the force-feedback capability of the robot, it was discovered that certain human operations could be performed robotically. The simple task of acquiring the handrail of an unrestrained door was at first difficult to perform with the robot, and later made easier by adding additional features, such as Handrail Fingers to the tool and adjusting the door capture operation to first press the door shut. The execution of operations to overcome door jams successfully combines the capabilities of a robotic tool and robot to perform a difficult task that requires two astronauts working in tandem. The results of developing the Come-Along Tool are applicable not only to the HRSDM, but also to future robotic missions as we begin to realize NASA’s vision for space exploration. Development of the Come- Along Tool furthers NASA’s goal to “. . .implement a safe, sustained, and affordable robotic and human program to explore and extend human presence across the solar system....”* The Come-Along Tool’s 356

Document

AMS_2006.pdf

ability to operate interfaces made for humans and coordinate with other robotic systems is pertinent to the next generation of space exploration in which robotics will play a key role. References 1. Tomlin, Faile, Hayashida, Frost, Wagner, Mitchell, Vaughn, and Galuska. “Space Tethers: Design Criteria.” NASA Technical Memorandum 708537 (July 1997). 2. Martin, Gary L. “Level 0 Exploration Requirements for the National Aeronautics and Space Administration.” NASA Document No.: SA-0001 (4 May 2004), p. 5. Acknowledgements Development of the Come-Along Tool would not have been possible without the thoughtful work of Dick McBirney, Jenny Xu, Carl Anders, Randal Frey, Kevin McMennamin, Justin Cassidy, Paul Nikulla, John Bishop, and Giles Robinson. Thank you for your help and insight. 357

Document

AMS_2006.pdf

Planetary Airplane Extraction System Development and Subscale Testing John E. Teter Jr.' Abstract The Aerial Regional-scale Environmental Survey (ARES) project will employ an airplane as the science platform from which to collect science data in the previously inaccessible, thin atmosphere of Mars. In order for the airplane to arrive safely in the Martian atmosphere, a number of sequences must occur. A critical element in the entry sequence at Mars is an extraction maneuver to separate the airplane quickly (in less than a second) from its protective backshell to reduce the possibility of re-contact, potentially leading to mission failure. This paper describes the development, testing, and lessons learned from building a 1/3 scale model of this airplane extraction system. This design, based on the successful Mars Exploration Rover (MER) extraction mechanism, employs a series of trucks rolling along tracks located on the surface of the central parachute can. Numerous tests using high speed video were conducted at the Langley Research Center to validate this concept. One area of concern was that that although the airplane released cleanly, a pitching moment could be introduced. While targeted for a Mars mission, this concept will enable environmental surveys by aircraft in other planetary bodies with a sensible atmosphere such as Venus or Saturn's moon, Titan. Introduction The ARES project will employ an airplane as the science platform to closely survey the surface, identify the constituents of the atmosphere, and assess the residual magnetism of Mars. In order for the airplane to arrive safely in the Martian atmosphere, a number of sequences must occur, starting with Earth launch and ending with deployment. The airplane will be launched from Earth inside a protective aeroshell attached to a spacecraft. It will cruise for almost a year from Earth to Mars. Then, arriving at Mars it will begin the Entry, Descent, and Deployment (EDD) sequence. Figure 1 shows the stages of EDD. Many sequences must occur quickly to allow the plane to fly in the Martian atmosphere. Upon arrival at Mars, the protective forward aeroshell will separate from the spacecraft and coast into the atmosphere of Mars. After atmospheric drag has slowed the assembly to approximately Mach 2, a supersonic parachute will deploy to slow the craft further allowing the forward heatshield to separate. At this point, the airplane will be safely tucked in the swinging and turning backshell, which is suspended from the parachute. Now the final deployment sequence of the airplane begins. The backshell must ascend 0.7 meter relative to the airplane extraction system to expose the airplane. The extraction system will then release the airplane. In less than two minutes, the airplane will fall under the restraint of a drogue chute, unfold, pull up, and fly above the surface of Mars. This paper concentrates on the mechanicaf extraction system developed to separate the airplane from its protective backshell. Since the folded tail of the airplane is not sufficient to support the airplane launch loads, a secondary structure is required to extend past the folded wings and tail to attach to the main body of the airplane. This multi-legged, tripod structure which connects the airplane to the backshell has been dubbed the Airplane Extraction System (AES). NASA Langley Research Center, Hampton, VA Proceedings of the 3@ Aerospace Mechanisms Symposium, Langley Research Center, May 17- 19,2006 359

Document

AMS_2006.pdf

e, I I I * rr\ 't ''J I Figure 1. ARES Entry, Descent, and Deployment sequence Concept Description The primary functions of the AES are to support the airplane through launch, interplanetary cruise, and entry; and then to guide the airplane safely out of the backshell during extraction. During launch, cruise, and entry, the airplane is held by three kinematic mechanisms to prevent stresses from building up in the airplane structure by allowing the aeroshell and airplane to deform independently. During the extraction phase, six pyrotechnic separation nuts will fire releasing the AES and airplane assembly. The backshell will be free to roll up the AES guided by rollers on the AES's central ring and tracks on the backshell's parachute can. The forces of differential aerodynamic drag between the backshell's high drag supersonic parachute and the low drag free falling AESIairplane assembly will cause the separation. As the parachute can reaches the end of the AES, a second set of pyrotechnic separation nuts will release the airplane from the AES. The folded airplane is then in free fall in the atmosphere until the drogue chute is deployed (Figure 2). The following tests verify the extraction function of the airplane extraction system. A w Figure 2. Extraction concept. Left illustration shows stowed airplane stowed inside the backshell just after heatshield release. Right illustration shows the extracted airplane prior to release from the AES. AES Requirements Key AES Requirements: Low mass Hold 175-kg airplane securely through Earth launch, interplanetary cruise, and Mars entry Guide the airplane out of the backshell and release it in the Martian atmosphere Reduce stresses on the airplane due to thermal expansion and contraction Fit within the volume of a bi-conic, 2.65-meter-diameter aeroshell Airplane/AES minimum natural frequencies, 15-Hz lateral, 35-Hz axial Withstand 15-9 launch loads 360

Document

AMS_2006.pdf

AES DescriDtion The AES shown in Figure 3 is approximately 2.4-meters wide, 0.9-meter tall, with a mass of 56 kg. It is composed primarily of titanium tubes. Hard stops on the top of the central ring prevent the AES from coming off the parachute can. Figure 3. Airplane Extraction System AES Roller Confiauration Central to the success of the AES is the roller configuration. The configuration is based on the successful MER extraction hardware modified to work with anairplane. Like MER, three tracks are equally spaced on the central parachute can. However, the vertical spacing of the rollers along the track is much greater. The nominal clearance between each pair of rollers and the track is +/-0.25 mm. The clearance in the system ensures there is no binding as the backshell is pulled away. Yet, the tolerances are close enough to guide the backshell without damaging the plane. The clearances are needed to compensate for machining tolerance stack-ups and thermal growth. Theoretically, this system will still work even if the rollers do not turn, although sliding friction would result. Figure 4 through Figure 8 show the various movements allowed. In each figure, the illustration on the left shows a schematic top view. The red ring and rollers represent the AES. The blue cylinder with three protuberances represent the parachute can and tracks. The right illustration shows a schematic side view for each figure. The four red circles represent one vertical set of AES rollers. The blue rectangle represents one parachute can track. Table 1 summarizes the movements. - Figure 4. AES roller configuration, nominal clearance 36 1

Document

AMS_2006.pdf

I Figure 5. AES roller configuration, axial rotation Figure 6. AES roller configuration, radial thrust I "$ r I - Figure 7. AES roller configuration, X tilt 362

Document

AMS_2006.pdf

1 E Figure 8. AES roller configuration, Y tilt Table 1. AES Clearances and Maximum Movements Nominal Radial Thrust Rotate X Tilt Y Tilt f 0.25 mm clearance between rollers and track * 0.28 mm side to side thrust * 0.1 deg axial rotation * 0.1 deg tilt from vertical axis * 0.09 deg tilt from vertical axis Test Description In order to demonstrate the extraction design approach and operation of the AES, a functional 1/3 scale model of the backshell, AES, and airplane was created. To simulate the potential orientations in which separation would occur, the model was statically held at various angles and rotations on an A-frame in the high bay of building 1250 at NASA Langley Research Center. Figure 9 shows the test apparatus. Earth gravity was used to simulate the differential drag between the backshell and AES/airplane assembly. Although in actuality the backshell, AES, and airplane are in freefall together, practical considerations for testing dictated that the backshell be held statically for this set of tests. The relative motion is still the same and most of the dynamics are captured. A scale of 1/3 was chosen for ease of manufacturing and handling while testing. High-speed video was used to determine proper extraction. Figure 10 shows a multi-exposure sequence of a typical test. Turntables AES Backshell Figure 9. AES test apparatus 363

Document

AMS_2006.pdf

Figure 10. Multiple exposure picture of a typical extraction test Kev Test Obiectives Key test objectives were: Determine electrical cable clearances. Demonstrate proof-of-concept for an airplane extraction system from a backshell under various axial and lateral loading conditions. Demonstrate no binding as the airplane extraction system/airplane assembly rolls down the parachute can. Demonstrate dynamic clearance between the backshell, the airplane extraction system and the airplane. Determine the timing sequence for airplane release. Determine effects of the kinematic mounts on release of the airplane. Determine airplane attitude after release. Test Hardware The 113 scale model of the backshell, AES, and airplane were not miniature replicas of the full scale concept. Because of cost, schedule, and practical considerations, some compromises were made. Figure 11 shows a CAD model of the test backshell, AES, and airplane. The following sections give a brief description of the major components. Figure 11. 113 scale model of the backshell, AES, and airplane 364

Document

AMS_2006.pdf

Backs hell The scale-model backshell is an aluminum skeleton structure that represents the interior volume of the full-scale backshell. Its mass properties are not represented because it is a form only, static structure. The backshell skin has been eliminated in order to have a clear view of the extraction sequence. - AES The scale-model AES is made of welded aluminum for ease of construction. It has approximately equivalent mass, cg, inertia and leg stiffness as the full scale titanium structure. The central cylinder is missing some stiffening rings for ease of manufacture, but they were needed only for high launch loads (1 59) and not lightly loaded (29) extraction loads. Airplane The scale-model plane is a foam, fiberglass, and wood structure. The airplane represents the correct mass properties and roughly the correct volume. However, the stiffness of the airplane has not been matched. The mass, cg, and inertias are scaled from the full-scale airplane. The plan form of the airplane is correct as well as the positioning of the tail booms. Kinematic Mounts The kinematic mounts duplicate the correct function but are greater mass because miniature spherical bearings were not readily available. The fixed point, hinge point, and swivel points of attachment between the AES and airplane functionally match the full-scale model. Release Mechanism The full-scale AES and airplane are released with pyrotechnic separation nuts (Figure 12). To reduce cost and safety concerns, electromagnets were used on the scale model. While electromagnets do not release as cleanly (residual magnetism, longer response time) as pyrotechnics, these devices allow for multiple tests without replacing hardware. The electromagnets also increase the mass of the AES. I. Figure 12. 1/3 scale model release mechanism Parachute Can The parachute can is made of thick aluminum for ease of manufacturing as opposed to the thin titanium on the full-scale hardware. Also, a steel track instead of an aluminum track was used because of its durability. Rollers The rollers for the scale model are mounted on three independent rings for ease of construction and the ability to move their locations easily (Figure 13). The full-scale AES has three axial trucks instead of three rings. The important parameter is the location of the rollers in relation to the tracks and not the structure that holds them. 365

Document

AMS_2006.pdf

Figure 13. 1/3 scale model rollers Fraction 1 /3 Scaling The apparatus is dynamically scaled to 113 with the proper mass, center of gravity, and inertia properties. The scaling factors are summarized in Table 2. Decimal .333 Table2. Sc Length (1 /3)"3 = 1 /27 (1 /3)/\0.5 (1 /3)"3/( 1 /3)"3 (1 /3)/( 1 /3/\0.5) .037 577 1 .192 m/sec Mass Time I I Force kg*m/sec' kg sec Pressure Rotation deg/sec Inertia (1 /3)"3*( 1 /3)"2 = 1/27*1/9 = 1/243 ng Factors 1/3 Scaling Factor I 1/3 Scaling Factor .0041 (1 /3)/( 1 /3/\0.5)"2 = 1 (1 /3)"3*1 = 1 /27 (1 /27)/( 1 /3)"2 = 1 /3 .333 I 1/(1/9.5) 1.732 I Mass Properties The mass properties of the apparatus are summarized in Table 3. The full scale information was extracted from a ProEngineer CAD model of the ARES concept. The actual scale model information was extracted from an as built ProEngineer CAD model with selected information verified by measurement. The actual hardware corresponds well to the calculated properties. The mass of the airplane increased from configuration 1 to 2 to better reflect the calculated mass. The mass of the AES was greater because the electromagnets holding the airplane are heavier than an equivalent pyrotechnic device would be. The full-scale airplane has a slight x cg offset, however, the x cg of airplane2 is essentially zero for ease of 366

Document

AMS_2006.pdf

manufacturing but the error is negligible. The airplane inertias about the center of gravity are higher than prescribed but still within reason. Figure 14. Airplane coordinate system Table 3. Mass Properties airplane coordinate 367

Document

AMS_2006.pdf

Results There were no major surprises in testing. The airplane released cleanly in all cases. The kinematic mounts did not interfere with the release of the airplane. Yet several improvements can be made. Figure 15 shows pictures of the first four tests. Test 1 Test 2 Test 3 I Test 4 U Figure 15. Setup and release of test 1 - 4

Document

AMS_2006.pdf

Table 4 shows the test matrix. The azimuth refers to the position of the airplane about the release axis. Zenith refers to the position of the apparatus in reference to the vertical. Zero degree is vertical. The roller configuration column refers to the placement of the rings of rollers vertically along the central canister. The first group of tests positioned rollers in all three possible location, top, middle, and bottom. The second group of tests positioned roller sets only on the top and bottom ring. The release trigger point refers to the point at which the airplane is released relative to the top of the parachute c2n. The airplane was tested in two configurations. The first one had a smaller mass than the second. The video number refers to the video file name. The frame rate was reduced from 500 to 250 frames/sec for some of the high zenith angle tests in order for the video data for each test to fit on one compact disc. Tests 1 - 39 were conducted with the video camera isometric to the test apparatus to capture movement in all three axes. For tests 40 - 47 several improvements were made. The video camera was moved perpendicular to the motion of the airplane and targets were added. These improvements allowed specific points on the airplane to be tracked without compensating for the angle of the video camera. LED indicator lights showing power to the AES and airplane electromagnets were placed in the camera’s field of view. This gave precise information as to when the AES and airplane were released. Table 4. Airplane Extraction System Test Matrix 369

Document

AMS_2006.pdf

Obiectives Met Table 5. Objectives Met assembly rolls down the parachute can I with damaged rollers - of the airplane Determine electrical cable clearances Determine airplane attitude after release Problems Discovered One point of interest was that the bottom rollers were damaged after many tests because the airplane negatively Yes No cable hang-ups Yes Determined from video extraction system would rebound after airplane release. Thisis not a problem for the actual flight since the AES must only work once and is then discarded, but it may be a problem if ground testing is required on flight hardware. A wedge or positive stop at the end of travel is being considered to eliminate this motion. High-speed video revealed a pitching motion in the airplane after it was released for some extreme orientations. This is a problem in two ways. First, the pitch may cause the airplane to hit the backshell or AES under certain circumstances, although it was not observed in this set of tests. Second, the airplane now is starting to tumble. This motion must be counteracted by the drogue chute to avoid problems while unfolding the tail and wings. The ideal case would be to have the airplane to separate without a pitching moment. There are several possible causes for this pitching. Since the AES and airplane are not symmetric about the vertical parachute can axis, there is a cg offset. This offset can cause the airplane to pitch after separation. Second, after the next to the last set of rollers leaves the track, the AES is free to pitch slightly under the influence of gravity and aeroloads. Third, the electromagnets used to release the airplane and the airplane extraction system contains residual magnetism after they are turned off. Sometimes this causes the aft end of the airplane to release after the wing points have separated. This influence appears small, and will be eliminated with pyrotechnics for the flight hardware. Design modifications are being considered to address the other issues. 370

Document

AMS_2006.pdf

Future Work The next series of subscale tests should have the entire bzckshell, AES, and airplane assembly free fall in order to capture the effects of dynamics as the backshell rotates and swings during entry. This complex interaction should reveal new insights. A full-scale high-altitude balloon drop test is scheduled for 2006. This test will include a form, fit and function backshell, AES, and airplane called the High Altitude Drop Demonstrator 2. The goal of the test is to verify all aspects of the EDD at simulated Mars conditions from 30,000 meters in Earth’s atmosphere. I+ Conclusions Extraction is a critical event in the entry, descent, and deployment sequence for the Mars airplane. This development and subscale testing proves the viability of the concept. Subscale testing demonstrated a clean release of the airplane in every instance. Yet testing also showed that pitching of the airplane needs to be addressed. References Wright, Henry S. et al., ARES Mission Overview - Capabilities and Requirements of the Robotic Aerial Platform, AlAA 2003-6577 Levine, Joel S. et al., Science from a Mars Airplane: The Aerial Regional-Scale Environmental Survey (ARES) of Mars, AlAA 2003-6576 MER Aeroshell Critical Design Review presentations, May 30-31, 2001 Acknowledgements I would like to acknowledge the contributions of the many people who made this development a success: * Gary Qualls for photogrammetry Henry Wright for funding and guidance Charles Bailey and Joseph Hickman for assembling and modifying hardware Tom Lash and Norman McRae for development of the electromagnetic release system Paul Bagby for capturing high-speed video 371

Document

AMS_2006.pdf

“Digital” SMA-Based Trailing Edge Tab Actuators for Aerospace Applications Robert M. McKillip, Jr.* Abstract A novel approach for providing aerodynamic control using collections of discrete active tab devices was completed recently under an Army-sponsored Phase II SBlR program, and is being developed further as part of a NASA-sponsored effort on UAV control. Each tab incorporates a patented bistable design that uses SMA wires for transitioning between one of two deflection positions, thereby providing a localized perturbation to the sectional aerodynamics at the location of the lifting surface trailing edge. By arranging several such tabs spanwise along a wing or rotor blade, incremental adjustments may be made using these tabs to perform aerodynamic reconfiguration, tracking adjustments on rotor blades, or primary flight control. The device has been prototyped as part of these SBlR research programs, and has been designed so that it may be retrofit on existing rotor blade systems and UAV platforms to support further application development in a variety of aeronautical applications. The underlying concept of using SMA- based actuation for configuration control is directly extendable to additional aerospace systems, including spacecraft applications. Rotor Blade Tracking Application Development of the underlying concept for SMA-based actuation of a bistable device originally grew from the desire to provide in-flight adjustment of rotor blades to minimize one-per-revolution vibration. Blade one-per-revolution vibration is often a result of blade aerodynamic and mass mismatch between blades of a given rotorcraft’s “blade set”. This mismatch may be a consequence of assembly tolerance or from unequal wear on in-service blades, and often can be alleviated through placement of balance weights and aerodynamic perturbations via adjustments of both swashplate links and trailing edge tab deflection angles. Aerodynamic adjustments that minimize vibration also typically minimize blade “tracking” deviations, and thus the blade’s track is used to optimize the adjustment process. Traditional blade tracking adjustments are performed by maintenance personnel using specialized hardware, coupled with a series of flight test measurements, to attempt to optimize blade aerodynamics across the helicopter flight envelope. This process is often expensive, as it tends to be iterative in nature. In fact, some current operational military helicopters must devote nearly 10% of their flight hours to track and balance activities. Clearly, in-flight tracking capability would significantly reduce this cost burden, as all tab tracking changes could be made in one flight without the requirement for periodic landings to stop the rotor and perform manual adjustments. In addition, since tracking adjustments would be easier, they could be performed more often, on an as-needed basis, thereby providing “trickle-down” benefits of reduced vibration exposure and hence longer service lives for dynamic components. This motivation spawned the development of the SMA-based actuation system described here, in that the tab deflection adjustments would be performed using SMA wires as the prime mover for an electrically-actuated tracking tab device. “Diaital” Tab lmolementation The system initially developed under U.S. Army Phase I SBlR funding initially replaced the maintenance specialist’s adjustment tools with a pair of agonist-antagonist SMA wires, so-that the imposed moments generated from the SMA wire’s phase change would provide plastic deformation of a metal trailing edge tracking tab mounted on the rotor blade (Figure 1). While successful at the conceptual demonstration stage, this design required a means of sensing the blade tab angle in order to properly control the SMA wire material properties to position the tab deflection to a desired angular orientation. In addition, the hysteresis inherent in the SMA wire’s thermal properties made closed-loop control challenging for this application implementation. Under Phase II, this concept was replaced with a “digital” tab device, whereby each discrete tab uses SMA wires to move between one of two positions, thereby significantly reducing * Continuum Dynamics, Inc., Ewing, NJ Proceedings of the 38‘” Aerospace Mechanisms Symposium, Langley Research Center, May 17- 19,2006 373

Document

AMS_2006.pdf

the complexity of both the associated control system (proportional/servo control becomes simply “on/off”) and the tab position sensing (“up/down”) (Figures 2 through 4). Aerodynamic “resolution” of adjustment is provided through the incorporation of several tabs in a spanwise fashion, thus providing true “digital” adjustment for blade tracking. plastically eformable tab I i- c F SMA actuator wire (one on each side) Figure 1. Phase I concept demonstration for agonist/antagonist SMA-based tracking tab concept. 1 \ L ,/- -- I. 4 I Figure 2. “Digital” rracning rap aernonsrrarion concepr on rraiiiny edge of a CH-47 blade. This design approach change incorporates all the features of the agonistlantagonist plastically-deforming tab, namely: Use of SMA wires as prime movers driving pre-stressed structural components in a hybrid fashion provides large strain and force capability for the actuator system, without the attendant bulk required with other “smart” or active materials for enhancing actuator mechanical advantage Electric power is only used during transitions for changing tab position The devices are retrofit-capable onto existing rotor blades The tabs are adjustable in-flight The physical size and weight is minimal due to the use of SMA wires as the prime mover 0 0 0 0 0 In addition, the digital concept includes the following unique advantages: 0 0 Position sensing is trivial, in that only one of two possible states may exist The precise shape of the device is dictated from geometric features of the base material Control of the device is simplified to a “change position’’ command Use of multiple spanwise units provides robustness as an individual unit failure implies only a partial degradation in total system performance 374

Document

AMS_2006.pdf

* Design of the SMA wire is simplified to specifications on maximum stressktrain required for transition Maximum power required is that required to heat onlv one SMA wire, since the tabs are transitioned sequentially The basic actuation concept is described referring to Figure 3. Figure 3a shows a notched flat plate element that deflects out-of-plane to the plate when points A-B are forced together. The surface buckles elastically into one of two stable positions, determined by the manner in which the points are drawn together. Figure 3b illustrates the plate prior to installation of the actuation wires but installed on the trailing edge of an airfoil. The deflection results from the pre-stress introduced in the plate, and studies to date have shown that trailing edge deflection angles of +25” are achievable for acceptable levels of plate stress. Figure 3c schematically shows the pre-strained SMA wires attached to the upper and lower surfaces which, when heated, result in shortening of the wire and a “snap-through” and locking of the plate into the “mirror” position. The heating is directly accomplished by running a current through the wire, since SMA wires have relatively high resistivity. The innovations in this actuator are two-fold: The structural or load-bearing component of the actuator is a very simple machine (a notched flat plate). This component is readily fabricated and encapsulated to form a useful aerodynamic surface, and has no bearing surfaces. Power is only required to transition the actuator from one position to the mirror position. Once the transition is made, the power to the actuating wire is cut-off and the actuator is “locked” into position as a consequence of its geometric compound curvature. Figure 3. Schematic of “SNAP” trailing edge tab: (a) undeformed base plate (b) stressed plate on airfoil t.e. (c) tab with SMA wires Rotor Blade Tracking Testing Activities Testing of this device to support the rotor blade tracking application included benchtop functional tests, high dynamic pressure tests, high centrifugal field tests, and actual on-rotor testing. These data were used both to support analytical model development for design, and to validate configuration choices for construction and wire attachment. Tab Actuator Benchtop Testing While an analytical model for tab design was under development, a series of tests were performed on an oversize representation of the tab configuration of Figure 3. These tests were undertaken with the goal of generating a database for analytical model correlation, as well as validating a scaling analysis for performing ”nomogram”-like design calculations for similar tab geometries. Aluminum sheets of approximately 4 cm by 8 cm were constructed with a center stress-relief hole, and an offset stud was mounted near the center of the tab, from which a wire cable was mounted in series with a strain gauge load cell. Tests were performed to measure the tab deflection angle with applied cable load, much like an SMA wire would generate on the actuator tab, and notes were made of the displacement and force levels just prior to tab snap-through. These data, taken over a range of tab length to width ratios, thicknesses, offset stud locations, and initial wire pre-strains were ultimately used to validate both in-house CDI finite element (FE) code predictions and commercial FE simulations. Whereas the super-scale tab testing only included one nominal wire orientation for applying pop-through forces on the base tab material, several different arrangements were fabricated and evaluated at design scale on the benchtop. Two of these configurations, termed pre-prototype configurations, are shown in Figures 2 above and 4 below. Figure 2 oriented the SMA wires in a nominal spanwise orientation, while Figure 4 shows the SMA actuator wires aligned with the blade chord direction. Both configurations could be made to operate successfully, but it was found through assembly of several units that wire pre- tensioning was easiest to perform for the chordwise orientation. Thus, the ultimate configuration, th&t of a 375

Document

AMS_2006.pdf

pair of wires (upper and lower) aligned in a "v"-pattern orientation about the stress relief hole borrowed more heavily from the design experience of the device of Figure 4. Figure 4. Pre-prototype benchtop demonstrator tab arrangements, SMA wires mounted on offset studs in chordwise orientation. Several base tab materials were evaluated for use, including spring steel, aluminum sheet, and various plastics. Based upon earlier discussions with Boeing engineers about the desire to remove metals from rotating components on aircraft to reduce aircraft signatures, it was felt that the tab design could best support U.S. Army interests if it were made of non-metallic material. Lexan plastic (polycarbonate) was found to have sufficient robustness to effects of temperature and UV radiation that it became the default base material for subsequent actuator design concepts. In addition, aluminum was ruled out when it was determined that the anticipated stresses built up in the outer layers of the base tab material could experience fatigue failure after many pop-through cycles of the tab actuator. Since further development called for testing of the tab assembly on a full-size rotor system, most likely provided from an aircraft manufacturer or helicopter operator, bonding of the tabs to helicopter blades was required to be a reversible process. Several adhesives and solvents, along with manual installation and removal techniques, were evaluated for their capability to provide sufficient strength in both shear (to counter centrifugal loads) and peel (to maintain bond strength during pop-through and under aerodynamic loading). Tab coupons were manufactured and bonded to aluminum plate and the trailing edge of a composite CH-47 blade section, in order to evaluate shear and peel strength of the combined tablbonding agentblade system. Shear tests were instrumented using an in-line load cell, and peel strength was checked manually. Many bonding agents exhibited good shear strength but poor peel performance. Some bonding materials showed good strength in both directions, but poor resistance to applied heat. Ultimately, an epoxy was identified that showed excellent bond strength, and acceptable solvent-based removal performance, for both metal and composite skin blades; this material was used in all subsequent tests performed on combined blades and active tab assemblies. Two-Dimensional (2D) Section Testing Two separate 20 tests were performed using the active tab assembly. The first was using CDl's 30-cm x 30-cm low turbulence wind tunnel to assess anticipated aerodynamic drag effects of the tab on rotorblade performance. The second was a high dynamic pressure test, to check the two-position tab's potential to pop-through under loading when the section is generating significant lift. This latter experiment used a novel mounting arrangement in order to perform hydrodynamic testing from the side of a small power boat in the Delaware River. Each is described below. CDl's low turbulence wind tunnel was used to conduct tests on two-dimensional airfoil sections to determine the effectiveness of proposed actuation concepts. The 30-cm x 30-cm tunnel can generate up to 36 mls flows, which, although not representative of the anticipated aerodynamic environment for the actuators, was suitable for providing a "first look" at aerodynamic drag effects generated by the addition of trailing edge treatments on the airfoil sections.

Document

AMS_2006.pdf

A two-dimensional section representative of a Bensen B8-M autogiro rotorblade was constructed and installed in the tunnel, mounted on top of a 6-component strain gauge balance. Data from comparison of the baseline section with a full-span tab system shows an increase in a few “counts” of drag on the section (Figure 5). Later testing that included mock-ups of the supporting SMA wires and associated electronics showed these added effects to be minimal, and thus the original data as shown here was used in generating pre-test predictions for full-size rotor tests to follow. 0.07 + - 0 + w/tab wlotab + 0.06 0.05 0.04 0.03 - 0 - + 0 0 + + - 0 +O +O +O - +O +O+a* +o +O Typical tracking tab locations are near the 75% radius location, so for a typical helicopter tip speed between 200 m/s and 213 m/s, the dynamic pressure at the tab location is approximately 40 N/m2. Duplicating this dynamic pressure in a wind tunnel or other aerodynamic facility is difficult and expensive, so an alternative testing method was employed. Since the density of water is approximately three orders of magnitude greater than air, equivalent dynamic pressure may be achieved at speeds of approximately 7.7 m/s. Thus, dynamic pressure tests on a tab-equipped UH-60 helicopter blade airfoil section were performed using a special “oar“ with a representative airfoil and two-position tab attached to the end, with the oar placed into the water aside a small power boat. The airfoil was mounted on a long pole to permit its being placed sufficiently far away from the side of the boat hull to eliminate any surface effects from the boat displacement when traveling under power. Instrumentation consisted of a digital inclinometer for measuring approximate angle of attack of the foil, and a handheld GPS unit for measuring boat hull speed. Runs were taken in both upstream and downstream directions on the Delaware River, and despite extremes of both dynamic pressure and high angles of incidence (up to 0.3 radian), the two-position tab did not “pop” through to its alternate position from differential pressure on the tab. Another operational concern for proper tab actuation was the anticipated high centrifugal environment anticipated for tabs mounted at the 75% radial location. For UH-60 size rotors, this loading can approach 550 g‘s, and thus the tab may “pop through” from inertial loads, or the SMA wire may deform on the tab under its own centrifugal load, effectively “using up” the available strain energy and rendering the tab inoperative. The latter was surmised to be the case for early versions of SMA-based flap actuators, developed by CDI for application to the V-22 tiltrotor blades on a model rotor set; ultimately, the wire- based actuation mechanism had to provide mechanical guides for the chordwise-oriented actuator wires to eliminate the centrifugal distortion. Although the prototype tab actuator developed here has wires 377

Document

AMS_2006.pdf

oriented chordwise, supporting calculations showed that they would not be subject to large strains due to their minimal size and weight. Testing was nonetheless performed on the tabs using a four-foot diameter model rotor, with the operating rpm boosted to 550 in order to expose the tabs to representative centrifugal loads. In these tests, the tabs were mounted at the blade tip to provide maximal centrifugal stresses for the wires. Testing revealed no problems with the tabs reversing their positions at these g-levels, thus providing confidence for ultimately testing the units on a full-size rotor system. Full Size Rotor Testinq The full-size rotor system used for evaluating the prototype actuation system was a Bensen B-8M autogiro ("gyrocopter"), available for use at CDI (Figure 6). The advantage of using the autogiro test platform is the ease of access to a man-rated rotor system, with all of the system integration issues associated with a full- size flight vehicle. Since that aircraft is an autogiro, and not a helicopter, an alternate means of spinning the rotor was required so to avoid having to fly the aircraft with the tab devices installed. This drive system consisted of a pre-rotator assembly that used a flexible shaft that could be driven by a pulley mounted inboard of the propeller hub on the autogiro's engine. I Figure 6. Autogir, .... .. ".. ... .....e r,u.e... uue.. ... .u.m-uize rotor tests. Test objectives were twofold for the full-size rotor experiments: first, it was desired to demonstrate tab effectiveness in controlling tracking angle of the rotor blades from positioning of individual tabs on the rotor; and second, it was desired to assess the rpm range over which reliable electrical actuation of tab displacement could be achieved. Tabs were installed onto the Bensen blade set from the 0.71 to the 0.80 radius location, with each tab assembly fabricated using an aluminum base plate holding four 2.5-cm by 5-cm tabs. Three tab plates were mounted on each blade, with a Lexan separator plate bonded between the blade surface and the mounting plate in order to facilitate removal of the tab system from the blade at the conclusion of the test program (the removal agent for the tab system is known to attack certain plastics). To simplify the test program, only one set of four tabs was fabricated using SMA wires, since the purpose of the test was to demonstrate performance and not provide a complete system installation. As a result, the control electronics were mounted at the rotor hub, and separate flat ribbon wires were attached to the bottom of the blade to provide electrical power to each SMA tab. While a true prototype installation would include local surface-mounted MOSFETs for control of individual SMA wire excitation at each tab plate, this simplification helped expedite the completion of the full-size rotor test program. Details of the tab mounting arrangement may be seen in Figure 7, and the hub electronics in Figure 8. 378

Document

AMS_2006.pdf

Figure 7. SMA-driven bi-modal tabs on underside trailing edge of Bensen B8-M blade section. LU iftj Rotor tachometer Figure 8. Details of hub on Bensen autogiro test platform. In order to check blade track angle, a separate video camera installation was mounted in the rotor plane, with a calibration card constructed and the blade ends colored (white/black) to differentiate them on the videotape. The videotape was then viewed on a frame-by-frame basis to measure the blade tip position as a function of tab deflection angle. For the extreme case of all tabs on the “black” blade deflected down, and all tabs on the “white” blade deflected up, a tip deflection difference of 3.9 cm was measured from scaling displacements from the video monitor. This compares favorably with a predicted value of 4.5 cm, since the predicted displacement does not account for any lift rolloff at the most inboard and outboard locations of the deflected tab assemblies. Screen capture of the rotor plane video may be seen in Figure 9. 379

Document

AMS_2006.pdf

Figure 9: Still video of white and black blade position from rotor plane camera. Post-test analysis and further development testing revealed that the Lexan plastic material used to form the base bender element eventually took a “set” due to the continued exposure to bending stresses, and thus developed a preferential orientation that would make the tabs less robust. As a result, further development produced a tab with equivalent stiff ness but no tendency to plastically creep, due to the use of spring steel for its bender element. Shown in Figure 10 is a demonstration article that incorporates a steel base with Kapton tape overcoat and flush attachments of the SMA wire assemblies, along with the flatpack 6 volt battery used for its actuation. Figure 10. Metal-based tab demonstration device with flatpack battery. Distributed UAV Flight Control Application High-altitude, long-endurance (HALE) remotely-operated aircraft (ROA), or unmanned aerial vehicles (UAVs), are designed to operate at extreme altitudes for many days, in order to serve as platforms for science missions, earth observation, or communications and data relay stations. In order to meet the long duration requirements for these missions, these aircraft are designed to minimize drag and weight, and thus are characterized by having large wing spans and significant aeroelastic response under both steady and dynamic loading conditions. This inherent flexibility in these aircraft may sometimes lead to flight dynamics issues and instabilities, as witnessed with the breakup of the HELIOS aircraft over Hawaii during flight investigations while operating with alternate energy sources. These aeroelastic issues may be mitigated to a considerable extent for these vehicles if the loading on the large aspect ratio wings may be redistributed in flight, using distributed multiple actuators. Such an approach provides enhanced robustness to the aircraft platform, in that multiple actuators would easily accommodate the failure of a few to respond to commands, but if these actuators were conventional, servo-type systems, this feature would be overshadowed by an unacceptable weight and power penalty. 380

Document

AMS_2006.pdf

Conventional electric servo actuators incorporate electric motors that drive a gear train that ultimately is connected to a control arm, on which is also mounted a potentiometer for sensing actuator displacement. Thus, the positioning control and the actuator impedance are provided only while the electric motor draws power from its energy source. Use of the SMA-based actuator described for the rotor blade tracking application would be an ideal alternative, in that it would permit spanwise distributed load modification with minimal electrical power expended during the reconfiguration process. NASA is presently sponsoring CDI in an SBlR effort directed at investigating the use of these snap-through SMA-based actuators in support of HALE-type UAV aircraft control. Results to date in this investigation are described below. UAV Application Desican Challenqes Since the detailed design specifications for a future NASA HALE RONUAV are yet to be completed, it was felt prudent to consider HALE vehicle operating features in the actuator design, but size the actual device to be more immediately applicable to potential test platforms, such as CDl’s 2m span electric radio control (FUC) model aircraft, and the two NASA Dryden’s APV-3 test aircraft. Fortunately, these two platforms are roughly the same size, and thus it was decided to base the actuation sizing on a compromise that would accommodate each aircraft type. Fundamental to each actuation concept is its application in a distributed fashion, so that several (possibly many) of them would be arrayed along the trailing edge of the UAVs lifting surfaces. Sizing of each individual actuator was selected to provide: Desired flight control forces/moments, dictating a minimum ratio of actuator to surface chord length; Appropriate speed of actuation, thus impacting maximum wire diameter used; Sufficient stiffness to support operational aerodynamic loads, fixing a minimum thickness for the base material. 0 0 Key factors in the design considerations applied included operational environment, endurance, robustness, actuator weight and power consumption. HALE RONUAV systems operate at altitudes in which the ambient air temperature is approximately -48”C, and thus the actuator may require insulation on its wires to mitigate the power lost through ambient cooling to the atmosphere. Endurance requirements for these actuators would flow from possible HALE missions lasting from several days to several weeks. While use of multiple, distributed actuators provides some robustness relief in that single actuator failures represent graceful degradation of control capability (vs. catastrophic), the actuators nonetheless should be capable of accommodating multiple flights/missions without replacement. Standard practice for SMA wire- based actuators has shown that as long as the wire strains are kept below 3% in actuation cycles, millions of repetitions are possible. For the actuation systems here, actual wire stress changes in the discrete actuation system varied from 0.5% to 1.0% depending upon end termination of the wire that stretches across the base tab material. Robustness design goals imply that all material stresses are kept well below limiting values, and the opportunity for failure be mitigated as much as possible. The actual stresses on the base plate assemblies are well within linear bounds below yield, and failure modes were reduced in the actuation elements by reducing the number of parts required in assembly. Toward this end, use was made of custom-fabricated SMA wires having metal balls welded onto their ends. By using a welded ball with a diameter twice that of the wire, the wire could be “dropped into place” on the base tab assembly through insertion into a slot created in the base manufacturing process. This approach provided a convenient mechanical means for coupling the SMA wire to the base material (Le., the actuator armature), but continued experimentation with this approach revealed that the stress imposed at the junction of the weld could be past yield if the wire were tensioned at an angle nearly parallel to the surface of the base tab. Put another way, the ball end fitting was not adequate if the wire had significant non-axial stresses at the weld point. This observation resulted in a design having an offset mount for the actuating SMA wires. Weight considerations on the actuation schemes, both continuous and discrete, were actually a secondary issue, as it was a fall-out effect from material choice based upon actuator performance requirements. Basic stiffness requirements dictate a minimum actuator thickness, and the fact that the buckling- enhanced actuation concept routinely induces stresses on the base material showed in previous actuator implementations that plastic base structures are inadequate due to their creep properties. Thus, the design of these actuation systems has routinely used tempered steel (spring sheet stock) for the base plate material. Despite the use of steel components, the net weight of these actuators, when compared against conventional servo-system components, is almost insignificant. 38 1

Document

AMS_2006.pdf

Power consumption issues, however, show a clear bias in favor of the use of discrete actuator concept, when possible, over any continuously deflectable device. When comparing power required to hold a given actuator position, the discrete actuation concept requires none, while a conventional servo-based (motor driven) actuator will require steady currents of roughly comparable levels. This is in large part due to the fact that static servo holding stiffness operates the underlying electric motor at near stall torque levels, a very inefficient regime for these electromechanical components. Protracted operation holding a given deflection would impose a significant power drain on the HALE aircraft, limiting its endurance. Two design variants were investigated, where the wire was electrically actuated at each end with a potential applied across the entire wire length, and a second where the wire was grounded on both ends and an additional electrical connection was made at the center of the SMA wire. Initial work with the center actuated design revealed that adjustment of wire tension through the height change of the center actuation point was difficult to achieve repeatedly. Subsequent work with the end-point actuated tab system showed that it suffered in being able to repeatedly actuate the system without shearing off the ball end fittings. Thus, the final configuration realized in this design process was a hybrid design that incorporated features from both the center point actuation design and the end point actuation system. Shown in Figure 11 is the dual tab assembly, where two SMA wire-actuated discrete tabs are mounted to a “sandwich” of printed circuit board, with the two tensioning straps having a triangular shape added to promote the snap-through effect of the base tab when the wires are tensioned. This tab design also incorporated customized screw-adjustable crimps at the inner edge of the tab assembly, permitting rapid wire tensioning for optimum tab performance. Figure 11. Final Phase I configuration for dual discrete tab system. Electronic lntewation Desiannestinq Electronics and software design for controlling the groups of individual tab actuators has been investigated as part of the integration work, since the collection of multiple tabs would be interfacing to a single standard servo connector for both the flap and the aileron assemblies for applications to both CDl’s R/C airplane and the APV-3 wing. While having multiple tab actuators provides a means of mechanical redundancy for providing trailing edge control, both signals and power must be supplied to all to make them function as a group. Brute force approaches for directly wiring each and every tab actuator to a central controller would be unacceptably heavy and complex, due to the massive amount of wire required for providing current to each individual SMA wire. Instead, electronics incorporating extremely small microcontrollers is being used for a serial-bus network that determines which actuator to energize and for how long. The target microcontroller is the Microchip PIC1 2F675 family, an 8-pin SO-8 surface-mounted device that has an internal clock and is reprogrammable in-circuit via FLASH memory. Two schemes were investigated for localized control and communication on this serial bus network for interconnected tab actuators. The first involves the direct device addressing from a “master” microcontroller for actuation, whereby a master/slave(s) network is constructed with the master requesting 382

Document

AMS_2006.pdf

an individual tab to change its state through serial commands that include the identifier of the tab device. The second scheme attempts to eliminate the assignment (and/or cataloging) of individual tab actuator identification codes by incorporating a “daisy-chain” interconnection of the tabs, where the most inboard tab microcontroller strips off the first command signal (pulse) and sends the rest down the serial databus to all the other actuators. This scheme allows for identical software to run on all tab actuator microcontrollers, but interrupts the signal coming from the master unit at each tab station, potentially impacting robustness. However, this same re-broadcasting technique can also be viewed as boosting the signal of the transmitted pulse train as the signals propagate spanwise from inboard to outboard actuator unit. Printed circuit boards compatible with the dual tab design seen in Figure 11 are shown in Figure 12, along side the tab assembly and a pen for scale comparison. The board has solder pads that directly accept a programming header so that software changes may be made in-circuit, if necessary. Figure 12. Dual dual power MOSFETS. Installation of the actuators on CDl’s model aircraft was expedited through the use of an extra control channel for the model. Since this model lacks a rudder, an additional control channel was available from the hand controller that had an output on the aircraft receiver, but not servo actuator that would connect to it. Thus, an auxiliary roll command channel was constructed using the output of the rudder command to drive a “master” microcontroller that, in turn, sent out serial daisy-chain tab command signals to both the left and right wing actuators. This model arrangement was flown, and found to exhibit moderate roll effectiveness, despite the fact that a full span complement of active tabs was in fact not driven from these serial command sequences. Testing on this configuration continues, along with integration efforts to support tests on NASA’s APV-3 aircraft. Other Possible Aerospace Applications Other aerospace applications suggest themselves for use of this actuation concept. While initially developed as a trailing edge control component, the fundamental nature of an SMA-actuated over-center spring like system has a variety of other possible uses. The essential features of this actuation concept are captured in a simple example. If a structural member, such as a beam or plate, is buckled and placed between two end constraints (either pinned or cantilevered), the beam (or, actuator “armature”) will assume one of two possible equilibrium shapes, representing the minimum stress condition that satisfies the end conditions (Figure 13). By attaching pre-strained, opposing SMA wires on each side of the buckled member, and heating the most elongated SMA wire, one may induce a moment within the buckled structure (at the SMA connection points) that forces the base material back toward its other equilibrium position. As this induced moment is increased, the buckled structure will eventually move to a shape where the incremental moment required to produce additional motion vanishes, and then rapidly switch, or “snap” or “pop”, to its other equilibrium position. This behavior is often called “snap-through” or “oil- canning” in the structural community, and is characterized by linearly elastic material properties but large (nonlinear) deformations. 383

Document

AMS_2006.pdf

upper SMA wire (stretched) I trailing edge snapped-through overhang configuration lower SMA wire (contracted) Figure 13. One-dimensional example of a “snap-through” SMA-driven actuator, showing pinned constraints and optional overhang. The actual work performed by this type of actuator as prime mover can come from several sources: 0 0 0 0 the linear motion of the center of the buckled beam may be used to drive a linkage between two different locations; the change in slope of the end points of the beam may be incorporated as a form of angular orientation or camber control; the snap-through in one direction may influence the convex or concave shape of a doubly-curved planar surface (such as on the trailing edge tab device above); or the buckled structure may induce relative forces and moments between two disparate locations on a flexible base structure (particularly if the base material has cantilever end constraints). A simple comparison with a notional miniature linear actuator for aerospace use [I] reveals this concept’s considerable advantages over motor-driven mechanisms. If one matches the end-point stroke and displacement of the device of [l] using the linear motion of the snapping arch shown above, one realizes a weight reduction of 91%, since the pre-stressed snap-through element performs the same useful work as a stepping motor/ball screw combination. This weight savings is effectively due to the replacement of a motor-driven system with a “reversible actuated spring” unit. Of course, similar devices may be combined in series or in parallel to provide complex, large deformations in multiple directions as a complete mechanism system. Several examples will be sketched below to better visualize these configurations. Figure 14 shows a method for forming a bi-stable actuator from a flat plate that produces a change in slope between its two ends. The center portion of the stamping comprises the buckled member, while the two side strips, when joined as shown, provide the geometric constraint that keeps the center piece pre- stressed. SMA wires mounted on the upper and lower surface of the buckled member are used to “pop” the device between the two possible stable positions. Figure 15 shows a three-position actuator, similar to that in Figure 14, but uses two nesting pre-stressed members that are cantilever-mounted on either side of a cut-out rectangular base support. By snapping one or the other pre-stressed arches through the cut- out area, one can generate three different slopes between the ends of the base support, or equivalently, an up-neutral-down type of output for the end displacement (Figure 16). Simple extensions to that mechanism, ones that include other nested buckled members above and below these two, provide even higher “resolution” of the relative angles between the two end points of the buckled members. Combinations of these three-or-more-position actuators distributed across a planar surface could provide complex warping of the base plate or planar surface to provide localized camber and shape control. Several actuators ganged in series would produce additive changes in base material slopes, or, very large end displacements (Figure 16). This large motion response is possible even while limiting the strain of the SMA wires to less than 3%, a level that is suitable for millions of cycles of operation. 384

Document

AMS_2006.pdf

join together (b) assembled, % view (c) popped-through, 34 view join E together (a) stamping Figure 14a-c. Bi-stable actuator formed from a flat plate, incorporating three parallel strips. Bender - _. d) demonstration device implementation and actuation response Figure 14d. Bi-stable actuator formed from a flat plate, incorporating three parallel strips. Partial views of additional, co-planar 3-position actuators Figure 15. Co-planar three-position actuators for compound geometry control. Force applied from SMA wires (not shown) snapped "up" snapped "down" Figure 16a. Three-position actuator showing base support bending from snap-through action. 385

Document

AMS_2006.pdf

Figure 16b. Three-position actuator in series arrangement for large displacement application. Conclusions An actuation system incorporating SMA wires to reorient a buckled member has been described, with a focus on its use as a trailing edge control device for aerodynamic modifications. Applications to helicopter rotor blade in-flight tracking, and UAV spanwise aerodynamic load control have shown that the small form factor, minimal power requirements, and simplicity of design make it an attractive alternative to conventional aerodynamic control effectors. Future use in other aerospace applications will benefit from the attractive weight and size offered in such a device. References Asadurian, A., “Miniature Linear Actuator,” Proc. 34th Aerospace Mechanisms Symposium, NASA CP- Bilanin, A. J., and McKillip, R.M., Jr., “Actuation Device with At Least Three Stable Positions,” U S. Patent McKillip, R.M., Jr., “Remotely Controllable Actuating Device,” US. Patent No. 5,752,672, Max 19, 1998. McKillip, R.M., Jr., “”Digital” Tracking Tabs for One-per-Rev Vibration Reduction,” Proc. 59 AHS Annual 2000-209895, May 2000. No. 6,345,792, February 12, 2002. Forum, Phoenix, AZ. Mav 2003. .. McKillip, R:M., Jr., “Actuating Device with Multiple Stable Positions,” U.S. Patent No. 6,220,550, April 24, 2001. 386

Document

AMS_2006.pdf

Development of a Forced Oscillation System for Measuring Dynamic Derivatives of Fluidic Vehicles B. C. Trieu’, T. R. Tyler*, 6. K. Stewa3, J. K. Charnock’, D. W. Fisher*, E. H. Heim’, J. Brandon, and S. 6. Grafton’* Abstract A new Forced Oscillation System (FOS) has been designed and built at NASA Langley Research Center that provides new capabilities for aerodynamic researchers to investigate the dynamic derivatives of vehicle configurations. Test vehicles may include high performance and general aviation aircraft, re-entry spacecraft, submarines and other fluidic vehicles. The measured data from forced oscillation testing is used in damping characteristic studies and in simulation databases for control algorithm development and performance analyses. The newly developed FOS hardware provides new flexibility for conducting dynamic derivative studies. The design is based on a tracking principle where a desired motion profile is achieved via a fast closed- loop positional controller. The motion profile for the tracking system is numerically generated and thus not limited to sinusoidal motion. This approach permits non-traditional profiles such as constant velocity and Schroeder sweeps. Also, the new system permits changes in profile parameters including nominal offset angle, waveform, and associated parameters such as amplitude and frequency. Most importantly, the changes may be made remotely without halting the FOS and the tunnel. System requirements, system analysis, and the resulting design are addressed for a new FOS in the 12- Foot Low-Speed Wind Tunnel (LSWT). The overall system including mechanical, electrical, and control subsystems is described. The design is complete, and the FOS has been built and installed in the 12- Foot LSWT. System integration and testing have verified design intent and safe operation. Currently it is being validated for wind-tunnel operations and aerodynamic testing. The system is a potential major enhancement to forced oscillation studies. The productivity gain from the motion profile automation will shorten the testing cycles needed for control surface and aircraft control algorithm development. The new motion capabilities also will serve as a test bed for researchers to study and to improve and/or alter future forced oscillation testing techniques. Introduction & Background Forced oscillation testing is traditionally used to investigate the dynamic derivatives of vehicle configurations. A model is oscillated one frequency at a time at various amplitudes. Balance data, angular position and rate data are measured and recorded. The data set corresponding with a model mass and inertias is reduced to determine the dynamic derivative coefficients. The coefficients are used in damping characteristic studies, and used in simulation database for control algorithm development and performance analyses. Test vehicles may include high performance and general aviation aircraft, re-entry spacecraft, submarines and other fluidic vehicles. Historically, the forced oscillation testing hardware has been an induction motor driving a crank and linkage mechanism to provide oscillatory motion. Figure 1 shows an existing system, formerly designed and built for the 30- by 60-fOOt Langley full-scale tunnel that can be configured to oscillate a model in roll, yaw and pitch. The system is limited to sinusoidal motion having the frequency controlled by motor speed and the amplitude determined by the mechanical linkage. Oscillatory frequency may be changed remotely by changing the motor speed. However, to change amplitude, the oscillation hardware and the tunnel must be halted and the linkage mechanism manually reconfigured. ** NASA Langley Research Center, Hampton, VA Vigyan, Inc., Hampton, VA Proceedings of the 38 Aerospace Mechanisms Symposium, Langley Research Center, May 17- 19,2006 387

Document

AMS_2006.pdf

YAW GONFIQURATW PITCH ooNpx3uwTIOw Figure 1. Existing Forced Oscillation System (FOS) The existing system is operationally inefficient thus keeping wind tunnel productivity low. Beside the need to halting the wind-tunnel and the FOS system for amplitude change, the inherent one frequency at a time data gathering is slow and time consuming. Many data points are needed to construct a frequency response curve for any given aerodynamic derivative coefficient, and there are many coefficients for any given vehicle configuration. In addition to operational inefficiency, the motion generated by the existing system is limited to sinusoidal profiles only. Other motion profiles including ramp and arbitrary waveforms are useful in aerodynamic testing. Low-speed constant ramp may be useful for quasi-static testing; and arbitrary waveforms as used in Schroeder sweeps can improve dynamic testing efficiency by measuring the full frequency response spectrum of the aerodynamics behavior instead of the traditional method of one frequency at time testing’. The vision for a new FOS is to create a system that greatly enhances forced oscillation testing capabilities. The new system should be able to oscillate not only in sinusoidal motion but also other profiles that may be beneficial for aerodynamic testing. The system should enhance productivity such as reducing model setup time and reducing run time needed to collect a data set. System Requirements The overall design goals of the new FOS included the ability and flexibility to oscillate a model in sinusoidal as well as non-sinusoidal motion profiles. The system should be able to operate safety and efficiently - it should be able to operate with different amplitudes and frequencies as well as different motion profiles without the need to stop the test and mechanically change the system. Technical requirements for use in the design of the system were derived from free-flight models. Often, it is desirable to use the same model for lower development cost; Le., the same model design may be used in wind tunnel, free flight, and drop model tests. Using the same model type of same scale has the added benefit of avoiding potential inconsistencies which can occur from dynamic scaling. The two model classes used to establish requirements were based on typical fighter configurations with characteristics listed in Table 1. 388

Document

AMS_2006.pdf

Table 1. Model Characteristics Traditionally, models are sinusoidally oscillated to determine the dynamics derivatives of a vehicle configuration and then scaled for use in the full-size vehicle control and stability system studies. For the model sizes in Table 1, Figure 2 shows the theoretical desired frequency versus amplitude of oscillation for a non-dimensional frequency constant k of 0.18. For the selected constant k and model classes, the dynamic free stream velocity is usually limited to low speed dynamic pressures'. 0 10 20 30 40 50 Oscillation Amplitude (Degrees) Figure 2. Theoretical frequency vs. amplitude of oscillation The 12-Ft Low-Speed Tunnel was selected as the primary facility for the new FOS. The tunnel has a dynamic pressure of up to a Q of 48 kPa (7 psf) (V =23.5 m/s (77 ft/sec) at standard sea level conditions) which meets the frequency scaling factor and satisfies the low speed requirement. In addition, this tunnel has an arc-sector model support system that provides pitch and yaw static positioning of a model and can be modified for use with the new forced oscillation system. However, the existing 12-Ft arc-sector support model system was designed for static testing only with loading limitations as listed in Table 2. The model support system needs to be analyzed for the additional dynamical forces to prevent overloading and to prevent undesirable structural vibrations. From the static load limitations, the existing model support system can sustain the 200-lb model class but with reduced normal force capacity (712 N or 160 Ibf). Along with model support system limits, the balance used to measure dynamics and aerodynamic forces during testing also imposes additional design constraints to the system. For both the 90-lb and 200-lb model classes, the FF-10 balance was used. Table 2 also lists balance hading limits. Additional derived requirements for performance and operational efficiency included: Range of displacements * 170 deg System accuracy * 0.05 deg Finally, the new FOS must be integrated with the existing model support system, data acquisition system and tunnel safety operations. 389

Document

AMS_2006.pdf

Table 2. Loading Limitations Design Analysis Concept studies were performed to arrive at a system design that would satisfy the design goals for conducting dynamic derivative studies. Instead of using mechanical linkages to achieve motion profiles, the new design is based on a tracking principle where a desired motion profile is achieved via a fast closed-loop positional controller. The motion profile for the tracking system is numerically generated and thus not limited to sinusoidal motion. It permits non-traditional profiles such as constant velocity and Schroeder sweeps. This new design approach simplifies the mechanism design but requires more emphasis on drive system and motion control design as well as system integration. From the model classes and motion profile requirements, inverse analyses were performed to determine the required actuating torque and speed to achieve the desired frequency and amplitude. Figure 3 is a typical inverse dynamic analysis. Using the roll inertia for the 200-lb model, the peak torque of 246 N-m is required to generate an oscillation of * 10 degrees at 2.5 Hz as shown. Analyses were performed for frequency and amplitude combinations as shown in Figure 2. Table 3 summarizes the inverse dynamic analysis results for sinusoidal motion profiles meeting the theoretical frequency versus amplitude of oscillation curve. Anaular Position Q E -0.2 o.~~ 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 08 09 1 Angular velocity 92 ;o (r -2 -4 0 0.1 0.2 0.3 0.4 0.5 06 0.7 08 0.9 1 Angular Acceleration N 50 50 -50 o 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 I Torque 200 -200 z fo 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Sec. Figure 3. Inverse analysis result for a motion profile of 10 degrees at 2.5 Hz 390

Document

AMS_2006.pdf

Table 3. Inverse dynamic analysis results for determining peak torque From the inverse dynamic analyses, the torque and speed required to generate the sinusoidal motion are used to design and size the actuator system. Table 3 indicates that a torque multiplier is needed to meet the high torque and low speed requirements. High performance servo drive systems (for low and medium inertia rotors) of interest are capable of about 3000 RPM at a rated constant torque. The required high output torque and cyclic loadings lead toward a cycloid drive for long life and high momentary overload margin. The reduction ratio of the cycloid torque multiplier is selected to match motor speed and rotor inertia with expected total inertia load3. Typical high performance servo drives are tuned to operate most stable when having a reflected inertia ratio of about 51. Optimization of design parameters including motor torque, speed, reduction ratio, reflected inertias and spatial constraints for packaging led to the system as designed. Another consideration is the dynamic loading on the existing arc-sector model support system generated by the new FOS. The system was designed for static loading only. The dynamic load induces unwanted vibration and may excite natural modes in the system. Modal analyses were performed to determine the natural frequencies and modes of the modified model support system. Figure 4 shows the first 5 modes and their frequencies. As shown, the lowest natural frequency for mode 1 is near 4.55 Hz. This is near the high frequency end of the theoretical oscillation curve shown on Figure 2. Care must be taken to operate the FOS near this high frequency range. Other than that, the modified model support system can accommodate the dynamic frequencies induced by forced oscillation testing. System Description Figure 5 shows the new FOS as installed in the 12-Foot LSWT. Figure 6 shows the schematic of the overall system as implemented in the tunnel. As shown, the system may be controlled from the test section or from the control room via a laptop computer. The LabVlEW based PXI real-time controller (RTC) from National Instruments has custom developed software modules that manage the FOS including system safety, communication, operations, and motion control according to user input and system feedback. The system safety module monitors and controls the 230VAC power to the servo drive system, safety interlocks including control-keyed switches, Emergency Stops, end-of-travel limits (software and hardware), and communications among user interface computers, and RTC controller. The system also monitors safe operations from the existing arc-sector model support system. The communication module connects the RTC with the user interfaces to process user inputs and provides system status. The motion control module executes the positional closed-loop algorithm to track motion profiles as requested by a user. The block diagram in Figure 7 shows how the motion tracking is achieved. As shown, the velocity loop of the motor is internally managed and controlled by the BDS4 servo amplifier. The control algorithm commands the velocity signal based on the error signal generated by the commanded motion profile and the encoder feedback signal. For system verification and validation, a PID algorithm is used. For more complex motion profiles, different control algorithms may be explored in the future. 391

Document

AMS_2006.pdf

i Mode 3 Q 6.23 Hz Twisting about Mode 1 @ 4.55 Hz Side to Side Bending u, * Mode 4 Q 6.41 Hz Sting Bending Mode 2 @ 4.90 Hz Fore & Aft Bending Mode 5 Q 39.94 Hz Side to Side Bending w, Figure 4. Modal analysis results for the arc-sector model support system 392

Document

AMS_2006.pdf

I Figure 5. FOS installed in the 12-Foot Low-Speed Wind Tunnel (LSWT) --=’- 12 FT. LOW-SPEED WIND TUNNEL FORCED OSCILLATION SYSTEM FOR ROLL AXIS Figure 6. Schematic of the overall FOS in 12-Foot LSWT 393

Document

AMS_2006.pdf

I 1 i veiouty Loop ! PXI: Real Time Controller I I: I ; (W Feedback) I ;( ---------------- - I- Tandler -1 sumnorno Servo- I Gearbox Match Gearbox I ~~g mcle (FS 00) ineaa Right Angle -+ (~~25) Figure 7. FOS block diagram The mechanical subsystem of the new FOS is defined by the hardware located in the test section of the wind tunnel. This subsystem consists of all the drive system components and mounting hardware. Figure 7 shows a FOS block diagram which provides the flow description for all the major components, and Figure 8 shows a cross-section of the mechanical subsystem. The existing arc-sector model support system provides static orientation in pitch and yaw; and the new FOS provides roll attitude and dynamic motion control. The FOS is driven by a high performance Kollmorgen servo motor. This servo motor orporates high energy rare earth neodymium-iron-boron magnets to provide a high torque-to-rotor inertia ratio as well as exceptional continuous torque and peak torque performance. This motor has a peak speed of 3600 RPM with a peak torque of 13.8 N-m and a continuous stall torque of 4.7 N-m. Right angle spiral. bevel gearb Angular contact bearing set Existing model support system Figure 8. Cross-section of the FOS assembly Minimizing intrusion to the tail end of the model and potential aerodynamic intetference, the overall length of the FOS was shortened by utilizing a right angle drive at the motor interface. Even though this increased the frontal aerodynamic area of the FOS, the motor is still located within the frontal area of the existing model support arc-sector and therefore overall tunnel flow is minimally affected. The right angle drive is a Tandler spiral bevel, low backlash gearbox. The main function of this gearbox is to optimize 394

Document

AMS_2006.pdf