cetiennec/heat-pipe-model-identification

Heat Pipe Control Data

Simulated time-series data from a 450-meter heated pipe loop, generated for the Heated Pipe System with Heat Loss tutorial project comparing PI, MPC, and Reinforcement Learning control strategies for regulating outlet water temperature near its boiling point.

Dataset Details

This is an open-loop system identification experiment: the heater command is driven by a fixed/exogenous input sequence (not a feedback controller), so the resulting inlet/outlet temperature trajectories can be used to identify the pipe's thermal dynamics (delay, time constants, heat-loss coefficient, etc.).

The data comes from a physics-based simulation of a circular pipe loop with:

• Continuous water flow through a 450 m pipe
• A heater at the pipe entrance
• Convective heat loss to the ambient environment along the pipe
• Water recirculation (closed loop), producing a stair-like temperature response
• ~12 minute transport delay between heater input and outlet temperature response

File

pipe_heater_data.csv — 288,366 rows, sampled at ~0.01 s intervals (100Hz, ~2884 s / ~48 min of simulated time).

Column	Description	Units
Time	Elapsed simulation time	seconds
Tinlet	Water temperature at the pipe inlet (heater outlet)	°C
Toutlet	Water temperature at the pipe outlet	°C
Tambiant	Ambient temperature surrounding the pipe	°C
HeaterCommand	Heater on/off command	0/1, normalized — actual heater power = HeaterCommand × 33 kW (heater power is clipped to the 0–33 kW range)

Physical Parameters

You can create a physics simulation using the following pipe/fluid parameters (from params.py), provided here for reference. Note these are the identified parameters used in the model, not raw physical measurements — e.g. the actual pipe is closer to 300 m long, but parametric identification against real behavior led to the values below :

Parameter	Value	Description
Pipe length (L)	450 m
Pipe diameter (D)	22.2 mm
Fluid velocity (u)	0.8 m/s
Convective heat transfer coefficient (h)	2.7 W/m²·K
Water density (rho)	1000 kg/m³
Specific heat capacity (c_p)	4186 J/kg·K
Number of pipe segments (N)	30
Initial pipe temperature (T_init)	15 °C
Heater power range	0–33 kW	HeaterCommand × 33 kW

Use Cases

• Training/evaluating time-series forecasting models
• System identification of the pipe's thermal dynamics
• Benchmarking control strategies (PI, MPC, RL) offline against real simulated trajectories
• Reinforcement learning environment replay/analysis

Source

Generated by the simulation code in the heat_pipe_control GitHub repository.