A Hybrid Hardware In-Loop Approach to Fault Prediction Simulation in satellite ADCS architecture based on 3-Axis Reaction Wheel: Integrating Probabilistic and Statistical ML Models with Traditional Control Systems
- Paper ID
93034
- author
- company
Indian Institute of Technology Kanpur
- country
India
- year
2025
- abstract
Satellites have played a crucial role in advancing science by enabling global communication, weather forecasting, and environmental monitoring. They provide valuable data for climate research, disaster management, and Earth observation. It also support space exploration, contributing to our understanding of the universe. Satellite system testings and ground based data validation is at the utmost importance in any space mission as it validates the system performance, identify potential failures, and ensures all components function properly before launch. It helps simulate space conditions, reducing risks and ensuring mission success by identifying and fixing issues early. HIL (Hardware In-Loop) Simulation used in satellite missions is used to validate the software systems with hardware in operational conditions. HIL involves running of real time sensor in a control loop to generate the resultant datasets, the paper will include IMU dataset (Euler Angle Rate) as the input values running in the loop. The system proposed integrates probabilistic and statistical methods like Hidden Markov Model (HMMs) for fault prediction and Statistical Machine Learning Models Like Random Forests(RFs) approach used by NASA Jet Propulsion Laboratory to predict on board failures and monitor the health of architecuture of Mars rovers. The hybrid methodology aims to combine the ML Models and Traditional control systems which uses KF (Kalman Filter) for sensor noise reduction. The Sensors used in the HIL are: IMU along with 3-Axis reaction wheel and Magnetotorqures as the primary and secondary actuators (Running in the Controller Loop) respectively. The study will utilize MATLAB SIMULINK to create the model and generate the datasets. The HIL will study the datasets of uncombined traditional control architecture and hybrid architecture with fault prediction, and study the importance of integrating ML with traditional control systems for future space missions.