J021: RW FI Onboard 3-Axis Controlled Satellite Using Enhanced Random Forest + Multidomain Features скачать в хорошем качестве

J021: RW FI Onboard 3-Axis Controlled Satellite Using Enhanced Random Forest + Multidomain Features

Rahimi A., Folami M.* (2021, Oct.) Reaction Wheels Fault Isolation Onboard 3-Axis Controlled Satellite Using Enhanced Random Forest with Multidomain Features, International Journal of Prognostics and Health Management, from PHM Society, Volume 12, Issue 2, pp. N/A, doi: 10.36001/ijphm.2021.v12i2.3078 This research paper details a data-driven methodology for identifying and isolating faults in satellite reaction wheels, which are essential components for maintaining spacecraft orientation. The authors propose an enhanced machine learning approach that utilizes an automated library to extract spectral, temporal, and statistical features from simulation data. By testing various classifiers such as Random Forest and Gradient Boosting, the study demonstrates high accuracy in detecting both abrupt and transient faults across different wheel configurations. The framework specifically addresses the computational limitations of smaller satellites by prioritizing efficient algorithms over resource-heavy deep learning models. Furthermore, the paper provides a rigorous sensitivity analysis to evaluate how the system handles sensor noise, missing values, and failed measurement components. Ultimately, the results indicate that Random Forest classifiers offer a robust and reliable solution for enhancing the safety and health monitoring of orbiting satellites.