ESTIMATION OF REACTION WHEEL DEGRADATION AND ITS IMPACT ON ATTITUDE STABILIZATION OF A LOW EARTH ORBIT SATELLITE

Reaction wheel (RW) bearing degradation is a primary driver of precision-pointing loss and premature end-of-life in low Earth orbit (LEO) satellites. This paper presents a physics-guided simulation framework for estimating RW degradation and quantifying its impact on three-axis attitude stabilization. The framework integrates rigid-body attitude dynamics with a thermally-coupled Stribeck friction model that captures four inter-dependent failure modes friction coefficient growth, lubricant depletion, rotor imbalance, and electrical resistance growth together with realistic LEO disturbance torques (gravity-gradient, aerodynamic drag, geomagnetic, and solar radiation pressure) and a machine-learning diagnostic layer, all implemented in a single reproducible Python simulation. Six degradation scenarios are simulated over an 8000 s horizon, yielding a Composite Health Index (CHI) and 480 006 labeled samples for leave-one-simulation-out (LOSO) cross-validation. Results show a friction-torque amplification of 5.72× (RMS) and 6.86× (peak) between healthy and severely degraded wheels, directly quantifying the attitude-stabilization penalty under a fixed-gain PD controller. Using fifteen on-board-observable signals (wheel speeds, motor currents, power draw, body angular rates, and bearing temperatures), a Random Forest classifier achieves 99.46% LOSO accuracy against a 76.96% motor-current threshold baseline, while a Gradient Boosting regressor delivers R^2=0.9909 for real-time CHI estimation. All results are obtained from simulation; validation against flight telemetry or hardware-in-the-loop data remains a subject for future work.