Gaussian Process Regression and Linear Degradation-Based RUL Prediction of DC-Link Capacitors in Electric Aircraft Powertrains

In modern electric aircraft, DC-link capacitors are critical for voltage regulation in power converters, directly affecting the reliability and performance of onboard electrical systems. Accurate prediction of capacitor degradation is essential for proactive maintenance and fault prevention. This paper investigates the remaining useful life (RUL) prediction of 270 V DC-link capacitors in electric aircraft powertrains using two distinct methods, linear degradation modeling (LDM) and gaussian process regression (GPR). Diagnostic features are extracted from simulated voltage waveforms representing extended flight operations, and a health indicator (HI) is constructed to quantify capacitor aging. The LDM provides incremental RUL estimation with confidence intervals, while GPR captures nonlinear degradation patterns and associated uncertainties, achieving higher prediction performance compared to LDM. Real-time validation is conducted using a dSPACE 1202 processor under different flight phases, coordinated by an energy management centre (EMC) compliant with MIL-STD-704F standards. Comparative results demonstrate the superior performance of GPR, enabling informed decisions for intelligent energy management and maintenance in aircraft power systems.