Research on Fault Diagnosis Method of Solid-State Power Controller of Rocket

Considering the scarcity of test nodes in solid-state power controller (SSPC) circuits, the challenges in acquiring fault data, and the frequent misdiagnoses and omissions by traditional algorithms, this study presents an SSPC circuit fault diagnosis method underpinned by a particle swarm optimized extreme learning machine (PSO-ELM) algorithm. This enhanced algorithm markedly boosts the accuracy and efficiency of fault diagnosis. Through the simulation of the electrical-thermal coupling effect, the actual operational conditions of the SSPC circuit are accurately mirrored. Feature values extracted from the temperature data during the circuit's power-up to stable operation process serve as the diagnostic model's input parameters. A PSO-ELM model is then established to optimize the input weights and hidden layer bias, thereby refining and enhancing the model, while elevating the diagnostic accuracy of the SSPC circuit. Experimental results affirm that the enhanced algorithm effectively improves fault diagnosis accuracy.