The application research of nuclear power mechanical equipment coupling-fault diagnosis technologies based on BP neural network and optimization algorithm

An intelligent coupled-fault diagnosis approach, based on the BP neural-network combined with optimization algorithms, is proposed to address challenges such as low data utilization, limited diagnostic accuracy, and the complexity of decoupling in manual failure diagnosis of nuclear power mechanical device. Utilizing device condition monitoring data and failure information, the diagnostic performance and error rates of three intelligent fault diagnosis algorithms are comparatively evaluated on typical nuclear power mechanical systems. The comparison results indicate that both the genetic algorithm and particle swarm optimization algorithm markedly enhance the diagnostic accuracy of the BP neural-network. In addition, a decoupling process is introduced for the analysis of coupled faults in nuclear power mechanical device, integrating the BP neural-network with optimization algorithms. The effectiveness and applicability of the proposed diagnostic method in identifying and analyzing coupled fault scenarios are verified, offering a novel and practical solution for complex failure diagnosis within this domain.