OIL AND WORKING CONDITION DATA-BASED POWER UNIT FAILURE DIAGNOSIS METHOD

The present invention discloses an oil and working condition data-based power unit failure diagnosis method, which includes: determining a parameter of a kernel function of a kernel principal component analysis algorithm using a dynamic inertia weight-based particle swarm optimization algorithm; performing an optimization process on an initial weight and threshold of a Back Propagation (BP) neural network model using the particle swarm optimization algorithm, and training the initialized BP neural network model using a training sample set to obtain a trained BP neural network model; determining whether monitoring data is failure data using the kernel principal component analysis algorithm with the determined parameter; and inputting the monitoring data to the trained BP neural network model to determine a failure type corresponding to the monitoring data when the monitoring data is failure data. According to the present invention, the particle swarm optimization algorithm, the kernel principal component analysis algorithm and the BP neural network model are combined to efficiently and accurately diagnose a failure of a power unit based on rich failure information in lubricating oil of the power unit.