基于EWT-NPDLPP-LSSVM的水泵机组关键部件故障诊断方法OA

To improve the efficiency and accuracy of fault diagnosis of key components in pump units,considering the operational environment of the pump system,a comprehensive diagnostic method in-tegrating signal denoising,feature extraction,dimensionality reduction,and fault identification was proposed.Firstly,empirical wavelet transform(EWT)was employed to denoise the original signals,effectively mitigating the influence of environmental noise and improving data quality.Secondly,to comprehensively characterize the operational state of the pump unit,a multi-source fusion feature ex-traction method was designed,incorporating multi-channel signals(including vibration,pressure pul-sation,electrical,and other signals)and multi-domain features(time domain,frequency domain,and time-frequency domain),based on the specific operating characteristics of the pump system.On this basis,an improved local preserving projections(LPP)method,termed nearby probability distance(NPP),was proposed to eliminate redundant information from the high-dimensional features.Further,least squares support vector machine(LSSVM)was applied to classify different fault types.The experi-mental results demonstrate that the proposed EWT-NPDLPP-LSSVM-based diagnostic method achieves a high diagnostic accuracy of 99.44％and superior computational efficiency,which confirms the validity and engineering practical applicability in scenarios.