Abstract:Traditional fault diagnosis methods rely on prior data and models, which have limitations.To solve this problem, a datadriven fault diagnosis method for rotating machinery was proposed. The empirical mode decomposition (EMD) algorithm was used to decompose the original fault signal, and the limited IMF components were obtained. The optimal cutoff threshold was obtained by optimizing the existing EMD algorithm, and the system noise interference was effectively separated. The time domain and frequency domain features of the fault signals were extracted from the multi-domain quantization perspective, and the feature classification and identification of the fault points in the fault signal of denoising rotating machinery were realized based on the EMD sample entropy.The simulation results show that the proposed datadriven algorithm can be used to accurately identify the weak features of fault signals under different load conditions, and it has higher training accuracy and diagnostic accuracy