Fault diagnosis method of rotating machinery based on improved deep forest model

Abstract

Here, to solve the problem of deep neural network being limited by hyperparameters and data volume, an improved deep forest model was proposed to realize efficient fault diagnosis of rotating machinery. Firstly, multi-granularity scanning link was used to extract features of initial input data, and obtain probability features. Then,a stacking layer was added to the place cascaded with multi-granularity scanning layer to do corresponding feature extraction of input data. Finally, the data processed by multi-granularity scanning and stacking layer were input into a cascade forest to obtain classification results. The experimental results showed that the fault diagnosis accuracy of the improved deep forest model is 99. 59% and 98. 05%, and it is better than commonly used fault diagnosis models.