Rotating machine fault diagnosis method based on multi-sensor information fusion and deep learning

Abstract

The invention discloses a rotating machinery fault diagnosis method based on multi-sensor information fusion and deep learning, and the method comprises the steps: obtaining data of a plurality of sensors, dividing the data into three groups, and carrying out the sample segmentation to form a sample set; extracting time domain and frequency domain features of the sample data; then constructing R, G and B three-channel features; a Fisher score method is adopted to obtain scores of each feature for various fault categories, sorting is carried out according to the scores, and the first 10-20 features of sorting are selected and constructed into an R, G and B three-channel feature matrix; and taking the feature matrix as three channels of an RGB image, enabling each sample to generate a two-dimensional RGB image, inputting the obtained RGB images into a deep learning model for training and fault diagnosis, and finally obtaining a diagnosis result of the state of the rotating machine. According to the invention, related data acquisition is carried out by adopting multiple distributed sensors, so that the mechanical state detection is more comprehensive, and the anti-interference performance of the fault diagnosis model is greatly improved.

Images