Fault identification of bearing based on multi-layer information entropy feature matrix

The rolling bearing may be damaged due to various reasons in the process of operation. Therefore, it is very important to determine the faults under various working conditions. In this case, how to make good use of the information characteristics of the time domain signal itself to improve the efficiency of fault diagnosis. Therefore, it is necessary for us to collect and analyze the original vibration signals of rolling bearing faults. And the characteristics of information itself can be measured by information entropy, in this paper, a bearing fault recognition method based on multi-layer information entropy feature matrix is proposed to extract the fault features of rolling bearing. The empirical mode decomposition(EMD) method is used to decompose the time domain signal and extract the first five IMF components. Each layer of signal is divided into eight segments, and one information entropy is calculated for each segment. Then a 5*8 information entropy eigenmatrix is established. Then, the average information entropy eigenmatrix of four conditions is calculated, and the average information entropy eigenmatrix is taken as the benchmark, the information entropy characteristic matrix of the test data is compared with the average information entropy characteristic matrix of the four working conditions to calculate the spatial distance. The working condition of the test data is the closest to which working condition, and the working condition of the test data belongs to this working condition. This method can effectively diagnose the fault of rolling bearing.