Application of a novel improved Manhattan distance on bearing fault diagnosis

Abstract

Rolling bearing is a common rotating machine in industry. Once it is damaged, the industrial machinery associated with it will be affected, and if it is serious, it may threaten life safety. Thus, an effective fault diagnosis method can reduce the occurrence of accidents. In the light of the principle of Manhattan distance and symmetrized dot pattern (SDP), the Manhattan distance is improved and a new variable is obtained. It is used as characteristic parameter to diagnose fault type. Firstly, sample data of rolling bearing under various working conditions is extracted, and it’s split into 10 pieces of data. Then, through the SDP principle, the equal part of the data is became a picture, and the symmetrical image in polar coordinate system is obtained. After binarization of the SDP image, the local area of the binarized SDP image is selected, and the mean value matrix of the local matrix is computed, and salt and pepper denoising is carried out for each local matrix and mean value matrix, and the maximum characteristic value of average value array after salt and pepper denoising is computed. Finally, local images are corresponded with their average images one by one, and the Manhattan distance between them is calculated. According to the feature method, the improved new value is obtained by linear transformation, which is called the improved Manhattan distance, and then the average value is obtained. Through the experimental data to verify whether the method can effectively distinguish the fault type.