Graph modeling of multi-scale energy entropy for intelligent fault detection

Bearing is a key component of rotating machinery, and the fault detection of bearing is of great significance to ensure the normal operation of machinery. How to extract fault features with rich information from the time-frequency domain after signal decomposition becomes the key to fault detection. However, the undershoot and overshoot phenomena affect the result of signal decomposition, and the defined fault features do not consider the correlation between local scales after decomposition, which hinders the accuracy of fault detection. To address this issue, this study proposes an intelligent fault detection algorithm, named multi-scale energy entropy-graph model fault detection algorithm. The algorithm first proposes a local envelope mode decomposition method to obtain the multi-scale components of the signal, effectively addressing the issues of undershoot and overshoot present in traditional decomposition methods while reducing signal decomposition errors. Subsequently, a multi-scale energy entropy-graph model is constructed to capture more fault information considering the correlation between local scales. Furthermore, the graph state similarity index is defined to describe the machine state, and fault detection is realized using hypothesis testing. The experimental results were compared with some benchmark methods reported in the existing literature. The experimental results show that the detection algorithm achieves the best performance in Precision of 83 %, Recall of 100 %, and F1 score of 91 %, and the detection results are satisfactory.