A novel feature reconstruction method for compound fault diagnosis based on vibration signal information

Compared with single fault, compound fault is more common, more dangerous, and more difficult to diagnose in practical engineering applications. Compound fault in mechanical equipment is the result of multiple faults occurring at the same time and coupling with each other. This can lead to interference, loss, or overlap of fault features, making fault identification difficult. In this paper, the time domain and frequency domain features of samples are extracted, and the importance of different features is analyzed by calculating the information gain of samples. The main characteristics of single fault and compound fault are extracted. Multiple machine learning algorithms are used to construct fault diagnosis models for comparative analysis. The extracted features of a single fault are weighted and combined, and the features of compound fault are reconstructed. And the subclass weights are updated by the optimization algorithm. By analyzing the weight of the subclasses of compound faults, the specific fault class of compound faults are determined. The results show that the feature reconstruction method can effectively identify the specific class of compound faults.