Fault diagnosis of rotating machinery with high-dimensional imbalance samples based on wavelet random forest

Abstract

Rotary machinery is the key equipment in industrial production, and its running state directly affects production safety and efficiency. However, in practical applications, rotating machinery often faces the problem of imbalanced data categories, which not only reduces the accuracy of fault diagnosis but also affects the generalization ability of the model. To solve this problem, this paper proposes a wavelet packet transform (WPT) and random forest (RF) combination model called Wavelet Random Forest. Specifically, WPT is used to extract the time domain and frequency domain features of the signal, to reduce the complexity of the signal and enhance its expression ability. Then, the extracted features are classified by RF to improve the efficiency of fault diagnosis and classification performance effectively. The experimental results on two unbalanced datasets show that the proposed method is superior to the traditional methods in fault diagnosis tasks, and has a better classification effect, especially in the identification of a few classes of faults.