Intelligent Fault Diagnosis Method for Gear Transmission Systems Based on Improved Multi-Scale Reverse Dispersion Entropy and Swarm Decomposition

Based on the non-stationary and non-linear acceleration signals, a rapid data-driven method for fault diagnosis in gear transmission systems, which is based on swarm decomposition (SWD) algorithm, improved multi-scale reverse dispersion entropy (improved MRDE) algorithm, and bidirectional long short-term memory (Bi-LSTM) network, is proposed. First, every segment in the original signals is decomposed into several oscillatory components (OCs) with simple fault information by the SWD algorithm. Second, the proposed improved MRDE algorithm is adopted to further extract the features of the original signal and the decomposed signals under different scale factors, and the features are combined into a next bigger feature vector. Finally, the datasets composed of feature vectors are divided into train and test datasets to train and validate the Bi-LSTM network, so as to recognize and classify different fault signals intelligently. The proposed method of fault diagnosis in this article is verified by the signals under different types of faults are collected from the wind turbine drivetrain diagnostics simulator (WTDDS). And the results of the experiment show that it can recognize and classify the types of gear transmission system’s fault diagnosis quickly and accurately, and has its advantages in stability, determination, and efficiency.