Planetary Gears Feature Extraction and Fault Diagnosis Method Based on VMD and CNN

Abstract

Given local weak feature information, a novel feature extraction and fault diagnosis method for planetary gears based on variational mode decomposition (VMD), singular value decomposition (SVD), and convolutional neural network (CNN) is proposed. VMD was used to decompose the original vibration signal to mode components. The mode matrix was partitioned into a number of submatrices and local feature information contained in each submatrix was extracted as a singular value vector using SVD. The singular value vector matrix corresponding to the current fault state was constructed according to the location of each submatrix. Finally, by training a CNN using singular value vector matrices as inputs, planetary gear fault state identification and classification was achieved. The experimental results confirm that the proposed method can successfully extract local weak feature information and accurately identify different faults. The singular value vector matrices of different fault states have a distinct difference in element size and waveform. The VMD-based partition extraction method is better than ensemble empirical mode decomposition (EEMD), resulting in a higher CNN total recognition rate of 100% with fewer training times (14 times). Further analysis demonstrated that the method can also be applied to the degradation recognition of planetary gears. Thus, the proposed method is an effective feature extraction and fault diagnosis technique for planetary gears.

Small‐batch‐size convolutional neural network based fault diagnosis system for nuclear energy production safety with big‐data environment

“…In the fault diagnosis field, CNN is constantly being explored and researched in different mechanical and energy production system: A planetary gear fault state identification and classification method by training a 1‐D CNN using singular value vector matrices as inputs was achieved by Liu in 2018 . Two CNN‐based refrigerant charge fault detection strategies were suggested by Yong Hwan Eom in 2019 .…”

End-To-End Convolutional Neural Network Model for Gear Fault Diagnosis Based on Sound Signals

“…In the past decade, a vibration signal analysis method has become one of the most used and effective methods for gear fault pattern recognition and fault localization due to the abundant information and clear physical meaning in the mechanical vibration signal [1][2][3][4][5][6][7][8][9][10][11][12]. For example, Li et al [1] proposed a method of planetary gear fault diagnosis via feature extraction based on multi-central frequencies and vibration signal frequency spectrum.…”

“…For example, Li et al [1] proposed a method of planetary gear fault diagnosis via feature extraction based on multi-central frequencies and vibration signal frequency spectrum. Liu et al [2] proposed a feature extraction and gear fault diagnosis method based on vibrational mode decomposition, singular value decomposition, and convolutional neural network (CNN). Kuai et al [3] used the method of permutation entropy of Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) Adaptive Neuro-fuzzy Inference System (ANFIS) to make gear fault diagnosis in a planetary gearbox.…”

“…At present, deep learning models have been widely applied in speech recognition [25,26], acoustic scene classification [27,28], and environmental sound classification [29,30]. Many researchers also applied deep learning models in vibration signal analysis to detect fault modes [1,2,4,31]. However, there are few studies applying deep learning in ABD for two reasons: (1) the scarcity of publicly available acoustical signal fault diagnosis datasets like Case Western Reserve University bearing vibration signal datasets, and (2) acoustical signal acquisition requires a higher standard for environment and for equipment.…”

“…(1) Build and release the first ever publicly available sound dataset of gear fault diagnosis based on acoustical signal. (2) We designed an end-to-end stacked CNN model for an ABD task, which obtained information directly from time and frequency domain without feature engineering.…”