Exploring Wavelet Transformations for Deep Learning-based Machine Condition Diagnosis

TL;DR

This work addresses motor-condition diagnosis by transforming time-series motor current signals into 2D time-frequency representations via Wavelet Transform (WT) and feeding them to CNNs. The WT is defined as $WT{x(t)}(a,b) = \frac{1}{\sqrt{a}} \int_{-\infty}^{\infty} x(t) \psi\left(\frac{t-b}{a}\right) dt$. Five WT variants (Amor, Bump, Morse, and two synchrosqueezed WSST variants) are evaluated with identical CNN architectures and 10-fold cross-validation on a dataset of 3,750 samples across five motor conditions. WT-Morse achieves the highest accuracy at 93.73%, slightly surpassing the prior best 93.20%, while the synchrosqueezed methods underperform and show instability, supporting wavelet-based 2D representations as a robust, cost-effective alternative to vibration-based monitoring.

Abstract

Deep learning (DL) strategies have recently been utilized to diagnose motor faults by simply analyzing motor phase current signals, offering a less costly and non-intrusive alternative to vibration sensors. This research transforms these time-series current signals into time-frequency 2D representations via Wavelet Transform (WT). The dataset for motor current signals includes 3,750 data points across five categories: one representing normal conditions and four representing artificially induced faults, each under five different load conditions: 0, 25, 50, 75, and 100%. The study employs five WT-based techniques: WT-Amor, WT-Bump, WT-Morse, WSST-Amor, and WSST-Bump. Subsequently, five DL models adopting prior Convolutional Neural Network (CNN) architecture were developed and tested using the transformed 2D plots from each method. The DL models for WT-Amor, WT-Bump, and WT-Morse showed remarkable effectiveness with peak model accuracy of 90.93, 89.20, and 93.73%, respectively, surpassing previous 2D-image-based methods that recorded accuracy of 80.25, 74.80, and 82.80% respectively using the identical dataset and validation protocol. Notably, the WT-Morse approach slightly exceeded the formerly highest ML technique, achieving a 93.20% accuracy. However, the two WSST methods that utilized synchrosqueezing techniques faced difficulty accurately classifying motor faults. The performance of Wavelet-based deep learning methods offers a compelling alternative for machine condition monitoring.

Figures (6)

• Figure 1: Sample plots showing the five Wavelet transform variants with 32x32 image resolution of the motor dataset across five condition classes under the full-load condition (load=100%).
• Figure 2: The Convolutional Neural Network (CNN) architecture of FOPCNN with the following input layer corresponding to the 32x32 RGB images.
• Figure 3: The training and validation classification accuracy performances of the five WT models
• Figure 4: The training and validation loss function graphs of the five WT models.
• Figure 5: A box plot of the five WT models performances under the 10-fold stratified cross validation step