A New Hybrid LSTM-GRU Model for Fault Diagnosis of Polymer Gears Using Vibration Signals

Abstract

Polymer gears have exhibited promising potential in power transmission due to their robust mechanical properties. However, expanding their use in power transmission requires the development of a reliable fault detection technique to minimize maintenance costs. Therefore, the primary aim of this study is to design, test, and compare the six different deep-learning models for the classification of the multiclass fault of polymer gears with minimum computational time. The proposed approach involves complete ensemble empirical mode decomposition with the adaptive noise (CEEMDAN) technique for getting an enhanced signal. Features extracted from the enhanced signal are fed to various design models for fault diagnosis. The result demonstrates that a maximum of 99.6% accuracy with 99.89% kappa and 99.6% F1-score could be achieved by hybrid long short-term memory and gated recurrent unit (LSTM-GRU) model with minimum computational time. The findings suggest that the proposed method has the potential to serve as an effective alternative for precise fault detection of gears. © 2023, Krishtel eMaging Solutions Private Limited.