Prediction of Gear Tooth Fatigue Life Using Vibration and Acoustic Wave Data Based on Deep Learning

In this study, the fatigue life of gear teeth was predicted using vibration and acoustic wave data based on deep learning. The dataset included vibration amplitudes and natural frequencies of gears under various loading conditions and damage states. To predict fatigue life, advanced machine learning algorithms such as linear regression, convolutional neural networks (CNN), and a hybrid CNN-LSTM model were employed. The results indicated that both vibration amplitude and natural frequencies significantly affect fatigue life, with every ۰.۰۱ g increase in vibration amplitude leading to a reduction of ۳ hours in gear life. Furthermore, the hybrid CNN-LSTM model demonstrated the best performance in fatigue life prediction, achieving an accuracy of ۹۴.۲% and an AUC-ROC of ۰.۹۸. Time-series and correlation analyses also confirmed that vibration amplitude and natural frequency are among the most critical parameters in gear failure simulation. Overall, this study demonstrates that utilizing vibration data and deep learning methods can effectively predict fatigue life and enhance gear performance in mechanical systems.

Keywords:

Gear fatigue , Fatigue life prediction , Vibration data , Acoustic waves , Deep learning , Convolutional neural network , CNN-LSTM model , Correlation analysis , Time-series analysis , Dynamic modeling