Analysis of Spur Gear Vibration Data on Backlash and Prediction of Backlash Based on Deep Learning

Gears are critical in mechanical systems, but wear-induced vibrations pose challenges. It is difficult to experimentally verify gear wear through vibrations. This study proposes a method to simulate wear by manipulating the backlash size and enabling the analysis of vibration changes. Gears with varying backlash sizes, resembling worn gears, facilitate the effective analysis of vibration signals. Collected vibration data is used with AI models to predict the gear wear degree. A deep learning model classifies backlash amounts, whereas a regression model predicts specific backlash sizes. Accurate classification and prediction enhance gear wear monitoring. This datadriven approach and AI techniques contribute to gear wear analysis, improving the monitoring accuracy.