PMDC Motor Brush Wear Condition Diagnosis using Dual Vibration Sensors and Windowed Feature-based XGBoost

Key Points

• The study aims to diagnose PMDC motor brush wear using data from dual vibration sensors and machine learning techniques.
• Installed dual vibration sensors on motor and gearbox to collect synchronized signals.
• Segmented continuous signals into 0.1 s windows with 50% overlap.
• Extracted 34 features from time, frequency, and envelope domains for model training.
• Applied XGBoost for both regression and classification tasks.
• Regression model achieved MAE of 0.1245 mm, RMSE of 0.2760 mm, and R² of 0.9846.
• Classification model reached an accuracy of 98.48% and F1-macro of 98.41%.
• Effectiveness of the proposed method was validated through comparisons with baseline models and configuration setups.

Abstract