Intelligent wear localization based on friction acoustic signals and machine learning

Wear localization is critical for assessing artificial joint longevity and failure. Traditional detection methods are invasive and unsuitable for real-time monitoring. This study proposes a non-invasive method for identifying wear locations based on friction acoustic signals generated by different surface textures (circular, square, hexagonal, triangular). Signals are collected using a smartphone sensor and analyzed through time-frequency domain features. Four machine learning models (Random Forest, XGBoost, a hybrid model (RF+XGBoost), and a stacked model with LightGBM) are used for classification. The results indicate that the strong correlation between surface texture and friction acoustic signal features significantly enhances the robustness of the model, and the classification accuracy can reach up to 92.38 %. The proposed method combines the advantages of acoustics, tribology, and artificial intelligence to provide a wearable solution for online health monitoring of artificial joint implants to provide personalized exercise guidance for patients, demonstrating great potential for clinical applications.