Tool wear prediction based on multisensor data fusion and machine learning

Abstract Modern manufacturing industries are revolutionized by smarter and more efficient operations, particularly the integration of sensor technology, data science, and machine learning. These technologies play a crucial role in enhancing productivity, maximizing equipment lifespan, and maintaining consistent product quality. In computer numerical control (CNC) machining, worn-out tools, associated with increased forces and vibrations, directly affect machine performance and part quality. Traditional tool condition monitoring methods involve manual offline inspection, resulting in machine downtime and reduced productivity. To overcome these limitations, automated monitoring systems based on multisensory data and machine learning techniques have been developed, allowing for continuous real-time tool wear monitoring and prediction. This study introduces a novel three-level data fusion architecture for predicting tool flank wear in CNC machining. Raw data was collected using a load cell, an accelerometer, and an acoustic emission sensor during CNC milling operations. A stacking ensemble method combines random forest (RF), artificial neural network (ANN), and extreme gradient boosting (XGBoost) to enhance predictive accuracy, obtaining an R 2 value of 0.955 and a root mean squared error (RMSE) of 58.237 µm. The proposed three-level data fusion architecture demonstrated exceptional effectiveness in predicting tool flank wear and advancing prognostics and health management (PHM) of machining equipment.