Tool wear prediction in turning process with very near field acoustic sensor using machine learning techniques

Tool Condition monitoring is an important factor in manufacturing industries for improving the product quality, reducing the manufacturing losses It also helpful to optimize the labour cost as well as to increase the productivity. Several monitoring methods are used to monitor the tool condition either directly or indirectly. The study focuses the application of very near field acoustic emission technique used as an indirect measuring method for monitoring the tool condition. In the proposed system, sound particle velocity signal is captured and is used to monitor the tool condition. The machine learning models like K–nearestneighbors,decisiontree,randomforest,logisticregressionandgradientboostinglearning were used as decision making in tool condition monitoring system. Among the selected model, gradient boosting has an F1 score value of 1.0, indicates the chosen model performs better for the given data set with weighted average of 0.99. The Kappa score value of the gradient boosting algorithm is 0.98,indicates that there is a good agreement between the predicted versus actual. The AUC-ROC Score value 0.9991 indicates that the gradient boosting model is performed well for thegiven data set with different machine learning algorithms for varioustool state conditions.