Multisensory based tool wear monitoring for practical applications in milling of titanium alloy

Abstract

Titanium alloys can be found widely in aerospace applications due to their ideal mechanical properties. However, the low machinability of titanium alloys results in a rapid wear rate of cutting tools. In order to improve the productivity and reliability of manufacturing systems, it is essential to monitor tool wear condition. Considering practical applications on shop floors, a tool wear monitoring model is developed for milling of titanium alloy. A series of milling tests were conducted using different machine tools on shop floors. Multiple sensors were used to collect signals including cutting force, vibration and cutting sound. Time, frequency and time-frequency domain features are obtained from original signals captured during machining process. Then, feature dimensionality reduction is performed by using an Information-Measurement based feature selection method. Symmetrical Uncertainty is introduced to select the features which have good correlation to classes. Based on the chosen features, multiclass support vector machine (SVM) is developed to recognize the wear stage of cutting tool. 96.7% overall recognition accuracy can be get by using multiple sensors. The performances of classifiers based on a single force sensor and vibration sensor are 96.7% and 92.5%, respectively. Therefore, the proposed approach is effective for practical applications.