Improved performance evaluation of tool condition identification by manufacturing loss consideration

When neural networks are used to identify tool states in machining processes, the main interest is often the recognition ability. It is usually believed that a higher classification rate from pattern recognition can improve the accuracy and reliability of tool condition monitoring, thereby reducing the manufacturing loss. Nevertheless, the two objectives are not identical in most practical manufacturing systems. The aim is to address this issue and propose a new performance evaluation function so that the recognition ability of tool condition monitoring can be evaluated more reasonably. On this basis, two kinds of manufacturing loss due to misclassification are analysed: the overprediction caused by misclassifying the worn tool condition; and the underprediction caused by misclassifying the fresh tool condition. By using both to calculate corresponding weights in the performance evaluation function, the potential manufacturing loss is introduced to evaluate the recognition performance of tool condition monitoring. Based on this performance evaluation function, a modified support vector machine approach with two regularization parameters is employed to learn the information of every tool state. In this support vector machine design, the effective feature set extracted from acoustic emission signals is used as inputs, and a five-fold cross-validation is used to tune the parameters. The experimental results show that the proposed method can reliably identify tool flank wear and reduce the overdue prediction of worn tool conditions and its relative loss. Experimental results show that this approach may effectively identify tool state over a range of cutting conditions and reduce the manufacturing loss in the practical industry process.