Tool condition monitoring using sound signal in machining process

Cutting tool wear and breakage are among the biggest obstacles to fully automated machining operations. An effective tool condition monitoring (TCM) system is the best solution for setting up fully automated manufacturing systems. Although a wide variety of monitoring techniques have been developed for online detection of tool status, the need for a reliable, simple and inexpensive solution still remains. The main purpose of this study is to classify the collected audio signals belonging to different wear status and cutting parameters during the face milling operation by using machine learning techniques and to develop a TCM model to detect the tool condition. In this study, it has been determined that the models created by training the machine learning algorithms with the audio signals collected during the face milling operation can successfully detect the tool wear conditions. A decision making system based on machine learning techniques including artificial neural networks, support vector machine and convolutional neural network approaches has been developed. The success of the installed systems was compared with the obtained prediction accuracies. The developed system is trained with preprocessed sound data. Significant prediction accuracy was obtained by testing the trained model. This proves that an effective TDI system can be established using audio signals and machine learning techniques. In addition, it has been evaluated whether the generated TDI system works with the same performance in different workpiece sizes. The sound signal collected from the face milling process of the workpiece at lower dimensions were tested in the generated model. The results show that the change in the size of the workpiece reduces the accuracy of the generated system, but it still has high accuracy values in the model generated using the artificial neural network. In addition, in this study, the performances of the models constructed with the audio signals obtained from two different audio signal acquisition equipment were compared. The results obtained show that the TDI system, which is created with audio signals obtained from hardware with high impedance and low sensitivity, gives worse results in workpiece size change. In conclusion, successful implementation of the proposed TDI system will enable to reduce downtime for tool changes, minimize the amount of scrap in the machining industry and make the most efficient use of cutting tools.