A Machine Learning Scheme for Tool Wear Monitoring and Replacement in IoT-Enabled Smart Manufacturing

Tool wear monitoring is an important task in a smart manufacturing industry. Detecting worn-out tools and replacing them in time can increase the efficiency significantly. Various sensors are being used in machine tools to integrate them into a smart manufacturing setup. Continuously decreasing the cost of the sensors is encouraging the use of low-cost indirect methods for the task. Using multiple sensors increases the precision of estimating tool health over the single sensor-based approach. Appropriate mathematical models relating tool wear parameters and sensors data can be used here, but machine learning models become more suitable in a large variety of applications over normal mathematical models. This paper proposes a methodology for multi-sensor-based indirect tool wear monitoring system and presents a comparison of accuracy among various machine learning models. Standard references are used to generate dummy training and testing data. Python is used to create and test the models. In the end, it has been found that Naïve Bayes and support vector machine algorithms are yielding up to 97% accuracy. This is the initial work in the development of an IoT enabled and fully automated manufacturing setup.