A Supervised Machine Learning Model for Tool Condition Monitoring in Smart Manufacturing

In the current industry 4.0 scenario, good quality cutting tools result in a good surface finish, minimum vibrations, low power consumption, and reduction of machining time. Monitoring tool wear plays a crucial role in manufacturing quality components. In addition to tool monitoring, wear prediction assists the manufacturing systems in making tool-changing decisions. This paper introduces an industrial use case supervised machine learning model to predict the turning tool wear. Cutting forces, the surface roughness of a specimen, and flank wear of tool insert are measured for corresponding spindle speed, feed rate, and depth of cut. Those turning test datasets are applied in machine learning for tool wear predictions. The test was conducted using SNMG TiN Coated Silicon Carbide tool insert in turning of EN8 steel specimen. The dataset of cutting forces, surface finish, and flank wear is extracted from 200 turning tests with varied spindle speed, feed rate, and depth of cut. Random forest regression, Support vector regression, K Nearest Neighbour regression machine learning algorithms are used to predict the tool wear. R squared, the technique shows the random forest machine learning model predicts the tool wear of 91.82% of accuracy validated with the experimental trials. The experimental results exhibit flank wear is mainly influenced by the feed rate followed by the spindle speed and depth of cut. The reduction of flank wear with a lower feed rate can be achieved with a good surface finish of the workpiece. The proposed model may be helpful in tool wear prediction and making tool-changing decisions, which leads to achieving good quality machined components. Moreover, the machine learning model is adaptable for industry 4.0 and cloud environments for intelligent manufacturing systems.