Tool Condition Monitoring Using Artificial Intelligence in Machining: An Application Directed towards Industry 4.0

Introduction: In response to the growing demand for machining process automation, both academic and industry experts have prioritized real-time tool condition monitoring. However, this approach has been difficult previously owing to several known and unknown machining parameters, as well as difficult working circumstances, such as high temperatures and pressure. Inprocess condition monitoring and real-time process optimization have become easier because of developments in sensor and digital technologies. These processes are now very precise, reliable, and efficient. Method: This paper presents a comprehensive solution for monitoring machining systems, introducing an Artificial Intelligence (AI) and Machine Learning (ML) framework that leverages integrated sensory data to detect and predict flaws and wear issues in cutting tools. In the present work, the time series data from CNC motors (in the X, Y, and Z axes and the spindle) recorded during 18 machining tests at 100 ms sampling rate were used for training and testing the system. The experimental data reported the tool condition, whether worn or unworn. Results: Data pre-processing involved importing essential libraries, converting non-numerical labels to numerical ones, and removing irrelevant columns. With feed rate and clamp pressure specified, a correlation heatmap indicated important features for tool condition prediction. Two classification algorithms have been implemented, i.e., decision tree and logistic regression models. Further, the adequacy of the models was compared based on tool condition prediction. Conclusion: Logistic regression demonstrated superior predictive performance compared to decision tree, establishing it as the preferred model for this analysis. Model performance was evaluated through training and testing with a 70/30 dataset split. Logistic regression effectively predicted tool condition as either worn or unworn based on input parameters of feed rate and clamp pressure.