Application of metaheuristic optimization based support vector machine for milling cutter health monitoring

With the arrival of Industry 4.0, intelligent condition-based maintenance has become a must, if not a need, for industries with significant capital investments in rotating machineries. Tool Condition Monitoring (TCM) is one of the strategic research domains in condition-based maintenance. Lately, supervised algorithms based on Machine Learning (ML) techniques assist classification of the cutting tool's condition in operation. One such algorithm is the Support Vector Machine (SVM) popularly used for training the data however, choosing optimal hyper-parameters for an SVM is essential in making the model robust. Owing to intermittent cutting in a milling operation, the modeling of tool conditions based on vibrations evolved during machining needs to be handled wisely. Consequently, there exists a need for meta-heuristic optimization algorithms to drive SVM for evaluating the robustness of the model and to increase accuracy, thereby minimizing the risk of false classification of tool bits. Over the past decade, meta-heuristic algorithms have found immense use in optimizing ML models and solving real-life engineering problems. This research paper aims to optimize hyperparameters of SVM - 'C' and 'gamma' using metaheuristic algorithms in the context of TCM. Further, the paper evaluates popular metaheuristic algorithms. It compares their respective efficacies, enabling researchers in the field of TCM to choose the appropriate algorithm for their optimization problem statement to get higher performance predictions from their SVM models.