Vibration Based Honeycomb Core Milling Diagnostics Using Machine Learning Approaches

In industrial machining processes, tool failures can result in a loss of surface and dimensional accuracy of the finished part, or even damage to both the workpiece and the machine. Consequently, tool condition monitoring has become essential for achieving high-quality machining and cost-effective production. Moreover, cutting tool degradation can vary considerably under different operational conditions and material behaviours. Therefore, identifying the tool state during machining is critical. In this study vibration responses during the milling of composite sandwich structures based on honeycomb cores are measured using a piezoelectric accelerometer. Numerous milling experiments were conducted with different machine settings and tool states. Features are then calculated in both the time and frequency domains for each time sliding window. It may be necessary to perform dimensionality reduction on the features table to avoid overfitting and reduce the computational time of the learning algorithms. The first part of this study focuses on the analysis of vibration behaviour during milling with different tool states in order to identify the best features. The second part centers on building a diagnostic model for tool states, based on supervised machine learning techniques (i.e. model training requires labelled data). The accuracy of the models is then evaluated using test data and the results are compared.