Milling Tool Condition Monitoring for Difficult-to-cut Materials Based on NCAE and IGWO-SVM

Tool wear will induce a series of problems such as surface quality degradation, poor dimensional accuracy, and low machining efficiency. Tool condition monitoring (TCM) is essential for meeting the demands of high precision and productivity in automatic machining. However, due to the high complexity of the machining process, accurately identifying the tool conditions during milling of difficult-to-cut materials remains a tough work. To handle this issue, this paper proposed a novel method for monitoring tool wear conditions based on the nonnegativity-constrained autoencoder (NCAE) and an improved grey wolf optimization algorithm for optimizing the support vector machine (IGWO-SVM). Features are first extracted from multi-source heterogeneous signals utilizing time-domain analysis, frequency-domain analysis, and wavelet packet domain analysis and then fused by a deep NCAE network (DNCAE) to better match the dynamic characteristics of high-dimensional nonlinear data. Subsequently, an IGWO with nonlinearly decreasing convergence coefficient and adjusting position weight is utilized to adaptively optimize the penalty factor and the kennel parameter of SVM. Ultimately, the IGWO-SVM model is employed for tool wear state prediction. A multi-perspective comparison of the proposed method with other commonly used algorithms is carried out. The experimental results demonstrate that the proposed method is considerably promising in TCM since it yields significant improvements in modeling efficiency and prediction accuracy.