Research on tool wear classification of milling 508III steel based on chip spectrum feature

In the milling process of high-strength steel, the tool and the chip produce severe contact and friction, resulting in severe tool failure. Tools need to be replaced in a timely manner to reduce the impact on workpiece quality and processing efficiency caused by severe wear. Therefore, tool wear condition monitoring is a key factor in improving processing efficiency, ensuring processing accuracy, and achieving intelligent manufacturing. In actual machining production, the color and shape of the chip are often relied upon to judge the time to change the tool. In this paper, chip spectrum is proposed as an important feature to identify tool wear condition. What’s more, a new support vector machine model optimized by wild horse optimizer (WHO-SVM model) was constructed to identify and classify tool wear conditions. Firstly, the chip spectrum features of different tool wear stages were extracted by the near-infrared spectrum acquisition system. Through the analysis, the chip spectrum can characterize the difference in chip color under different wear conditions. The mapping relationship between chip spectrum and tool wear condition was established. Secondly, the chip spectrum was preprocessed by standard normal variable transformation (SNV) and multivariate scattering correction (MSC). What’s more, principal component analysis was used to reduce the dimensionality of the spectral data to extract 7-dimensional chip spectrum feature most correlated with tool wear amount. Finally, the WHO-SVM model was constructed, and the extracted chip spectrum was used as sensitive features to identify and classify the tool wear condition. The accuracy rate of the test set reached 90.3%. The experimental results confirm the feasibility of using the WHO-SVM model and monitoring tool wear based on chip spectrum feature. Compared with PSO-SVM, CS-SVM, and SSA-SVM models, it is concluded that WHO-SVM has a better training effect and higher prediction accuracy.