Development of tool condition monitoring system in end milling process using wavelet features and Hoelder's exponent with machine learning algorithms

An effort was made to monitor the flank wear using wavelet analysis by extracting the Hoelder's exponent as a feature and using various machine learning algorithms to forecast the tool condition. The test was conducted on a Tungsten carbide insert with selected cutting parameters and the acquired vibration signals were used to develop the prediction model. The wavelet coefficients, Hoelder's exponent, and statistical features were extracted from the vibration signals. These features were used in machine learning algorithms like SVM, KNN, Kernel Bayes, Multilayer perceptron, and Decision trees to forecast the flank wear. The accuracy of the machining algorithm was analyzed through the confusion matrix and accuracy. The results revealed that HE along with wavelet coefficients performed better than statistical features. From the analysis, it was found that DT and SVM had the highest accuracy of 100% and 99.86% respectively. The performance of the selected ML was verified with benchmarking datasets and proves its accuracy.