Tool wear monitoring during electric pulse assisted milling with acoustic emission signal based on IA-VMD noise reduction and GA-RFE-RBF feature selection

Regarding the problem that electroplastic effect and sliding electrical contact have an impact on acoustic emission (AE) signals during electric pulse assisted milling, making it difficult to achieve accurate online monitoring of tool wear, the influences of electroplastic effect and sliding electrical contact on the AE signals were studied, a noise reduction method based on immune algorithm (IA)-variational mode decomposition (VMD) was proposed, a feature section algorithm based on genetic algorithm (GA)-recursive feature elimination (RFE)- radial basis function (RBF) was established, and tool wear monitoring was achieved using a BP neural network model. The experiment results showed that the AE signals during the EPAM process became more complex due to the electroplastic effect and the sliding electrical contact. The noise of the AE signals was effectively reduced using the IA-VMD algorithm, the signal effectively suppressed random noise interference while retaining the key features of the pulse waveform, resulting in a signal-to-noise ratio of 13.7 dB. The GA-RFE-RBF algorithm integrated the global search ability of the GA, the local fine-grained filtering characteristics of the RFE and the non-linear mapping advantages of the RBF, the best feature value combinations were as the input of the monitoring model with combination number of 7 and 9. And the predictive accuracy of the tool wear using BP neural network reached 98.5%.