Prediction of Surface Roughness Based on Least Square Support Vector Machine in Low-Frequency Vibration Cutting Process

Abstract

A prediction method based on least square support vector machine is introduced into the surface roughness prediction model in low-frequency vibration cutting. The model is created with low-frequency vibration cutting experiment for the corresponding relationship between vibration parameters and cutting parameters and the workpiece surface roughness. The training sample set is constructed to train regression models of least square support vector machine through experimental data. Identification of training sample set is done to gain the regression parametersaandb. The amplitude ofA, vibration frequencyf, feedf1and spindle speednare used as the input variable in Xi. Predicted values of surface roughness are forecasted with the model. Evaluation is made with the difference between the predicted value and experiment. Comparison with BP neural network and support vector machine method has shown that the least square support vector machine prediction model works faster than SVM method, the prediction error is about 29% of that by support vector machine, and the prediction accuracy is higher than the BP model.