Use of least square support vector machine in surface roughness prediction model

Abstract

This paper aims to introduce a novel model into prediction field for surface roughness in machining process, and report the results of comparison between the novel model and the other two prediction models in the experiments that have been examined. The novel model is based on least square support vector machine (LS-SVM), while the other two models are based on BP neural network and standard support vector machine (SVM) respectively. In the study, 54 groups of data about surface roughness and four kinds of parameters were obtained by full factorial experiments. And then, the data were analyzed by contrast experiments: set up prediction models with BP neural networks, standard SVM and LS-SVM respectively. The results have indicated that the mean deviation of LS-SVM model is only about 25% of that of SVM method, and 2~3 orders smaller than that of BP method. Furthermore, it takes the least time to set up the models by LS-SVM model among these approaches. In summing up it may be stated that the proposed model is faster in speed, higher in accuracy, and more suitable for prediction of surface roughness.