关键词: 表面粗糙度, 成形磨削, 预测, 最小二乘支持矢量机

Abstract: A new prediction method based on least squares support vector machine is put forward for surface roughness of profile grinding after comparing and analyzing the common prediction methods. For one thing, it can solve the small sample learning problem better and avoid such disadvantages as over-learning and weak generalization ability that the artificial neural network prediction has, because support vector machine recognition model is based on structure risk minimization. For another thing, this method is more simple and quick in finding answer because it uses equality constraints instead of inequality constraints. The application example shows that the model is higher in accuracy and learning speed, easier to realize and so on. So it is more suitable for prediction for calorific value of coal. With the precise prediction model, a feasible plan for optimizing grinding parameters is put forward and a figure about the relation between grinding parameters and surface roughness is set up. Therefore, it has certain guiding significance to optimizing grinding parameters and improving the surface quality of finished parts.

Key words: Least squares support vector machine, Prediction, Profile grinding, Surface roughness