Prediction of surface roughness in cylindrical longitudinal grinding based on evolutionary neural networks

Artificial neural networks are introduced in the area of grinding. There are some disadvantages in BP (Back Propagation) algorithm, such as low rate of convergence speed, easily falling into local minimum point and weak global search capability. In order to settle these problems, this paper presents a new learning algorithm that uses GA (Genetic Algorithm) to train BP neural networks. The prediction model of surface roughness in cylindrical longitudinal grinding based on evolutionary neural networks is proposed in detail. The experimental and the simulating results shows that the combination of BP and GA can effectively overcome the problem of easily falling into local minimum point, and this method can get higher accuracy of prediction. By motoring the grinding parameters, this method using in this paper can realize the on-line prediction for the roughness of the workpiece.