Abstract:Improved artificial neural network was applied to the prediction model of parallel machine tool roughness, which effectively predicted the influence on roughness by change process parameters such as machine tool feed rate, spindle speed, processing angle, machining force and number of times of machining, and etc. Results show that the steps when the training of the network controlled from 200 to 400, the training sample mean square error of the entire network model is stable and convergent, the prediction error of added inspection training samples can be controlled under 5%, which satisfies the requirement of the prediction model in training. The improved neural network prediction model is proved to be feasible with higher precision when used to forecast the actual machining process.