Error Prediction And Pre Compensation Of CNC System

Numerical control technology is the key technology to achieve single-processing,integrated technology in industry,has an important impact on national economy.With the development of society,CNC processing scene is gradually changing from batch manufacturing to small batch manufacturing.However,the traditional way of CNC processing has the weakness of low processing efficiency in small batch manufacturing.This paper combines the characteristics of numerical control technology,proposes an error prediction and pre-compensation algorithm to improve the processing quality.In view of the situation which machined data can not be used to improve tracking performance in multi-species small batch processing like single batch mass processing,a method of tracking error and error compensation is proposed to improve tracking performance before system processing.Through the analysis of the typical motion control system,a single-axis linear model of the CNC servo system is established.In view of the study of tracking error by the simplified servo system model combining the relevant knowledge of control theory,the influence factor of following error is determined by the combination of theory and experiment.On this basis of simplified model,a method for predicting following error by neural network is proposed based on the influence factors of tracking error.Trajectory generation module is established,the module achieves the change of track bending degree by changing the parameters,and the equivalent distance output is composed of geometric trajectory at the equidistant point.Set the qualification to ensure that the generation trajectory is connected at the beginning and end and the starting point is consistent,and the amount of data is expanded by overlay.The geometric trajectory with sufficient data volume is obtained by the existing interpolation module,and the reference input is used as the sample set.After the sample set passes through the data processing module,the position,velocity and acceleration information is obtained as the input of the neural network,and the tracking error is followed as the output value corresponding to the input.Using input and output data,NARX neural network training is used to obtain a tracking error prediction model,and the prediction ability of the model is verified by generalization experiments.On this basis of trained model,based on the limitation of experimental equipment,open-loop compensation algorithm is used to predict the follow error instead of the previous follow error to compensate,andthe compensation experiment is performed on XY platform,which greatly reduces the following error.