Research And Application Of Contour Error Calculation And Compensation Method Based On Feed System Response Model

Contour error is definitely produced when the machine tool processes curves and surfaces,which seriously affects the processing quality of the workpiece.The prediction and compensation of contour errors is a crucial technique to improve the contour performance of CNC machine tools to achieve higher machining accuracy of parts.The research objective of this subject is to design an offline contour error compensation algorithm that modifies on the original machining instruction trajectory to reduce the contour error of the processing curve and improve the contour accuracy.A necessary condition for the offline compensation algorithm is to obtain a simulation model of the feed system.Based on the machining data of the machine tool,the reference position and the actual machining position(table position),this topic establishes a contour error prediction model based on the nonlinear autoregressive long-short-term memory network(NAR-LSTM).The model simulates the mapping between the input command data and the output position of the table.The prediction error meets the requirements of the compensation algorithm for the model.In the process of solving the compensation value,because the change of the previous compensation value will affect the conversion of the latter compensation value,and the large number of compensation values also makes direct conversion very difficult,the problem is converted into an alternative compensation parameter.According to the consistency and fast advantages of the structure of the deep reinforcement learning algorithm and the compensation value optimization algorithm,a deep Q network technology combined with the feed system response model,a search strategy based on optimization is proposed,which is named DQN-CVC algorithm.Combining with the research object of this subject,we have optimized the algorithm in detail.The algorithm can quickly find the optimal compensation parameters by inputting the initial compensation parameters defined by loop iteration training.And,for different processing curves,the algorithm can solve the corresponding compensation parameters to achieve compensation.The next problem to be solved is the application of the compensation value in the CNC process.Through the S-G smoothing processing of the compensation data,the compression of the data amount based on the linear difference,and the development and design of the internal interface of the CNC system,the compensation during the CNC machine tool processing is realized.In order to verify the proposed method,a set of experiments were carried out to prove the feasibility of contour error modeling and pre-compensation methods.