Research And Implementation Of Five Axis Machine Tool Error Tracing Method Based On Deep Learning And S Test Part

This year, we are proud of our homemade second aircraft carriers put into use and big aircraft C919 completed first flight. Among them, Five-axis machine tools which have excellent performance in cutting free surface play an important role here. Five-axis machine tools have increased two rotating shafts, which greatly increase the machining performance but also increase the complexity of mechanism movement. At present, we still lack of corresponding evaluation standards for machining performance testing of Five-axis machine tools at home and abroad.Chengdu Aircraft Industry designed S test part according to the defects of foreign test parts and actual production experience. By analyzing S test part which cut by Five-axis machine tool, we can comprehensively evaluate the machining performance and dynamic characteristics of machine tools. However, it is difficult to establish the mapping function between the S test part and Five-axis machine tools, so the theory of machine tools error traceability based on S test part still needs to be perfected.Deep learning is one kind of machine learning algorithm that can directly process high-dimensional structural data such as images and establish accurate mapping functions automatically. Which achieve great success in several related fields.Based on the existing methods of error traceability and refers to the application methods of deep learning in related tasks, this paper presents an error traceability method for Five-axis machine tools based on S test part and deep learning, which has great significance for improving the theory of machine tool error traceability and improving the machining performance of Five-axis machine tool for S test part. The main contents of the paper as follows:1. Designing a simulation algorithm about the process of five-axis tools to cut S test part and establishing a positive mapping function between the errors of Five-axis machine tools and S test part profile error, including establishing space error model of Five-axis machine tools based on multi-body theory, then analyze the location corresponding relation between individual error of Five-axis machine tools and profile error of S test part by MATLAB.2. Designing a special convolutional neural network structure for the task of five-axis machine tools error traceability based on S test part and deep learning algorithm, including designing a three-path network for solve the difference between the three-dimension point cloud data of S test part and traditional two-dimension data, then operating batch normalization to solve the problem that error data has uncertain data distribution and high precision argument,which will affect the network performance,then using the compression method to optimize the structure of network.3. Designing the complete experiment scheme about how use deep learning algorithm and S test part to establish the mapping function about five-axis machine tools error traceability, including setting the training strategy and specification standard of data set; Building an online、visual client testing framework based on IPython notebook; Building and maintaining the deep learning server of Lab and providing remote login service. Finally, we carry out the experiment based on Caffe platform to verify the feasibility and correctness of the overall scheme.