Research On Key Technology Of Intelligent NC Programming System For Wedge Parts Of Automobile Panel Die

The wedge part is an important component of automobile panel die,mainly using CNC machining to complete the processing.At present,enterprises use the interactive graphical programming method to program the NC machining program of the wedge parts.This method is cumbersome,error-prone,low-efficiency and relies too much on manual experience,which greatly affects the manufacturing efficiency of the mold and increases the mold delivery cycle and cost.Therefore,it is urgent to research and develop a more efficient and intelligent CNC programming system.After deeply analyzing the structure of the wedge part and its numerical control process,this paper summarizes the characteristics of wedge parts' structure and its numerical control process,proposes a concept of the core plane of the manufacturing feature.On this basis,the existing key technologies of intelligent NC programming are studied and summarized.Aiming at the problems existing in the current intelligent programming system in the manufacturing feature recognition and processing boundary extraction,a novel solution combining the machine learning method and two-dimensional image processing technology is proposed.Aiming at the problem of manufacturing feature recognition,this paper proposes two manufacturing feature recognition algorithms: manufacturing feature recognition algorithm based on B-rep key information of core plane and manufacturing feature recognition algorithm based on convolutional neural network.Experimental studies show that both algorithms achieve higher recognition accuracy on the wedge part.The first algorithm identifies the manufacturing features through a large number of inference rules,which leads to the flaws in the generality of the algorithm.For this reason,the second algorithm adopts the supervised learning method to learn the manufacturing feature recognition ability by introducing the machine learning theoretical achievements.The algorithm has made great breakthroughs in versatility while maintaining extremely high recognition accuracy.Aiming at the problem of processing boundary extraction,this paper proposes a processing boundary extraction algorithm based on core planar surface ring and a processing boundary extraction algorithm based on two-dimensional image skeleton.The first algorithm calculates the machining boundary by traversing the surface ring of the core plane of the manufacturing feature,which is mainly applied to the case where the machining boundary is formed by the edge of the manufacturing feature.The second algorithm projects the manufacturing features onto a two-dimensional plane,then obtains the pixel curve of single pixel width by the skeleton extraction algorithm,and finally projects the pixel curve into the three-dimensional space to obtain the processing boundary.The algorithm is mainly applied to the case where the processing boundary was formed by the manufacturing feature center line.Experiments show that the two algorithms can effectively improve the extraction efficiency and accuracy of the processing boundary.Based on NX/OpenAPI and TensorFlow machine learning platform,the paper developed an intelligent numerical control programming system on Siemens NX10.0.The feasibility and usability of the system were verified by the typical wedge parts of the enterprise.