Research And System Development Of NC Automatic Programming Technology Based On Machine Learning

Based on processing information and knowledge,automatic programming is a technique about calculating NC toolpath automatically and outputting NC files.Nowadays,automation programming is mainly developed by integrating original functions of CAM or setting special rules for specific products.When meet a new requirement,NC programmers have to make a lot of adjustments to old programming templates,which is failed to adapt to changeable environment and processing requirement.With the development of machine learning,the method of product driven by data provides new solution for many complex difficultiesBy studying process knowledge and programming logic,this study adopts an automatic programming method based on machine learning and develop related systems,which apply programming experience from historical programming project to new requirements.This method gets rid of the limitation of specific processing object of automatic programming system while improving the ability to adapt complex processing environmentThe research work of this paper is as follows1.A basic geometric feature data framework is reconstructed from the STEP model file The approach to decompose processing model is studied based on the feature surface concave-convex and boundary relation obtained from this framework while processing order among features is planned.An improved feature recognition method is proposed based on area distribution algorithm calculated with triangular surface data collected from the STL model file.2.The theory and realization technology of intelligent mapping of feature and process method are studied.For each type of atomic feature,many process template are designed and constructed.Then,based on muti-classification BP algorithm,the feature and process mapping model,which is trained and simulated by sample sets collected from history programming project,can solve the optimization problem to choice proper process template for objective feature3.A pre-learning weight optimization method is proposed to improve parameter fitting performance of IPSO-BP algorithm.With this method,the model for cutting parameters,which is trained and simulated by processing parameters collected from history programming project,can solve the optimization problem to give recommending parameters in different situationFinally,based on the research result stated above,an automatic programming system,which is developed by the method of secondary development on Powermill,has the functions of automatic craft creation,calculation and simulation of machining tool path and generation of NC program.