Online Recognition Of Material Parameters In Straightening Process

Press straightening is a technology which erases or reduces bending of long steel parts. It is simple, flexible and economic. At present, the researches on straightening technology theory have developed better, especially in shafts, but the intelligent controlling study on automatic straightening technology is insufficient. In order to meet the needs of automatic straightening machine developing, it is necessary to do the straightening machine intellectualization research. This article acts as a following research based on Li Jun's model of press straightening process. After analysis of the model, in order to reduce or eliminate the influence caused by the material parameter's fluctuation, material parameters online recognition is studied. At the same time, it can expand the straightening object scope particular in the indefinite material parameters and the work-pieces which have the stress strengthening in the using process. It also combines the research results of computation model with artificial intelligence technology. The main work is as follows:In order to get the solution of the model, the numerical analysis method is used. To obtain the parameters needed for computation, the interface is programmed in DELPHI. The main straightening computation programs are completed by MATLAB.The model can be applied in parameter computation and control domain, the controlling model is also constructed in the article. There is a problem about how to obtain the material parameters accurately. The method of online recognition is used to solve this problem. The Yong's modulus and the yield stress are the recognition objects, the intrinsic and the external factors which affected these two parameters are also analyzed in the article.BP and the RBF neural network and the support vector machine are used to recognize the parameters in the straightening process, the article study the neural network and SVM on how to decide the structure, the algorithm choice as well as the essential parameter optimization and compare the good and bad points of the three methods.Through the stretching experiment and the straightening experiment, it is proved that all of three recognition methods all can solve material parameter undulation problems, RBF neural network recognition's precision is highest and structure determination is also easiest. The experiment also proves that the on-line recognition is suitable for solving material parameters undulation problem.