Research For Numerical Optimization Method On Stamping Composite Parameter Based On Artificial Intelligence

Sheet metal stamping is an important process for manufacture of sheet metal parts and is widely used in automobile, aircraft, electronic instrument, shipbuilding, apparatus and other industries. While the requirement of stamping is increasing, every country, especially developed one, attaches much importance to this field. In sheet metal forming, many factors could influence the quality of this process, amongst which springback is the most important one. There are two jobs, springback forecasting and springback controlling, have to been done in the research of springback. Springback forecasting is the base to control the phenomenon of springback. Furthermore, springback controlling is just the objective which we have to achieve, which improve the quality of stamping.As the development of computer science, it can satisfy the requirement of large scale numerical calculation which makes the stamping simulation based on FEA popular in industry and brings the stamping innovation. Moreover, with the powerful computing capacity, computer can simulate the process of stamping through the mathematical model based on the experiment. Furthermore, numerical optimization can be done in the base of simulation, which make the stamping industry has a great development.In this paper, the research is carried out around springback phenomenon in the technical background introduced before. As we known, the springback phenomenon mostly appears in the U-shaped sheet stamping. Therefore, it is the research object in this paper. Firstly, Orthogonal Experiment (OE) design will be launched to build the base of the whole research. The material performance, stamping process and the springback angel will be set as experiment genes. The value in the range of real project will be set as experiment levels. The tolerance between simulation result and original model will be set as experiment target. Secondly, according to the Orthogonal Experiment form, stamping forming and springback simulation will be carried out in DYNAFORM. Besides, the springback angel and the tolerance could be gained through the springback evaluation module to finish the Orthogonal Experiment form. Thirdly, this form can be imported into the Springback Control System as the database to train the Artificial Neural Network (ANN) which could simulate the process of stamping. In the meanwhile, this database also can be used as the checking data to make sure the tolerance of ANN can meet the requirement in industry application. Then, the ANN is used as a Fit Function in the Genetic Algorithm (GA), with which can give the most appropriate parameter.The characters of this paper consist in the following several factors:1) Implementing the transforming from geometry to parameter through the springback measuring which sets the material performance, stamping process and the springback angel as optimization object.2) Implementing the stamping simulation through mathematical model and gaining the optimized result with ANN and GA.3) Succeeding to develop a application valuable software platform which integrate OE, stamping simulation, springback evaluating and optimization.