Research On The Neural Network Optimization Module In Sheet Metal Forming CAD/CAE Integrate System

The drawing of the stamping workpieces is a complex and dynamic process. The workpieces are often cracked or winkled and even become useless during the actual production. At present, the analyses and optimization of the defects absolutely depend on the knowledge and experience of the technocrats. Therefore, we can't get the satisfying result on the average. In order to solve the problem above, we can use a numerical finite element approach during the CAE stage to simulate the sheet forming process after the CAD stage. On this ground, the potential sources of defects and failures can be forecasted in the early time. The method introduced above is what we called Integration of CAD/CAE. Although the development and application of the numerical simulation technology have gotten an active effect over the past decades. Several problems still exist during the actual application. For examples: the errors between the numerical simulation software and the physical model; the deflection of the software because of the difference between the simulation model and the really production environment; the intelligent transmission of the data between CAD and the optimizing stage, etc. In view of the reasons above, we use the neural network technology to develop the optimization system corresponding. The result has proved that: this system can solve the problems above well and get the optimal precision.In this problem, we can't obtain the experiment data through the equipments and conditions available in view of the drawing of the covering parts need higher manufacturing and experimentation conditions. Therefore , we choose the columnar parts which is the most typical and familiar in drawing as the experiment base. Through a large amount of physical experiments, we have gotten some useful data about the Blank Holder Force, which is one of the most important factors of the drawing process. Second, we utilized the LS-DYNA-based sheet metal forming simulation solution software Dynaform to simulate the whole drawing process of columnar parts on the base of the physical experiment. Then we compared the data of the simulation and the physical experiment to check the accuracy of the simulation. Third, we took advantage of the neural network technology and design method to construct the optimization system of BHF. Then we used the numerical software Matlab 7.0 , the specialized knowledge of sheet metal forming and the Object Oriented Design(OOD) to develop the system in detail. Through the work above, the optimization system of BHF based on the BP neural-network was been established. At last, we made some tests to prove that this system could not only be suitable for columnar parts but also for the covering parts.The research of this paper suggests that: the BHF optimization system based on the the neural-network has the high optimal precision and it is not only suitable for columnar parts but also for some complex parts; numerical software Dynaform can simulate the drawing process and it's defects rightly, so it can provide accurate training data to the optimization system. The combination of neural network technology, sheet metal forming theories and numerical simulation technology can solve the optimal problem of the stamping. It plays an important role to the realization of the sheet metal forming CAD/CAE integration and even to the application of the concurrent engineering technology in sheet metal forming. And the research on CAD/CAE integrate system of sheet metal forming in this paper is of great practical value. Moreover it can bring obvious economical and social benefit.