Research On The Optimization Of Variable Blank Holder Force Based On RBF Neural Network

Stamping forming was a very important manufacturing method, and was widely used in the fields of automobile and aircraft. It was a very important part of industrial manufacturing. The material flow of parts having complex body surfaces needed to be strictly controlled by adjusting the size of the blank holder force in the process of production. So that, the forming defects, such as crack, wrinkle and springback, which were caused by the large flow rate, could be reduced or even eliminated.In the development of new products, it needed to get the molds meeting the production requirements by repeatedly tryout, which often caused the entire molds scrapped due to human factors. These virtually increase the tooling cost and cycle. The numerical simulation and approximate model optimization technology were applied to sheet forming, which can not only greatly shorten the development cycle of new products, but also obtain the optimal combination of process parameters. Based on these, this paper applied numerical simulation and approximate model optimization technology to blank holding force of sheet forming. The related research was as follows:Firstly, the Gray correlation analysis method was used to carry out correlation analysis for sheet forming process parameters, which took sheet forming qualities of crack, wrinkle and springback into full consideration. The correlation degrees about sheet forming qualities were obtained. By comparing the correlations degrees, the result showed that the importance of the blank holding force control for improving the qualities of sheet forming.Secondly, in order to improve the convergence rate, while ensuring the diversity of the population, the search performance of artificial immune algorithm was improved by adding the elite cross and combining with the fitness probability and the density suppression probability. According to the basic principle of RBF neural network, the artificial immune algorithm was used to obtain the optimal centers and width parameters in the training of RBF neural network. So that, a RBF neural network approximate model based on the artificial immune algorithm was established.Finally, the box and S beam from NUMISHEET were regarded as research objects, and the variable blank holder force was regarded as design variable, and the largest thickening and thinning after forming were regarded as quality targets, and Latin hypercube was used to sample. The simulation software Dynaform was used to simulate sheet forming to obtain the training samples, and the artificial immune algorithm was regarded as training method. The RBF neural network approximate models between variable blank holding force with change over time or stroke or position and forming quality were established respectively. The artificial immune algorithm was used to optimize the models to obtain the optimal variable blank holder forces. Through the Comparison with constant blank holder forming, the results showed that the quality of forming parts could be improved by applying the variable blank holder force in sheet forming.