Study On Injection Molding Process Multiobjective Optimization And Quality Estimation Based On RBF Neural Network

The quality of injection-molded plastic parts is affected by many factors, and an inappropriate setting of the process parameters will cause quality defects such as warpage, weld-lines and air-traps, which will impact the appearance and the performance of the parts. This study focused on the impact of processing parameters on the warpage, weld-lines and air-traps of plastic part by uniform design based on CAE simulation. The method of injection molding process optimization by RBF neural network and Genetic Algorithm is studied, and we can get the optimal process parameters that make the value of three defects minimum. Improved RBF neural network is used to research on multiobjective optimization of warpage, weld-lines and air-traps. Three target values are evaluated comprehensively, and we can get the optimal process parameters of the best evaluation. Main missions of this article includes:1. There is only the value of warpage can be read in analytical data of CAE software. The size of air-trap and the length of weld-line present with image form. In order to make the following mathematics modeling easily, source files of air-trap and weld-line in Moldflow are processed by MATLAB.2. An accurate and reliable CAE simulation model is set up. Injection molding process simulation analysis is arranged by uniform experimental design. According to analysis result, regression equations of warpage, weld-line length and air-trap size are established. Regression coefficients are test of significance and analyzed, we can get influences of process parameters to the three defect.3. The multiple uniform design of experiment is applied to arrange sample points, taking melt temperature, mold temperature and packing pressure as inputs and taking warpage, air-traps size and weld-lines length as outputs, RBF neural networks are established. Combination with Genetic Algorithm and global optimization in the networks, we can get the optimal process parameters.4. The related multiobjective optimization and fuzzy mathematics assessment method are studied. Choosing the appropriate membership function and weighting coefficient to improve original RBF neural networks, taking process parameters as inputs and taking values of comprehensive evaluation as outputs, improved RBF neural network is established. It can realize the estimation of quality of plastic parts, and get the optimal process parameters of the best evaluation. 5. An injection mold was designed and processed for real injection molding experiment. The related defects are measured and compared. It can validate reliability and effectiveness the methods of modeling and optimization.