Warpage Optimization Methods Based On Artificial Neural Network In Injection Molding

The theory of artificial neural network (ANN) surrogate model is first described. It has so many advantages such as parallel-operation, distributed-memory, nonlinear-mapping, self-organizing, self-adaptability, self-learning, error-permissibility, recollection and vision, etc., which is widely used in many research fields such as control, data compression, forecasting, optimization, pattern recognition, classification and so on. With the continuous development and gradual maturity of artificial neural network technology, it has widely used in injection molding field.The main disadvantages of conventional BP methods arise from the facts that the optimal procedure is easily trapped into local minimum value and the convergence of algorithms employing such methods is very slow and unsteady. Aimed these problems, an improved activation sigmoid function conjunction with some advanced technologies for ANN is proposed in this paper. Numerical results show that the improved algorithm can converge faster and effectively avoid tracking into the local minimum.A sequential optimization design method based on integrated artificial neural network (ANN) and expected improvement (EI) sampling criterion is proposed to optimize the injection molding process. The ANN surrogate model is used to build an approximate function relationship between design variables and quality index, replacing the expensive simulation analysis in the optimization iterations. The adaptive process is implemented by El sampling criterion. This criterion can not only consider the predictor and its uncertainty, but also balance local and global search. It can improve the accuracy of the ANN surrogate model and quickly approach to the global optimization solution. As the applications, a cellular phone cover and a scanner are investigated. The mumerical results show that this method can effective reduce warpage of injection molded parts.A sequential optimization design method based on ANN surrogate model with Parametric Sampling Evaluation (PSE) sampling criterion is represented. The ANN model is used to build the relationship between the design variables and quality index to decrease simulation analysis time in the optimization process. The adaptive process is implemented by PSE sampling criterion. The PSE sampling criterion can not only overcome disadvantage of the traditional El sampling criterion that the value of El function jumps from one sampling point to another as the optimization proceeds, but also take the relatively unexpected space into consideration to improve the accuracy of the ANN model and quickly tend to the global optimization solution in the design space with a few sample points. Compared with the El sampling criterion, the numerical results show that the sequential optimization method based on PSE sampling criterion can converge faster and effectively approach to the global optimization solution.A sequential optimization design method based on ANN surrogate model and Weighted Expected Improvement (EWI) sampling criterion is developed to optimize the injection molding process. The EWI sampling criterion can flexibly balance local and global search to improve the accuracy of the ANN surrogate model and quickly approach to the global optimization solution. Then a sequential optimization design method based on Expected Improvement with the optimal weight (WEI) sampling criterion is developed. The WEI sampling criterion can not only overcome disadvantage of the El sampling criterion that usually takes longer to converge, but also precisely balance the local and global search and quickly tend to find the global optimization solution in the design space. Through the mathematic functions and engineering examples-an investigation is done for the propsed optimization method. Compared to El sampling criterion, the examples results show that the proposed optimization method can quickly and effectively approach to the global optimization solution.The authors gratefully acknowledge financial support for this work from the Major program (No.10590354) of the National Natural Science Foundation of China.