Global Sensitivity Analysis And Parameter Identification Based On Approximate Bayes Of Sheet Metal Forming

Sheet metal forming technology is widely used in manufacturing industry.With the improvement of computer technology and theory of numerical simulation analysis,CAE analysis technology is used in sheet metal forming simulation analysis in order to reduce development lead times and cost of sheet metal forming part.However,the forming parts are affected by many factors and forming defects often occur,such as cracking,wrinkling and so on.In engineering practice,parameters of forming process entirely designed by experience and verified by CAE analysis technology,then a feasible solution is obtained,which is very time-consuming and laborious.At the same time,it is lack of analysis on the sensitivity of each parameter and automatic optimization of final results.This process depends too much on subjective experience design and cycle test.Therefore,this paper provides a method to determine the ranking of sensitivity parameters,and establishes a feasible parameter interval recognition method based on image and algorithm,which is of great significance.For sheet metal cold forming process,the sensitivity of each parameter is determined firstly in this paper.Then,the Variational Auto-Encoder model in deep learning combined with approximate Bayesian method is used to obtain the feasible posterior probability distribution which is the interval estimation of each parameter.Compared with the traditional numerical hybrid algorithm,the results show that this method has certain advantages.The main research contents are as follows.1.The basic theory and simulation flow of numerical simulation in sheet metal forming are reviewed,and the numerical simulation analysis of the case is carried out.2.According to the design parameters of sheet metal stamping process,the global sensitivity analysis based on surrogate model is carried out.By comparing the evaluation index of several commonly used surrogate models,the optimal model is selected.Then the global sensitivity analysis based on surrogate model is carried out on the design parameters in sheet metal forming process.3.In order to get the design interval of parameters and consider the uncertainty of parameters,based on the approximate Bayesian framework,a method to identify the feasible interval of design parameters in sheet m etal stamping process is proposed,and the results obtained by this method are compared with those obtained by traditional multi-objective optimization method.4.The closed-loop system of parameter identification in sheet metal forming process is established,and the graphical user interface is developed.The dependence on experience can be greatly reduced,and get the parameter interval estimation when the formability of part is satisfied.