| Sheet metal parts are widely used in aeronautics, astronautics, automobile, weapons and other fields, so sheet metal forming techniques decide the international competition of a country’s manufacturing. Optimization design of forming technology parameters based on the method of finite element simulation combined with optimization algorithms can improve sheet metal forming performance and extricate people from heavy error test work, but the uncertainty factors throughout the forming process restrict the further application of sheet metal forming optimization design. In this work, fully considered the influence of cognitive uncertainty, abnormal sample points and parameter uncertainty factors during the sheet metal parts forming process, the key problems of finite element simulation, robust optimization model construction, robust surrogate modle establishment and robust optimization solution are researched, then the technique system of sheet metal parts forming technology robust optimization can be proposed to guarantee the stability and reliability of forming quality under the low cost condition. Therefore, rejection rate can be reduced and the manufacturing cost can be saved. The related research results are obtained as follows:A complex part matching algorithm based on contour curve feature is advanced to improve the automatic and intelligent degree of finite element simulation. Moreover, the initial solution for robust optimization can be obtained. Three dimensional part contour curve matching algorithm based on the skills of multiscale filter, feature calculation, partition equivalent class by Hausdorff distance and ISODATA cluster analysis method can enhance the accuracy, fault tolerance and robustness of the matching algorithm. This algotithm can search the best matching similar part of the simulating part in successful simulation parts database, then the successful simulation experience of the best matching similar part can be refered. Therefore, the pre-processing of simulating part can be set quickly and the finite element simulation cycle can be reduced. Moreover, the technology parameters of matching part can be used as the initial solution for robust optimization. So, this research can decrease the optimization iterations.The influence of sheet metal parts forming process parameters on forming quality indicators significance analysis method based on gray relational combined with experiment design is improved to advance the robust optimization efficiency and feasibility. The relationships of sheet metal parts forming process parameters and forming quality indicators are calculated to analysis the main influencing factors of the forming quality. Then the design variables and the noise factors for robust optimization can be selected more objectively and more accurately. This method can decrease the influence of sheet metal parts performance influencing factors cognitive uncertainty on model construction, and convert the traditional method by experience, analogy to the mathematical and intelligent method.The robust surrogate model construction method based on support vector machine is improved, and the fast training algorithm for support vector machine robust surrogate model is proposed. Support vector machine method is applied to construct the surrogate model of sheet metal forming in order to solve the problem that the traditional surrogate models nonlinear expression ability is not strong and its accuracy is not high. So, the surrogate model robustness and training efficiency can be advanced. Weight determination method based on the normal distribution of probability density function with mean med{ei} can decrease the excessive fitting caused by redundant sample datas. Moreover, the median of regression error is used as standard to calculate the weight value in order to enhance the robustness of least squares support vector machine surrogate model. Then abnormal sample points uncertainty factor can be reduced and the prediction precision of surrogate model can be improved. Furthermore, based on the matrix inversion recursive computation method using matrix relation of two successive iteration steps, fast training algorithm of iteratively weithed least squares support vector machine is put forward to decrease computation complexity, and shorten the modeling time of robust surrogate model.Non-nesting robust optimization method based on interval analysis is advanced to greatly reduce the robust optimization dependence on uncertainty information data, and increase the interval type robust optimization efficiency. The uncertain objective and constraint functions can be transformed to certain objective and constraint functions based on the interval theory, which combined with the surrogate model technology and multi-objective genetic algorithm construct the non-probabilistic robust optimization method based on interval analysis. This method can greatly reduce the robust optimization requests for original data and expand the research objects and the practical scope of robust optimization. Moreover, based on the method of function best consistent approximation instead of double nested optimization, the upper and lower bounds of the interval function response can be calculated, so the computation complexity of the optimization process can be greatly reduced by avoiding double nested optimization. The polygon blank shape is robustly optimized by such method, and the result, compared with deterministic optimization, can decrease the fluctuation of product quality and reduce the rejection rate.On the basis of deeply research on the above key technologies, the robust optimization technology route for rubber parts forming process is proposed. Rubber forming process robust optimization is carried for a practical rubber forming part.Then process validation experiment is carried under the process condition of robust optimization design, and the better forming quality parts can be obtained. |
