Research On Optimization Algorithm Of Structural Reliability Based On UGF

Traditional structural optimization design methods often fail to fully consider the uncertainty in engineering practice,resulting in the design of the structure can not meet the expected requirements.In order to solve the disadvantages of the traditional optimization design methods,reliability-based design optimization(RBDO)method is proposed,which can reasonably characterize the impact of uncertainty on structural design.However,there are still some deficiencies in the existing reliability optimization design methods for solving the problems of non normal random variables and highly nonlinear performance function.Therefore,in order to solve the above problems,this paper carries out in-depth research,mainly including:1.The universal generating function(UGF)method with high accuracy,which is not affected by non normal random variables and highly nonlinear performance function,is introduced to analyze the reliability.Several common solutions are proposed for the combination explosion problem of UGF method,such as merging similar items,non-uniform dispersion of continuous variables,and non-uniform clustering.The non-uniform discretization is mainly carried out according to the proportional series,which makes the discrete points in the neighborhood of the sensitive points to be dense,so as to ensure that the total number of discrete states is less,and the higher reliability analysis accuracy can also be obtained.By selecting the sensitive points near the limit state surface as the clustering center,non-uniform clustering can ensure the accuracy and efficiency of the solution.2.In order to improve the accuracy of reliability-based design optimization problem,an UGF-offset vector method is proposed.The algorithm introduces high-precision reliability analysis method and offset vector solving strategy to complete the optimization,and the iteration process is divided into three parts.In the first step,the least square method is used to fit the response surface regression model of the migration function,and the offset vector is solved according to the model and the allowable reliability index;in the second step,the deterministic optimization is completed according to the obtained offset vector,and the current design point is obtained;in the third step,the UGF method is used for reliability analysis and evaluation,and the offset function is reconstructed according to the relevant constraints.An example shows that the proposed method not only guarantees the efficiency of the solution,but also improves the optimization accuracy,and solves the problem that the optimization result cannot converge when the performance function is highly nonlinear.3.To improve the stability of the UGF-offset vector method and simplify the optimization process,an UGF-direct mapping method is proposed.On the one hand,the dynamic mapping between design variables and probability index is established directly through a series of response surface,which eliminates the inner loop of traditional double cycle method,and decouples the nested problem into reliability evaluation and optimization design sequence.On the other hand,the UGF method is used for reliability analysis,which avoids the disadvantages of low accuracy or convergence of traditional moment method.The results of numerical examples show that the proposed method has significantly higher accuracy under controllable computational cost,and has higher robustness for nonlinear performance function and non normal random variables.Two new methods for RBDO problem solving and three key technologies for solving UGF combinatorial explosion problem are proposed.The proposed methods and techniques have high accuracy and controllable calculation cost.At the same time,it provides a new way to solve the bottleneck problems in RBDO field,such as highly nonlinear performance function and non normal random variable,which has theoretical significance and practical application value.