Reliability-based Design Optimization Using Decoupling Technology

Structural reliability optimization design plays an important role in engineering.The double-loop method is a traditional method to solve reliability optimization.In the reliability optimization design,the double-loop method has high accuracy.However,the calculation efficiency of double-loop method is very low for complex engineering problems.Thus,to improve efficiency of structural reliability optimization design,this paper proposes two decoupling method,which are as follows:(1)three-phase decoupling technique for structural reliability optimizationThe first stage: The safety factor is used to optimize the structure.the optimization result using reliability safety factor is utilized as the initial sample point of the second stage optimization;The second stage: the reliability sensitivity and weight factor is used to adjust the sample points to obtain the data set of the objective function and the failure probability;The third stage: the objective function and the failure probability relation curve is fitted,the new objective function can be obtained by using the curve and acceptable failure probability,and the relation curve between the objective function and the failure probability is fitted again until the convergence criterion is satisfied,.the final optimization design variable can be obtained.The examples shows that the method has high accuracy and efficiency for non-linear functions,and calculation results are not sensitive to the reliability safety factor and the weight factor.(2)structural reliability optimization based on strength factorFirstly,the deterministic optimization of structure is used to find the initial optimal design.The strength factor is derived by using the limit state function and reliability index,and the most possible failure point is obtained by using the first order reliability method.The deterministic optimization and reliability analysis are operated alternately until the convergence criterion is satisfied.The results show that the proposed method is applicable to multiple constraint functions and has high computational accuracy and efficiency for non-linear limit state functions.