Multidisciplinary Reliability Design Optimization Method Based On Evidence Theory

Complex engineering problems often involve many disciplines which coupled with each other, and exhibit a strong non-linear characteristic. Only Multidisciplinary Design Optimization method can handle this particularity situation. At the same time, uncertain exist in the multidisciplinary system due to model uncertainty and parameter uncertainty. Therefore, optimization based on the uncertainty often gets a more efficient design results. So researchers pay more attention to Reliability-based multidisciplinary design optimization (RBMDO).Based on RBMDO, common method to description the uncertain parameter is probability model, but it requires more information to determine the probability distribution function and that is more difficult in a project. Researchers increasingly concerned about the non-probabilistic model and non-exact model which need less information about uncertain variables. Evidence theory as a non-exact model to describe the uncertain parameter has the advantages of less demand for information and clear expression of the degree of uncertainty. But because of evidence theory is discontinuous, its application in mechanical field has been limited. This article mainly about the multidisciplinary optimization design method based on evidence theory.First, evidence theory is used to describe uncertain variables in multidisciplinary system and Bayesian method is applied to accurate the evidence, then the approximate mean and standard deviation of uncertain variables can be obtained. Advanced first order second moment method is utilized to complete the system reliability analysis, Monte Carlo simulation is used to verify the accuracy and efficiency of the advanced method.Second, based on the sequential optimization strategy, the three nested reliability-based multidisciplinary design optimization problem is transformed into single-layer sequence optimization mode. It can improve the solve efficiency of optimization problem. In deterministic optimization process, Bi-Level Integrated System Synthesis process is used to complete the multi-disciplinary decoupling and seek the optimal solution.Next, an optimization method based on mobile response surface and evidence theory is developed to deal with the problem that multidisciplinary optimization use approximate model. Evidence theory can be seen as interval model with no-exact probability distribution, and then the performance measurement method is employed to obtain the most unreliable points. By comprehensive use of sequence optimization strategy and the response surface method based on trust region to solve the optimal result of the system.The last, low-pressure casting is a typical multidisciplinary optimization design problem, so we construct a multidiscipline approximate model between the casting solidification time and casting parameters, and evidence theory model is used to description uncertain parameters. We use the proposed method to seek optimal casting parameters, and the result proves this method can be integrated to solve multidisciplinary engineering problem.