A Research Of The Application Of Cluster Analysis And Agent Model In The Optimized Design Of Multidisciplinary Reliability

According to the research and utilization of the parallel operation principle interacted in the engineering system,multidisciplinary design optimization constructs solutions to complex system,reduces the time cost of design structure,and obtains the best method of structure design.There are a lot of uncertainties in the practical engineering problems,and the reliability theory is developed by considering the uncertainties of the engineering structure.Probability method can well describe the uncertainties of the variables,but it needs a large amount of experimental sample information.It is very difficult to obtain enough sample points in the practical engineering.If the probability distribution of the variables is assumed,it may lead to large errors,resulting in the limitation of probability methods in some engineering problems.Non-probabilistic convex model describes the uncertainties of variables with interval and ellipsoid,it requires relatively few sample points and doesn't need considering distribution law.It has unique advantage in uncertainty modeling and multidisciplinary reliability optimization design.However,the engineering problems we are currently facing are mainly large and complex engineering systems,and these problems involve many uncertain variables,and when considering the uncertainty of multidisciplinary reliability design,not only the problem becomes complex,the amount of calculation is quite amazing.Therefore,one of the common challenges faced by multidisciplinary reliability design is computational cost.In recent years,many scholars have put forward some methods to improve the computational efficiency of reliability design.Focusing on the above problems,and on the basis of multidisciplinary reliability optimization research based on the convex model,this paper studies the application of cluster analysis in the reliability analysis based on convex model as well as the application of proxy model in multidisciplinary reliability optimization design based on convex model with the objective of improving solution efficiency.The main contents are as follows:(1)When the convex model is used to describe the uncertain variables,the uncertain variables are usually classified by empirical methods,which may lead to the low solution efficiency of the reliability analysis based on convex model.This paper classifies uncertain variables according to the correlation of uncertain variables and by adopting single linkage method as well as clustering analysis algorithm.One category is described by a convex model,and then I will carry out the reliability analysis and solution.Finally,the obtained data is compared with the non-probabilistic reliability index obtained by Monte Carlo to analyze the effect of different classification results on the computational accuracy and computationalefficiency.(2)Sequential optimization and reliability assessment method(SORA)is also called single loop strategy.Multidisciplinary reliability design based on SORA is mainly using SORA frame to separate reliability from optimization,both the two are MDO problems.In order to improve the calculation accuracy,this paper adopts the multidisciplinary feasible method(MDF)to carry out the reliability analysis and optimization.The MDF method is with the character of reliable results,but low computational efficiency.To improve the computational efficiency,the multidisciplinary analysis under MDF method is adopted,which is similar to Kriging model,to construct multidisciplinary reliability optimization design model and process with the application of proxy model.Finally,the results are verified through numerical and engineering examples.The results show that the multidisciplinary reliability optimization design with the application of proxy model can improve the computational efficiency under the premise of ensuring the accuracy of the optimization results.