Bi-level Integrated System Synthesis Collaborative Optimization Method Based On Approximate Model

Complex products design often involves a lot of disciplines, and the relationship betweendisciplines is complicated. So, the role of multi-disciplinary design optimization (MDO)which is able to solve the complicated, strongly coupled design problems has becomeincreasingly prominent. MDO optimization process is the most central content inmulti-disciplinary design optimization, which can directly determine the application reliabilityof MDO technology using in specific engineering optimal problems. Domestic and foreignscholars have made achievement in deeply studying the MDO optimization process, andpresented a series of method. However, due to the diversity and complexity of the products, itnot yet formed a method can be widely used in various MDO problems. On the other hand, asapproximate model techniques not only can ensure certain accuracy, but also can significantlyreduce the computation cost on repeat calling the accurate simulation models, it has becomean indispensable tool for studying MDO problems. This article is devoted to study the MDOoptimization process and approximate model, and to find a high efficiency and high reliabilityMDO optimal method which can search the global solution more quickly and accurately.Firstly, this paper analyzed the research status of MDO optimization processes andapproximate models, gave the iterative flow of general decomposition-coordinationoptimization process from discussing the relationship between system and sub-systemoptimization. It studied several multi-level MDO optimization processes, and compared theirpros and cons from decomposition technique, convergence, solution efficiency and so on.Secondly, based on the existing research, this paper made an in-depth study of Bi-LevelIntegrated System Synthesis Collaborative Optimization (BLSCO) Method which owns agood synthesis optimization performance. In view of it need call high computational accuratesimulation analysis and its discipline optimization can be effected by numerical noise easily,this paper proposed an advanced BLSICO-AM method based on approximate model. Itsfeasibility was confirmed by a strongly coupled non-linear optimization problem. Comparedwith the original BLISCO method, the advanced method greatly reduced the number ofsubsystem analysis and improved the solution efficiency. At the same time, the consistencyconstraints in advanced method have also been studied by comparing the different optimalresults using different allowed tolerance. The results show that it is feasible and effective using inequality constraints instead of strict equality constraints.Finally, on the basis of above findings, this paper studied the influence of using differentapproximate model techniques on optimize results, and applied them in the gear deceleratorengineering optimization problem. Test results show that they obtained good convergenceperformance and reduced the computation cost no matter using the BLSCO-AM methodbased on responsible surface or Kriging model, and the constructed approximate model’saccuracy is higher, the optimal result is better. The study also proved that not for all theproblems, Kriging model’s approximate accuracy is higher than response surface model’s, sowe need to choose appropriate approximate technical depending on specific issues.