Research On Multidisciplinary Design Optimization Method Based On Collaboration Model Surrogate Optimization And Set Strategy

The design of current products is becoming more and more complex,including more subsystems or sub-disciplines and more complicated coupling effect.The calculation is more and more time-consuming.In the process of modern product design,it is very hard to address the strong coupling effect and the large amount of calculation.One of the solutions is to adopt multidisciplinary design optimization(MDO),which makes the development of more efficient MDO methods a research hotspot.Because actual engineering problems often contain various uncertainties,the quality and performance of actual engineering products will be greatly affected by these uncertain factors.Therefore,the research on multidisciplinary design optimization methods in uncertainty is more in line with the real situationn of engineering problems.This doctoral dissertation focuses on the MDO method.Using with the approximate model technique,this doctoral dissertation proposes new MDO methods that can solve the multidisciplinary design optimization problem of complex engineering system without calling complex and time-consuming multidisciplinary analysis and cumbersome sensitivity calculation.The actual engineering problems often contain uncertainties,and the research focuses on robust design.A set strategy approach for multidisciplinary robust design optimization(MRDO)under uncertainty is proposed,The engineering example results verify that the proposed method is feasible,effective and practical.Firstly,the basic principles and characteristics of seven common multidisciplinary design optimization methods are introduced.Then the advantages and disadvantages of various multidisciplinary design optimization methods are comprehensively compared.Finally,the research framework of the thesis is proposed.Secondly,a collaboration model surrogate optimization(CMSO)method for MDO is proposed.A collaboration model is constructed as a filter to select feasible sample points that satisfy system analysis or multidisciplinary analysis.These selected sample points are used to construct the surrogate models and verify and confirm the surrogate models,then the best surrogate model is selected to approximate MDO model.The optimization model is optimized by SQP method.The MDO framework based on collaboration model(CM)for collaborative sampling(CS)and combined with approximate model is built.The effectiveness,accuracy and efficiency of the method are verified by two examples.The example results show that the method does not need to repeatedly call complex and timeconsuming multidisciplinary analysis and cumbersome sensitivity calculation,which improves the calculation efficiency.Thirdly,a multidisciplinary design optimization method based on CMSO and artificial bee colony algorithm(CMSO-ALMABC)is proposed.The augmented Lagrangian multiplier method(ALM)is used to address the complex equality constraints contained in the MDO model,and transform the equality constraints into unconstrained optimization problems by Lagrange multipliers.The framework of CMSO-ALMABC method is built by using CMSO with artificial bee colony algorithm,and the steps and processes of the method are elaborated.Finally,the effectiveness and feasibility of the method are verified by the examples.The method improves the global optimization ability,reduces the engineering product design cycle,and provides a possibility for more effectively solving modern engineering design problems.Fourthly,a set strategy-based MRDO(SSMRDO)method is proposed.The maximum variation analysis method is used to construct a robust design model and used for uncertainty analysis.The system optimization model is constructed by using the set strategy method to coordinate the coupling variables between fully autonomous subsystems,to obtain a new design space,and to optimize the sequence.The system obtains a robust optimal solution and an optimal robust design space.The framework of SSMRDO method is built and the steps and flow of the method are elaborated.The accuracy and effectiveness of the method are verified by two examples.This approach allows designers to start with a wide range of design variables and then gradually reduces the set to get more information.The proposed method can reduce design time and improve system robustness.Fifthly,the design of the concept ship design,four-high rolling mill and air cooling battery thermal management system have been optimized by the proposed methods in this doctoral dissertation.The engineering example optimization results show that the proposed methods can effectively reduce the time-consuming and cumbersome calculation in engineering product design optimization and meet the requirements of modern engineering product design.Finally,the main work of this doctoral dissertation is summarized,and the future direction worthy of further research is expected.