Multidisciplinary Design Optimization Algorithms And Their Applications In Underwater Vehicle

Multidisciplinary Design Optimization(MDO)is the development direction of the complex electromechanical system.This paper presents an in-depth theoretical study of the problems existing in the multidisciplinary design optimization algorithms and its application in Unmanned Underwater Vehicle(UUV),the innovative research results obtained in this paper are listed as follows.(1)For testing how the performance of each MDO algorithms changes with the com-plexity of MDO problem,this paper firstly presents a variable complexity optimization problem which allows one to obtain a MDO problem with arbitrary complexity by speci-fying its changeable parameters.Then the mathematical model of each MDO algorithms to solve the presented optimization problem is built and the corresponding optimization program is coded.Finally,four investigations are performed to evaluate the performance of different MDO algorithms by varying the number of disciplines,global variables,local variables and coupling variables of the presented variable complexity problem,respec-tively,and the results supply a reference for the selection of MDO architectures in solving engineering problem.(2)According to the present situation of the low efficiency of traditional Multi-Disciplinary Feasible(MDF),this paper proposes a MDF algorithm with efficient gradi-ent calculation through studying the gradient computation method.Firstly,the MDF algorithm based on analytical theory is presented to improve the optimization efficiency through introducing the analytical theory to calculate the multidisciplinary system gra-dient information.Then the MDF algorithms based on discipline-merging method is pro-posed to further improve the optimization efficiency by merging the adjacent disciplines reasonably.Finally,the presented algorithm is tested by a typical MDO problem and the results show that it is an effective and efficient monolithic multidisciplinary optimization algorithm.(3)For the situation that many distributed MDO algorithms have difficulties in nu-merical convergence,this paper proposes two distributed multidisciplinary optimization algorithms through studying the extreme conditions of optimization problem with in-equality constraints and the penalty function method:the distributed multidisciplinary optimization algorithm based on Karush-Kuhn-Tucker(KKT)condition and the dis-tributed multidisciplinary optimization algorithm based on penalty function method.The former decomposes the original multidisciplinary optimization problem into a set of non-linear equations and multiple sub discipline optimization problems through a two level optimization process,which realizes the distributed parallel optimization of each sub discipline.The latter decomposes the original multidisciplinary optimization prob-lem into a system level optimization problem and multiple sub discipline optimization problems,and derives the gradient expression of the system level optimization problem based on the optimal sensitivity theory,which improves the solving efficiency of the sys-tem level optimization problem.Both two algorithms can be used to solve the general multidisciplinary optimization problem and their convergence proof is provided.Finally,two multidisciplinary optimization problems are introduced to perform a comprehensive evaluation and verification of the presented two algorithms and the results demonstrate that it has a good performance both in convergence and efficiency.(4)Based on the research of MDO algorithms and parallel computation,This paper applies MDO algorithms into the design of Energy-saving Unmanned Underwater Vehicle(EUUV)and realizes the efficient optimization design of EUUV system.Firstly,according to the characteristics of multidisciplinary design optimization,this paper divides the EUUV system into several disciplines and establishes a multidisciplinary optimization model of EUUV system based on these disciplines.Then the characteristics of EUUV's multidisciplinary optimization model are analyzed and the MDF algorithm based on discipline-merging method are selected as the efficient optimizer.Furthermore,the partial derivative information and the gradient analytic equation of the coupled system are solved in parallel in the high performance cluster,which make the parallel characteristics of the EUUV's multidisciplinary optimization process fully excavated.Finally,the whole EUUV multidisciplinary optimization model is solved efficiently and the result is satisfactory.