Research On The Multi-objective And Multidisciplinary Optimization In The Conceptual Design Of Mechatronic Products

With the rapid increase of the complexity of modern mechatronic products,researchers pay more and more attention on the design optimization of multidisciplinary mechatronic products during the system conceptual phase. The main difficulties arise from two aspects: multiple interdisciplinary domains are involved during the system conceptual phase and a lot of conflict design objectives must be taken into account. To obtain the satisfactory solutions the conceptual design process of the product system must be executed iteratively. In addition, it’s imperative to implement the automatic and efficient integration betweeen the system design and the system optimization. Although there are some related research work at present,some shortcomings are still imposed: (1) The efficiency of the meta-heuristics multi-objective optimization algorithms applied on the concetual design of the complex mechatronic products are not satisfatory; (2) The existing multi-objective algorithms can not handle the constraints in a good manner; (3) Complex mechatronic products involve multiple domain knowledge, the corresponding multidisciplinary otpimization problems usually present highly inner couplings, which increase the computation cost to a great extent. The perfomances of the existing decoupling strategies can not satisfy the practical requirements; (4) Most of the existing conceptual design platforms make the design decisions based on the design experiences, some highly depend on the external optimization tools,in which there exist the interation difficulty and the complexity of user operations. Therefore, with respect to the above problems, this dissertation proposes the following work:(1) A geometric structure based multi-objective particle swarm optimization algorithm.The most important characteristic is that the whole population will be evolved under the help of the geometric structure of the Pareto front. The current Pareto front is considered as a set of scattered points and fit to construct its geometric parameter space, compute the normal direction for each point in the objective space to precisely obtain the corresponding guiding points, finally the population will be evolved towards the better regions under the help of the guiding points.(2) An efficient hybrid search mode algorithm is proposed to handle the constrained multi-objective optimization problems. The algorithm mainly consists of two steps:Firstly, feasibility search mode. In this step the constraint conditions will be handled using an adaptive differential evolution strategy. Secondly, optimality search mode. On the basis of the first step an evolution process is conducted for each feasible individual using its corresponding local elitist and globla elitist.(3) A serialization partial decoupling strategy is proposed for the multidisciplinary design optimization, which consists of three steps: Firstly, the system is clustered into three subsystems through analyzing the sensitivities; Secondly, the serialization decoupling operation is conducted for each subsystem to ensure there is no coupling loops; Thirdly, a local optimization process is executed for each subsystem to ensure its consistency.(4) An integration framework based on pattern between system design and optimization is proposed. In this framework three steps are involved: Firstly, the construction of optimization problems. Extract the optimization variables and define other optimization elements according to the defined stereotypes; Secondly,the automatic selection of optimization method based on the semantics similarity;Thirdly, the update of design solution based on the feedback of the optimization results.