Multi-objective Dual-system Co-Evolutionary Algorithm And Its Application

Taking a kind of spacecraft (i.e. satellite and airship) module layout design problems asengineering background, this thesis studies a multi-objective co-evolutionary algorithm andits application.Recently, the co-evolutionary algorithms have been becoming an important researchfield and applied to solve the complex engineering problems extensively. In 2006, theComplementary Bi-system Co-Evolutionary Algorithm (CBCEA) was proposed by ourresearch group, whose feasibility and efficiency were verified by the numerical experiments.However, there are still many issues to be further improved. To overcome the shortcomings ofCBCEA, this thesis focused on two aspects. The first is to deal with the multi-objective layoutdesign problems using the non-dominated sorting strategy and incorporate it within thedual-system algorithm framework. The second is to develop an appropriate variable-graincooperative strategy of dual-system. Thus, the Multi-objective Dual-system Co-EvolutionaryAlgorithm (MDCEA) is proposed.(1) Comparing with CBCEA, MDCEA has two distinctive features as follows: a) Themulti-objective optimization strategy based on non-dominated sorting is employed inMDCEA to overcome the shortcomings of the multi-objective optimization strategy based onthe penalty function in CBCEA; b) A new variable-grain cooperative strategy is presented. InMDCEA, the original system is decomposed into two systems denoted by system A andsystem B, so-called a dual-system framework. System A adopts the variable-grain strategy ofthe design variables, while system B has all the design variables of the original problem and isdecomposed into several sub-systems coeVolving in parallel. Both of system A and system Bhave the same design objectives and constraints with the original problem.(2) The proposed algorithm (MDCEA) is implemented by C++ language for twopurposes. The first is to test the performance of MDCEA: the numerical experiments on thelayout design of a simplified satellite module are carried out. And the experimental resultsshow that MDCEA can solve the layout design problem more efficiently and effectively andoutperforms other co-evolutionary algorithms, such as CBCEA, MCBCEA and MCCEA. Thesecond is that the software module of MDCEA is incorporated into the layout designsimulation platform of satellite module for practical application. It's hoped that this study would be beneficial to the theoretical research of themulti-objective co-evolutionary algorithms. Meanwhile, it is also expected that the proposedapproach (MDCEA) would be expanded into the applications of other complex engineeringdesign problems such as the equipment layout problems of tank and tunnel boring machine.