Application Of Particle Swarm Optimizer And Support Vector Machine In Ship Structural Optimization

As an import part of ship design, ship structural optimization tries to findoptimized structure. For ship structural optimization problems, design variables are oflarge numbers and different kinds. Also, constraints is very complex. All adds to thenonlinear characteristic of ship structural optimization problems and the difficulty tofind optimized solutions. Optimization algorithm is a crucial part of structuraloptimization. Particle swarm optimizer(PSO) is a newly developed heuristic algorithm，with a simple form and easy to implement, could be used in structural optimizationprocess. This article introduces particle swarm optimizer into structural optimization.After validating the convergence and effectiveness of particle swarm optimizer instructural optimization through optimization of two classical truss structures, anoptimization strategy based on particle swarm optimizer through integration ofMATLAB and Finite Element software MSC/NASTRAN is proposed in this article. Tovalidate the effectiveness of the proposed method, a finite element model of the cargoarea of a multi-purpose ship is built and optimized using the above mentioned method,which yields a good result, thus validating its effectiveness.Iterations of Finite Element Analysis is often an unavoidable part of shipstructural optimization so as to get structural response to set objective or constrains.Due to the complexity of ship structural optimization, quite a few number of iterationis needed and the time consumption of optimization process could be huge. Regressionmodel could help reduce the time consumption of the ship structural optimizationproblems. Support vector machine (SVM) is an effective method to build regressionmodel to handle various nonlinear regression problems. For structural optimizationproblems, it could be used to build regression model for structural response so as topredict structural response and replace time-consuming finite element analysis.However, it’s usually very difficult to set proper parameters for support vectorregression model. Traditional selection methods are often empirical and it’s hard tofind proper parameters for a particular problem using traditional methods. In thisarticle parameter selection problems are abstracted as optimization problems.Optimization methods are adopted to solve parameter selection problems. To bespecific, particle swarm optimizer is used in this article to select proper parameters for support vector machine. The thesis uses empirical parameter and the parameterselected using method proposed above to build SVM to predict structural response andthe results are used to validate the effectiveness of the proposed method.A structural optimization approach integrating particle swarm optimizer andsupport vector machine are proposed in this article. Based on the parameter selectionmethod, a support vector machine with proper parameters could be built and integratedwith support vector machine for structural optimization. To validate the effectivenessof the integrated structural optimization process, the above method is used to optimizeship structures.