Study On Structural Multi-objective Optimization Based On Particle Swarm Optimization And Approximation Model

Structural optimization is an important field of structural design. For large-scale engineering structure, the demanding of lightening the weight, the complex static and dynamic conditions and numerous high performance indexes, making it urgent to carry out the research and application of structural optimization. Through structural optimization, on the one hand, it needs to find an optimization algorithm with good searching capability, which can take reasonable strategy to deal with complex problems in the process of structural optimization, effectively seeking the optimal solutions. On the other hand, it also needs to construct approximation models, which can predict the structural responses, substituting the time-consuming finite element analysis, simplifying the optimization process and reducing the time.First of all, multi-objective particle swarm optimization is proposed. Base on the standard PSO algorithm, a strategy of decreasing the inertia weight is utilized, and particles that violated the constraints are punished respectively, then the mutation operator is introduced to enhance the diversity of swarms, giving this algorithm a better capability of global optimization. Secondly, some common approximation models are summarized. The theory of support vector machine is mainly introduced and the method of establishing support vector machine approximation model is proposed. Then MOPSO-SVM method is proposed. And then, a satellite structure multi-objective optimization model is established, which targets on structural weight and dynamic response. Based on this model, the satellite structure is optimized in Matlab software combining with multi-objective particle swarm optimization. Results of PSO and GA are compared, and then for the optimal solution, results of approximation model and FEA are compared. The efficiency of MOPSO-SVM method is counted. Last but not least, a multi-objective optimization model of ship liquid tank is established. According to the regulations of sloshing load, sixteen calculation conditions are determined by taking into account the different loading positions and loading rates. Based on this model, which targets on structural weight and total price of steel, the liquid tank structure is optimized in Matlab software with MOPSO.Optimal results of satellite and ship tank show that MOPSO is an effective optimization algorithm. The MOPSO-SVM method can improve the optimization efficiency and have high calculation precision. MOPSO-SVM method can offer an effective solution to the large-scale and complex structure optimization problem.