Structural Optimization Research Based On Support Vector Machine And Genetic Algorithm

Genetic algorithm (GA) is a bionics algorithm, which simulates the biologic theory and is recently widely concerned in the fields of computer science and optimization. As an algorithm of global optimization, GA has many remarkable characteristics such as simplicity, common applicability, strong robustness, being applicable in parallel processing and vast application area. But in solving the large-scale problem of structural optimization, genetic algorithm each iteration required huge finite element compute to obtain the objective function value. This compute process significantly reduces the efficiency of the genetic algorithm. In order to improve the operating efficiency of genetic algorithms, scholars have tried to use approximate structure analysis method to replace the finite element compute. The common models of approximate structure analysis model are Taylor approximate model, response surface model and Kriging model, which have their own advantages and limitations. This paper presents a new structure approximate model established by support vector machine. Support Vector Machine (SVM), which was put forward by V.apnik, is a new and outstanding learning machine. It's the kernel content of statistical learning theory, and is a valid machine-learning tool in dealing with small samples. SVM overcomes some shortcomings of neural network, such as slow convergence, unstable solution, and bad generalization. So it has been widely applied to many areas such as pattern recognition, signal processing, automation, communication, et c. SVM had become a hot spot of research in the area of machine learning.In this paper, we first studied the genetic algorithm and its application in structure optimization design, including genetic algorithm's development, basic structure, main characteristics and its research background. Secondly, on the basis of simple genetic algorithm, we studied the fusion technology of genetic algorithm and support vector machine, we focus on how to use support vector machines to approximate the structures model, including the training samples chosen, the kernel function and its parameters chosen, and how to use genetic algorithms to solve the approximate model. Simulation computation of 10-bar truss and 25-bar truss were studied to verify the effectiveness of the algorithm. As support vector machines have good generalization ability and the complexity related to the dimension of the problem. So the algorithms discussed in this paper can use to any large structure optimization, is a new method for large-scale structure optimization.