Research On Structural Reliability Analysis And Optimization Based On Intelligence Algorithm

Structural reliability, as one of major factors to judge quality of structure, has been highly concerned by the engineering. Structural reliability is an integrated engineering course, and it mainly includes reliability analysis, optimization design, evaluation, use and control. Reliability analysis and optimization is the key of structural reliability, and it fundamentally decides structural reliability and service life. Structural reliability analysis and optimization is a calculated way to obtain optimal design variable, which combines reliability analysis method with optimization design completely, and define reliability index as constraint condition or goal. The ideas and models of structural reliability analysis and optimization methods are more reasonable than traditional design methods. The engineering practices also show that the method can significantly improve structure design quality and inherent reliability, and good economy benefit can be obtained. Therefore, the current research on structural reliability analysis and optimization has become an important actively exploring domain for domestic and foreign scholars.According to the structural risk minimization principle in statistical learning theory, support vector machines theory is a method to deal with machine learning by optimization methods. Support vector machines theory possesses outstanding advantages in solving small sample learning and prediction, nonlinear and high dimensional pattern recognition problem. Particle swarm optimization which simulates birds flock’s looking for food can make the flock find the optimal solution through working together. This algorithm is simple and easy to realize, and it has few parameters to set. Particle swarm optimization is an effective optimization tool for nonlinear continuous optimization problem, combination optimization problem and integer nonlinear optimization problem. Therefore, according to support vector machines theory, particle swarm optimization and its improvement method and structure reliability model, the reliability analysis and optimization design of structure combined with the National Science Fundamental Project (51175442) is mainly discussed in this paper.Combining support vector machines theory with advanced first order second moment method, a structural probabilistic reliability analysis method based on support vector regression is presented for the structure with implicit limit state function. One numerical test and two practical engineering examples in this paper are provided to demonstrate the prognostication precision of the support vector machine model, and the examples also show the feasibility, veracity and higher engineering practical value of the method proposed. Combining interval analysis with support vector machines, a non-probabilistic reliability analysis method based on support vector machines is developed for the structure with implicit limit state function. This method is an iterative method, which has appropriate iterative procedure, and high accuracy non-probabilistic reliability index is calculated by the method. The paper also shows that the method is feasible, correct and accuracy by other four examples. Therefore, the method has a certain reference value to non-prbabilistic reliability analysis of other uncertainty structure.According to the premature phenomenon of particle swarm optimization, chaos particle swarm optimization takes full advantage of particle swarm optimization’s global search capability and chaos optimization’s local search ability. Chaos particle swarm optimization and support vector machines are applied in structural probabilistic reliability optimization design. A method of structural probabilistic reliability optimization based on chaos particle swarm optimization is presented and a plane rigid frame is improved. For reliability analysis of implicit limit state function, the probabilistic reliability optimized design of a plane truss structure is performed through the structural probabilistic reliability analysis method, which is based on support vector machines. Two calculation examples both indicate that the results are superior to particle swarm optimization, best vector method and penalty function method. The method has the advantage of global convergence and high accuracy. It applies to complicated structure probabilistic reliability optimization design, and has better practical value and exploitation ability in engineering problem.Structure non-probabilistic reliability index is obtained by convex model method. Combining interval analysis and support vector regression, implicit structure non-probabilistic reliability index is calculated. The method of structure non-probabilistic reliability optimization is presented on the basis of simulate anneal-particle swarm optimization, and two plane truss structure are improved by the method. The results of examples prove that the method has global optimization quality and strong probabilistic jumping quality. The examples also show that the method can quickly find the global optimal solution, and the optimal results of the method are much better than particle swarm optimization and penalty function method.A new method for implicit structure hybrid reliability analysis based on support vector machines and hybrid reliability model is proposed in this paper. The validity and precision of the method are verified though hybrid reliability analysis of a cantilever and a plane truss structure, and it also shows that the structure has a certain risk if it is analyzed by probabilistic reliability model, when it includes both probabilistic and non-probabilistic parameters. According to the structure of multi-failure mode, a hybrid reliability analysis method is proposed. Implicit and explicit structure system hybrid reliability is analyzed by the method, and the results is an interval-valued. A method for structure hybrid reliability optimization based on intelligent single particle optimizer is proposed in this paper, combining with intelligent single particle optimizer and hybrid reliability analysis method. The results of the study show that intelligent single particle optimizer is superior to chaos particle swarm optimization and simulate anneal-particle swarm optimization in optimize performance, and its solution much closer to global optimal one.Structure reliability analysis and optimization is conducted in this paper using support vector machines and three modified particle swarm optimizations. This paper mainly studies the problems of probability and non-probability reliability analysis and optimization of structure with implicit limit state function, and the problem of hybrid structure reliability analysis and optimization. The proposed methods have high value on both theory and engineering practice, and establish the foundations for further research.