Applications Of Optimal Algorithm In Structure Design

Over the past 100 years structure optimization design has experienced considerable progress. Especially in recently 40 years, there were great developments either in theory or algorithm. Now, many researches focus on topology optimization. Simply state, topology optimization consists of determining the best arrangement of a limited volume of structural material within a given spatial domain so as to obtain the optimal mechanical performance of the concept design. A main difficult in topology optimization is the singular optimum exists, which performs dimensions mutated and un-connectively in spatial domain (the singular optimum point was hard to get). So topology optimization is as regarded as a global optimization problem in concave feasible fields. There are no effective methods to solve the problem. This paper did a research on truss structure optimization. The optimal model of truss structure is established, in which the cross sectional areas of bar are taken as design variables, the structure weight is taken as objective function. In the process, the reliability of structural displacement and bar stress and the fundamental frequency are taken as constraint functions. From the engineering practice, all the reliability constraints, which are implicit generally with the design variables, are equalized and transferred to the conventional explicit constraints. Therefore, the original optimization model is transformed into the problem of cross sectional area optimization.This paper had great research on the development of optimization algorithm by analyzing typical optimal search method, such as greedy algorithm, simulated annealing algorithm, neural network and genetic algorithm (GA). According to the characteristics of truss structure, we choose genetic algorithm as the solution way. GA uses simply coding technology to express complicated structure, and executes simple genetic operation (selection, crossover and mutation) and prevail surviving mechanism on a group of codes to conduct searching on solution scope. At the same time, GA doesn't have to know every feature of problem, so it can accomplish solution only by genetic procedure embodying evolving mechanism. We develop a adaptive evolutionary algorithm based on neighborhood searching. Firstly, class the individuals by the fitness number to realize the division of population. Secondly, mutate each individual, that is mutating a good individual in a small range and a bad individual a large range. By this way, the good individual makes the mutation direction and the bad individual searches the possible good individual. The mutated individual follows the advantage direction of current population, so it can jump local optimum and get the global optimum and it can avoid the semi-blind of conventional Darwinian-type evolutionary computation. We should specially point out that by the function of the mutation operator, which ensures a stable convergence of the algorithm into a global optimum.