Genetic Arithmetic And It's Application To Design Of Composite Laminate

Genetic Algorithms as a good optimization method, which has been rapid developed in the last decades, have feature of robustness and global convergence. The goal of this dissertation is to study the theory of Genetic Algorithm(GA) and to apply it on optimization of composite laminate, then to develop a new gradient free optimization design method.At first, based on the simple Genetic Algorithm, an improved GA which adopts a combination of emigration policy and adaptive operator is presented. Results of numerical optimization show that the improved GA is effective on preventing premature problem. Second, by combining theories of strength of laminate with GA, a new gradient free optimization method with new code strategy is discussed. The representation itself does not change the nature of the problem, but effects the searching efficiency of the algorithm. So the representation of the parameters to be optimized is a very important factor and the new code strategy is also easy used with crossover operator and variation operator in GA applications. In order to solve the premature problem in strength optimization design of laminate, to combine the part-degenerate operator and adaptive operator is a good idea.Poisson's ratio is a parameter which measures the volume-changed of objects. Generally, the value of Poisson's ratio is around 0.3, but elaborate laminates can have especial Poisson's ratio more than 1.0, moreover the negative. For sake of using this property, GA is used to design the Poisson's ration of laminate, and some optimization samples are given to show that laminates with this especial performance are available.Thermal expansion coefficient is an important property of composite laminate. In order to satisfy the special request in aero field, zero expansion coefficient is perfect. On the basis of analyzing the behavior of the thermal expansion coefficient of laminate, a new method used GA has been developed. Compared with results of two methods, GA method is much better than the 'isotropy' method . Furthermore, GA method also overcomes the shortcoming of the ' isotropy' method that can be only used for laminate composed with one kind of material.Design of composite laminate is a multi-objective optimization problem. The targets of each optimization are not consistent and it is very difficult to find the best solution to satisfy all the targets, which is the nature of multi-objective problem. In order to solve this question, an improved linear weighting method is presented of which the weight factors are changeable. In this method, the weight factors are used for optimize variables during thealgorithm. Based on the method, combined with GA, a multiobjective optimum design of composite laminates is discussed. The results indicate that GA is an efficient method in solving multi-objective problems.