Evolutionary Identification Of Nonlinear Material Model

With the increasingly deepened understanding of material nonlinear behavior, the effects of the nonlinear properties to the results of mechanical computation have got more and more attention. Nonlinear material modeling has been a hot topic to mechanic researchers. Recently, the back analysis method based on measured data becomes one of the most important research approach, which has been used in different area and obtained a number of achievements. However, because of the extremely complexity of engineering materials, the mechanism of its deformation and damage is not well known yet. In this case current identification methods encountered troubles that are difficult to overcome, there are much work to be done.This thesis introduced the evolutionary concept to the field of material mechanic modeling, by use of the global optimization technique of Genetic Algorithms(GAs) and Genetic Programming(GP), the author studied the evolutionary approach for identification method for nonlinear material modeling. The following are the main study content and major works.Using the advantages in global optimization of GAs, combined with nonlinear mechanic analysis method and Finite Element Method(FEM), The author proposed a new algorithm—evolutionary identification of the parameters in nonlinear material constitutive model. Simulating the evolution of nature life-forms, the algorithm applies genetic evolution to a set of trial solutions which will change themselves adaptively during the performance to find the reasonable solution globally. It can solve the difficulty of local optimize occurred in traditional methods.A new approved GP method—structural and parametric coupled evolution method—was proposed for model identification. Under condition of only having qualitative knowledge of studied problems, this method establish a evolutionary general expression cover the whole solution space based on the qualitative construct set of these problems. According to the evolutionary general expression, this method can perform coupled identification of the structure and parameters globally through genetic evolution. The new method needs not prior knowledge about the size and structure of the final solution, and it generalized greatly the application of evolution algorithms in engineering.Using structural and parametric coupled evolution method, combined with nonlinear mechanic analysis method, thesis also proposed a new algorithm—evolutionary identification of coupled structure and parameters in material nonlinear constitutive model. In complex cases that the mechanism of deformation and damage of materials are not well known, according to the qualitative knowledge of material nonlinear behavior, this algorithm build a infinite model set firstly, then adjust the structures of a subset of trial models positively through genetic evolution to find more satisfied solutions. Thus, the structure and parameters of the material nonlinear constitutive model is identified globally in the dynamic self-adaptive progress. It overcame the limit that the traditional methods can only check few models.Engineering media effect by multi-factor form a complex mechanic system. The performance of subtle numeric simulation is a very difficult work. To establish a equivalent empirical analysis model is a good approach. Using structural and parametric coupled evolution method, a new algorithm—evolutionary identification of empirical analysis model for complex mechanic system—was propose by the author. It can identify the structure and parameters of empirical analysis model from successful engineering examples.Using the algorithms and software, identification of the nonlinear constitutive model of composite material and the empirical analysis model of slope stability analysis were performed. The satisfied results proved that these algorithms could get global reasonable results. During...