Study On Optimization Methods Of Computer-aided Welding Material Design

This paper is about the computer aided welding material design and optimization problem.By surveying the development of the computer aided welding material optimization and the related region, this paper formed a solution to the welding material computer aided design and optimization, that includes experiment design method, mathematic modeling method and optimization method.The mathematic modeling method is the critical part in the welding material optimization problem. The ANN was chosen as the mathematic modeling method for welding material design in this case. This is because the effect of component in welding material is very complex and no explicit function expression can be gotten before build up the relationship between the recipes of welding material and its performance. ANN can be a universal function approximation tool if an ANN has a structure of three layer: input layer, hidden layer and output layer.Through theoretic analysis, the uniformity aimed experiment design method was chosen as the best experiment design method to work with the chosen mathematic modeling method ANN (Artificial Neural Network). This paper formed a new method to get the wanted uniform design by computer sampling a number of experiment plan, counting their uniformities and choosing the best one of them as the last experiment plan. This experiment design method is simple and efficient and is not worse than experiment plan gotten from the uniform design tables.Genetic Algorithms (GA) was chosen as the optimization method for the welding material optimization problem, since the optimization problem of welding material design is a multi-objective optimization problem (MOP). The answer of MOP is not one solution but a set of solutions according the Pareto-Solution of MOP. This paper worked out a new GA. In this GA the multi-objective value of a welding material was normalized, and the every given designed welding material can be sorted by the permutation formed according the normalized multi-objective value from large to less. The best one in a permutation must be in the current Pareto font line. And all the best ones in every permutation formed a set of current best Pareto solution. By the evolvement of the population in the GA, the Pareto solution of the MOP of welding material can be gotten. By comparing between the jobs worked out by this paper with the traditional welding material design method and the traditional welding material optimization method, the conclusion that the solution formed out by this paper for computer aided welding material design and optimization is better can be gotten.