Abstract: Whether the superplastic constitutive model considering grain growth can successfully simulate the forming process depends on the quality of parameter identification. However, it is difficult to obtain satisfactory parameters by experiment directly. The reasons are due to that it considers many physical processes and it is difficult to differentiate one physical process from other physical processes, at the same time, the model is involved many parameters. The material parameters are identified by the inverse analysis. The superplastic model is given; and the objective function is designed through minimizing the weighted square sum of the difference between the calculated value and the experimental value for the stress-strain relation and grain size-time relation; the zone of the parameter value is given based on the physical significance and the numerical result. A global optimization method is developed on the basis of the characteristics of the objective function. The developed optimization method incorporates the advantages of making global search for the genetic algorithm; at the same time, the method incorporates the strengths of relatively fast convergent speed for the Levenberg-Marquardt algorithm and the augmented Gauss-Newton method. To the question of the relatively large zone of the parameter value, the hybrid coding genetic operators, which make use of the real coding and exponential coding simultaneously, are designed. At last, taking Ti-6Al-4V as an example, a set of satisfactory material parameters are obtained by using the developed method. The calculated results agree with the experimental results relatively well.

Key words: Constitutive model, Global optimization, Inverse analysis, Parameter identification