Parameter Identification Of Viscoplastic Model Considering Dynamic Recrystallization

Microstructure prediction is one of the most important ways to improve the quality of the final forging. The leading difficulty, which is encountered in simulating the microstructure evolution, is how to identify accurately the material parameters included in the constitutive relation considering microstructure evolution at present. So, taking the constitutive relation considering dynamic recrystallization as an example, this paper does research on parameter identification theory and method. The research has not only strong engineering background, but also academic significance in developing parameter identification theory and method.We first construct the objective function, which is defined as the weighted square sum of the differences between the numerical results and the experimental results of the microstructure and the load. The experimental results are obtained from the cylindrical upsetting experiments, and the numerical results are obtained from the rigid-plastic element software by the author. We introduce the influence of strain rate on the mobility of dynamic recovery and the influence of the accumulated dislocation energy in the newly recrystallized grains on the driving force of dynamic recrystallization, in order to make the model be in better agreement with the actual physical process. Then this paper studies the characters of the objective function by the uniformly sampling method. It can be concluded that the characteristics of the objective function include the following aspects mainly: (1) non-convexity, (2) poor degree of sensitivity in the parameter space, (3) difficulty in defining the numerically feasible solution space clearly, (4) huge computational burden. The solution, which is generated by the traditional sampling operator, biases toward the end whose absolute value is large. To overcome the above problem, this paper constructs a mixed sampling operator, which couples the samplings of the numerical value and the power exponent.Based on the characteristics of the objective function, we develop a global optimization method. To guarantee the global search ability, genetic algorithm is regarded as the fundamental algorithm of the designed algorithm. We construct thehybrid type genetic operators, in order to prevent the algorithm from premature and overcome the problem that the obtained solution from the traditional sampling operator biases toward the end whose absolute value is large. Due to the slow convergent speed of the genetic algorithm and the poor degree of sensitivity in the parameter space, the solution from the genetic algorithm is refined by the Gaussian-Newton algorithm and the augmented Gaussian-Newton algorithm. During the procedure of applying the augmented Gaussian-Newton algorithm and the Gaussian-Newton algorithm, if the obtained solution is numerically infeasible, the flexible tolerance polyhedral method is applied to find a numerically feasible solution as a substitute in order to make the optimization continue.Based on the constructed global optimization method, parameter identification software PIIA is developed. Taking Richards model and MMF model as examples, the developed method is proven to be an efficient and effective global optimization method for the problem with large differences of orders between the upper limits and the lower limits of the parameters. Then, the parameters for 26Cr2Ni4MoV are obtained through the developed software. The simulated results agree with the experimental results well. The above facts prove that the developed software and algorithm are reliable, accurate and practical.