Development of numerical methodologies for parameter identification and shape optimization in metal forming simulations =Desenvolvimento de metodologias para identificação de parâmetros e otimização de forma em simulações numéricas de processos de conform

摘要

The interest of the stamping industry in the numerical simulation of sheet metal forming, including inverse engineering approaches, is increasing. This fact occurs mainly because trial and error design procedures, commonly used in the past, are no longer economically competitive. The use of simulation codes is currently a common practice in the industrial forming environment, as the results typically obtained by means of the Finite Element Method (FEM) are well accepted by both the industrial and scientific communities. In order to obtain accurate stress and strain fields, an effective FEM analysis requires reliable input data such as geometry, mesh, non-linear material behaviour laws, loading cases, friction laws, etc.. In order to overcome these difficulties, a possible approach is based on inverse problems. In this work, the following inverse problems in computational Mechanics are presented and analysed: (i) parameter identification problem, that refer to the definition of input parameters to be used in constitutive models for numerical simulations, based on experimental data, and (ii) initial blank and tool design problem, where the aim would be to estimate the initial shape of a blank or a tool in order to achieve the desired geometry after the forming process. New optimization strategies in parameter identification problems that lead more efficiently to accurate material parameters are introduced and implemented. The aim of these strategies is to take advantage of the strength of each selected algorithm and improve the overall robustness and efficiency of classical optimization methodologies based on single stages. Deterministic algorithms, evolutionary-inspired algorithms or even the combination of these two algorithms are used in the proposed strategies. Strategies such as cascade, parallel and hybrid approaches are analysed in detail. In hybrid strategies, cascade and parallel approaches are integrated. Two different approaches are presented and analyzed for the evaluation of the objective functions in parameter identification processes. The approaches considered are single-point and FE analyses. The single infinitesimal point evaluation seems to characterize an infinitesimal amount of material subjected to all kind of deformation history. On the other hand, in all FE analysis codes, the constitutive model is implemented and accounted for in each element integration point. Inverse problems, such as blank and tool design, are presented and described. In the case of the initial blank optimization process the design of a carter is presented. Also related to the initial blank optimization process, a study of the influence of the initial geometry definition in the optimization process is conducted. This study is performed considering the NURBS formulation to model the blank upper surface that will be changed during the optimization process. In the case of the tool design problem, a two-stage forging process is presented. In order to achieve a straight cylinder after forging, two different approaches are analyzed. In the first one, the initial geometry of the cylinder is optimized and, in the other one, the shape of the first stage tool is optimized. To parameterize the free surface of the cylinder different methods are presented. Furthermore, in order to define the tool in this example, different parameterizations are presented. The optimisation strategies proposed in this work efficiently solve optimisation problems for the industrial metal forming.