On the Objective Function Evaluation in Parameter Identification of Material Constitutive Models - Single-point or FE Analysis

Abstract

This work deals with the identification of constitutive parameters by inverse methodology. Two different approaches are presented and analysed: the single-point and FE analysis. The use of these two different methodologies for the evaluation of objective functions in the identification process is still an open question and the interest in this field has been increasing among the metal forming community. To discuss this issue, two different constitutive models suitable for metals were used, i.e. an elastoplastic hardening model and an elastoplastic model with isotropic and kinematic hardening. The determined material parameters for the two models, the respective objective function values and the CPU time required to perform the simulations are presented and discussed.

Blank optimization in a stamping process—Influence of the geometry definition

“…The FE analysis was performed with Abaqus [23], while the optimization process was carried out with the Sdl optimization software [18,26,27]. The communication between these two programs was ensured by a FORTRAN interface, where the new geometry is parameterized and then the relevant input variables written in the finite element input file.…”

“…Table 1 NURBS algorithm [19]. The success of the optimization procedure strongly depends on the correct definition of the objective function, which turns out to be essential to the efficient determination of the optimization variables effectively minimizing the objective function [26,28]. One of the most used objective functions consists of the sum of squares and for this specific problem can be defined as…”

Inspection of two sophisticated models for sand based on generalized plasticity: Monotonic loading and Monte Carlo analysis

“…Furthermore, the parametric study of each model, based on simple Monte Carlo simulations, is performed to identify the influence of group of parameters on the simulation of element tests and to detect possible interaction between parameters. In recent years, different methods have been developed to identify the relative contribution of parameters in mathematical models, such as gradient algorithms, numerical methods, knowledgebased systems or artificial intelligence algorithms [e.g., [37][38][39][40][41][42] ]. Among these methods, parameter sensitivity analyzes help identify the relative importance of model parameters, including global and local sensitivity analyzes [e.g., [43][44][45][46] ].…”

On the Single-Point Calculation of Stress–Strain Data under Large Deformations with Stress and Mixed Control

“…In addition, a developed complex constitutive model needs to be tested with experimental results or microscopic calculations, before it is regarded as a reliable model. Thus, it is not trivial to have a simple calculation framework to test advanced constitutive models, debug the program for models, and facilitate parameter calibration [ 7 ]. In addition, a potential application of the sophisticated constitutive models is to generate microscopic calculation data, for macroscopic model-free data-driven mechanics [ 8 ] or for establishing macroscopic data-driven machine learning models [ 9 , 10 , 11 , 12 , 13 ].…”