Abstract The mechanical characterization of thermoplastics submitted to finite strain is a non-trivial procedure since they may present necking and cold-drawing. These phenomena are associated with heterogeneous strain fields that may mask the real stress–strain curve and lead to mistaken conclusions about the capability of constitutive models to represent the material mechanical response. Aiming to shed light on this issue, this paper presents a numerical study based on a Finite Element Method Updating (FEMU) technique to obtain the true stress–strain curve of a thermoplastic specimen. FEMU technique is employed to characterize three elastoplastic models with different mathematical frameworks. Inverse problems are constructed considering force response and displacement data of a heterogeneous strain field on the specimen due to the necking and neck propagation kinematics. Results shown that even a very simple multilinear elastoplastic model, available in commercial software, is capable of representing force response of a tensile test when only experimental force data is considered. However, when the necking kinematics is taken into account, the studied models present different results and poor mechanical behavior representation, indicating that only the force data obtained from a tensile test is not enough to establish conclusions about the performance of a constitutive model when heterogeneous strain field occurs. Since none of the models was capable of accurately reproduce the experimental force and kinematics data, to evaluate which model performs better, it is proposed a multi-objective analysis of Pareto front results.

中文翻译：

---

有限应变下热塑性材料本构特征的数值研究

摘要 承受有限应变的热塑性塑料的机械特性是一个重要的过程，因为它们可能会出现颈缩和冷拔。这些现象与异质应变场有关，这些应变场可能会掩盖真实的应力-应变曲线，并导致关于本构模型表示材料机械响应的能力的错误结论。为了阐明这个问题，本文提出了基于有限元方法更新 (FEMU) 技术的数值研究，以获得热塑性试样的真实应力 - 应变曲线。FEMU 技术用于表征具有不同数学框架的三种弹塑性模型。由于颈缩和颈部传播运动学，考虑到试样上异质应变场的力响应和位移数据，构造了反问题。结果表明，即使是商业软件中提供的非常简单的多线性弹塑性模型，也能够在仅考虑实验力数据时表示拉伸试验的力响应。然而，当考虑颈缩运动学时，所研究的模型呈现出不同的结果和较差的力学行为表示，表明仅从拉伸试验中获得的力数据不足以得出非均质应变时本构模型性能的结论场发生。由于没有一个模型能够准确地再现实验力和运动学数据，