Abstract
Nowadays, finite element (FE) codes are increasingly employed for simulating large deformation problems. Thus, to reliably represent the strain hardening behavior, a proper calibration of constitutive laws is essential. Focusing on tensile tests, the main issue with ductile metals is necking occurrence, because of the consequent triaxiality and non-uniformity of the strain and stress states. Over the past decades many strain hardening identification approaches have been proposed. Among them, FE-based inverse methods are widely used, but computationally expensive and time consuming. Hence, the authors propose an efficient method which exploits a database for relating the plastic flow rule and the specimen necking profile. The explicit solver of the nonlinear FE code LS-DYNA was used to build the database, whose size could be limited thanks to physical considerations. The developed methodology was applied to experimental quasi-static tensile tests performed on different metals. The predicted hardening laws showed good agreement with those identified with FE-based inverse methods, thus verifying the applicability of the proposed strategy. This study paves the way for machine learning tools having as main input the necking shape: indeed, the present work suggests their feasibility and provides insights into how to establish datasets for a proper and efficient training.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have