On the prediction of material failure in LS-DYNA®: A comparison between GISSMO and DIEM

As a consequence of the worldwide tendency in reducing CO2 emissions by producing lighter and more energy-efficient products, the demand for accurate predictions regarding material behavior and material failure has greatly increased in recent years. In particular in the automotive industry, there is also an increasing interest in effectively closing the gap between forming and crash, since the forming operations may highly affect the crashworthiness of the produced parts. In this scenario, a correct depiction of material mechanical degradation and material fracture seems indispensable. Currently, there are several models implemented in LS-DYNA which have been developed to deal with material damage and failure. Many of them are complete constitutive models which consider elasto-plasticity coupled with damage formulations as well as with embedded failure criteria (e.g., *MAT_015, *MAT_052, *MAT_081, *MAT_104, *MAT_120, *MAT_153, among others). Alternatively, LS-DYNA also makes possible the definition of failure and damage through the keyword *MAT_ADD_EROSION, where the user can choose different failure models and fracture criteria which are, in turn, coupled with the selected plasticity model in an ad-hoc fashion. In this context, GISSMO (Generalized Incremental Stress-State dependent Damage Model) and DIEM (Damage Initiation and Evolution Model) are good candidates for the task of predicting ductile failure using LS-DYNA. However, many users still seem to have difficulties in using these models, meanwhile other users, who already master either GISSMO or DIEM, feel somewhat insecure in employing the concurrent model. These difficulties arise mainly because GISSMO and DIEM have been conceived following quite different interpretations of the phenomena that influence failure. For instance, in GISSMO the user has to input a failure curve as a function of the triaxiality (and also of the Lode parameter, in the case of solid elements) where this curve is used for the nonlinear accumulation of damage. This strategy intrinsically takes the strain path change into account, for which a numerical calibration based on experimental data is required. Furthermore, an instability curve may also be defined in GISSMO, where in this case, if instability achieves a critical value, the stresses are assumed to be coupled with damage, leading to a ductile dissipation of energy upon fracture. DIEM, on the other hand, allows the user to define multiple damage initiation indicators which evolve simultaneously. For example, the user can define a normal and a shear failure initiation criterion, the former as a function of triaxiality, the latter depending on the so called shear stress function. Additionally, a forming limit curve (FLC) can also be input in DIEM, where this criterion also evolves along the other two failure initiation criteria. The different damage initiation criteria can then be combined in a global damage evolution rule. Similarly to GISSMO, a certain number of experiments is required in order to properly fit the parameters necessary for DIEM. This contribution is an attempt to compare and better understand the differences between GISSMO and DIEM. In this respect, the main differences between both models and how they are intended to predict failure will be comprehensively discussed. Additionally, the calibration of a dual-phase steel using GISSMO and DIEM will be used to better highlight the differences between the models and how these are reflected in the final parameter fitting.