Stacked-shell approach - Low bending stiffness

Hi, I am a student of the Aeronautical Engineering Master's Degree course of Politecnico di Milano and I am working on my Master's Thesis. In particular, my work's objective is to predict delamination and failure of composite structures subjected to bird strike events. In order to do so, I am modelling composite panels adopting the stacked-shell approach by using LS-DYNA.

 

During my work, I found some difficulties in correctly represent the bending stiffness of the laminates. At the moment, I believe that this difference between the numerical model and the real specimen depends on the AUTOMATIC_SURFACE_TO_SURFACE_TIEBREAK card used in the analyses. 

After several tests, under suggestion of my supervisors, I am writing to kindly ask you for support.

In order to clearly present the problem, I am briefly illustrating the work path followed below.

 

At first, a material characterization campaign was carried out for the composites which I am now simulating and MAT058 was chosen to represent such materials. The second step was to simulate the characterization tests in LS-DYNA by using an equivalent thin shell model. This step aimed to verify the quality of the material cards developed before moving to the more complex stacked-shell approach.

The results show that numerical curves mimic the experimental ones with a high degree of fidelity.  Then, the stacked-shell approach was chosen to obtain the CONTACT_GAP output, which is related to the delamination phenomenon.  Several models were developed varying the material, the number of plies, the number of shells (from 2 to 4) and the contact parameters.

Instead, in the compression and 3-point bending test simulations, there is a clear problem due to the reduced bending stiffness. Indeed, the compression tests show a quite good initial behaviour of the numerical specimen under compressive load, but then an early failure due to buckling is obtained. Hence, this highlights that the critical load of the numerical model is lower than the real specimen's one. Moreover, the 3-point bending test simulations exhibit unphysical deformation and reduced level of stress with respect to the experimental curves. In both cases the results make worse with the increasing number of contact interfaces.

Then, the same analyses were repeated with SOFT=1 and SOFT=2.  For the first choice, the sensitivity with respect to SOFSLC parameter (declared relevant in terms of contact stiffness) was studied. However, this study does not show appreciable variations of the results.

By choosing SOFT=2, on the other hand, high numerical instability is obtained. Indeed, for the 2-shell and 3-shell models the instability causes the sudden deletion of all the elements. A qualitatively correct deformation is instead obtained for the 4-shell model, nevertheless, once the simulation outputs have been extracted, it is clear that the laminate does not transmit the load and therefore is deformed without exhibiting consistent reactions.

 

Eventually, option 11 was considered. Starting from literature energy release rate values for fracture opening modes I and II, the parameters ERATEN and ERATES were chosen. Also in this case, the d3plot file shows a "plastic" deformation of the composite laminate, making clear its low bending stiffness. Additional parametric studies were conducted for the parameter CT2CN and CN with no appreciable variations in terms of flexural stiffness.

Therefore, the difficulty found in correctly simulating the behaviour of the composite specimens does not allow to obtain reliable impact test simulations.