Assignment 3 - Comparison of base setup and recommended properties
Responses for questions 1 - 5:
The base simulation has the following shell element properties.
By refering the Hyperworks documentation, one can understand the different default values for each of the shell element flags.
IShell = 0 (This denotes the type of element formulation. 0 means I shell is set to 1. This is a Belytschko element formulation - visco-elastic hourglass modes orthogonal to deformation and rigid modes)
Ismstr = 0 (This is for considering the large deformations happening during the course of analysis. 0 means the flag is set to 2. Which means - full geometric non-linearities with possible small strain formulation activation in RADIOSS Engine
Iplas = 1 (This flag is for the shell plane stress plasticity. 1 means iterative projection with 3 Newton iterations. This is accurate whereas Iplas =2 is faster)
Ithick = 1 (This flag is used for considering the thickness change during the deformation and the resultant stresses calculation)
Changing the run time:
In order to change the run time to 55 ms, T stop card (under the RUN keyword) be changed. Or the third line of the engine file (ending with 0001.rad) can be changed to 55.
Change the number of animation steps:
To have a minium number of animation steps between 25 and 60, the T freq card should be between 0.9167(=55/60) and 2.2(=55/25). The T start is set to 0.
I have chosen a value of TSOP = 2, arbitrarily.
For the selected shell element properties, the results seem to be acceptable but not so great. As per the guidelines provided in the Radioss Ebook, they have mentioned that the results where the energy error is less than 15% could be accepted. Based on this, I have to oblige accepting the results. But I would rather change the shell material properties to reduce the energy error as much as possible. In this case, the energy error was about -10.3%
Link to model:
https://drive.google.com/open?id=1KQFEBSKkCNDLNvh8S5L0PlxgFPpH1S5v
Energy error snapshot:
Plot of energies:
Responses for questions 6 - 9:
The recommended properties of
I shell = 24, I smstr = 2, Ish3n = 2, N = 5, I thick = 1 and I plas = 1 are used in this simulation.
It was observed that the energy error has been drastically reduced to a mere 0.1%. The model uses I sheel flag value of 24 which means improved under integrated elements (QEPH) with physical hourglass stabilization are used with 1 integration point on the surface.
These QEPH elements took approximately 15% more computational time than the I shell = 1 element but the accuracy has increased drastically. (When compared from the energy error perspective)
The deformation looks realistic compared to the base model.
Link to model:
https://drive.google.com/open?id=1zt9rEPyAaCeqLWdBL8Abt9FzPOcC78uu
Energy error snapshot:
Plot of energies:
It can also be observed that the total energy is constant from the above plot - which should usually be the case. The kinetic energy has to decrease obviously as the moving object\'s velocity keeps reducing. This deformation happeing due to the crash results in strainwhich in turn results in strain energy which in turn results in an increase in the internal energy.
Overall observations:
I shell = 24 seems to have the least energy error and at the same time computationally less resource intensive.
I shell = 24 also does not have hourglassing energy.
Use I plas = 1 for accurate predictions.
Always consider large strain with full geometric nonlinearities, i.e. I smstr = 2 if you are not sure about the extent of deformations.
If I smstr =2, the strain is computed based on the current element shape instead of the initial element shape. This is important in crash simulations.
And use N=5 , i.e., the recommended number of intergations points in the thickness direction.
Other models - Just for understanding:
I just played with the properties and have captured the energy errors and the energy plots for those models. This was just done for a better understanding.
I shell = 24 I smstr = 1:
I shell = 24 I smstr = 2 I plas =2: