JK7070
Structural
- Oct 16, 2011
- 20
Hi all,
I have an issue with convergence when modelling a steel brace member with two lines of symmetry and using shell elements (S4R).
I am trying to compare the cyclic hysteresis response of the brace to existing test data. I am using a symmetric tensile/compressive displacement controlled input. This input increases every three cycles until failure of the brace. As I am not modelling fracture explicitly I need to monitor the strain results in an area of interest until they reach the strain at which fracture occurred in the experiments. My issue is as follows:
In order to obtain accurate strain readings I need to refine my mesh in the area of interest (where local buckling occurs). It is upon refinement of my mesh that the model breaks down before I am able to complete a similar number of cycles as in the experiments.
The frustrating thing is that if I use a course mesh I can complete more than enough cycles which match up quite well to the experiments except after the point when the brace should fracture.
I have tried numerous things to fix the problem without any success. These include reducing the time step, further refinements with the mesh, changing material properties. Nothing works once I have such a refined mesh.
Does anyone have any ideas regarding possible reasons for non-convergence. It might also be worth mentioning that the model always breaks down during a compressive cycle just after the brace has buckled. I am also applying a lateral geometric imperfection to initiate buckling of the brace and using large deflection effects.
Regards,
JK7070
I have an issue with convergence when modelling a steel brace member with two lines of symmetry and using shell elements (S4R).
I am trying to compare the cyclic hysteresis response of the brace to existing test data. I am using a symmetric tensile/compressive displacement controlled input. This input increases every three cycles until failure of the brace. As I am not modelling fracture explicitly I need to monitor the strain results in an area of interest until they reach the strain at which fracture occurred in the experiments. My issue is as follows:
In order to obtain accurate strain readings I need to refine my mesh in the area of interest (where local buckling occurs). It is upon refinement of my mesh that the model breaks down before I am able to complete a similar number of cycles as in the experiments.
The frustrating thing is that if I use a course mesh I can complete more than enough cycles which match up quite well to the experiments except after the point when the brace should fracture.
I have tried numerous things to fix the problem without any success. These include reducing the time step, further refinements with the mesh, changing material properties. Nothing works once I have such a refined mesh.
Does anyone have any ideas regarding possible reasons for non-convergence. It might also be worth mentioning that the model always breaks down during a compressive cycle just after the brace has buckled. I am also applying a lateral geometric imperfection to initiate buckling of the brace and using large deflection effects.
Regards,
JK7070