I am carrying out dynamic analysis of reinforced earth wall in FLAC 2D. Every time 'bad geometry;impossible to continue' message is coming even after i gave 'set large' command. distortion is happening at the base of the panel (where panel beam meets the foundation soil). What could be the possible reason?
Hi Anusree K V,
As you have seen, the bad geometry error occurs when and element (FLAC zone) becomes distorted beyond the default limit internal to the FLAC code. My understanding is that this is the ratio of the area of one of the 4 smaller triangular subzones within the element to the area of the quadrilateral element (FLAC zone). This ratio is calculated every step.
You can set this ratio to a higher limit if you want by using "set geom ##" command, where the ## in the quotes is a ratio input by the user, like 0.25 for example, but that wouldn't help you in this case. A higher ratio means that the simulation would stop slightly earlier than a lower ratio, or the default. I don't recall the default ratio at the moment, but it is lower than 0.25.
Using the set large command won't stop the bad geometry from occurring.
What is the nature of the distortion? If it is considered an "element blow-up" where the FLAC zone is just wildly and unrealistically "blowing up" into the air, then you may have an issue with the water bulk modulus value during loading relative to if you are using the config gwflow command. This is pretty obvious when it occurs, but you can look at the velocity vectors plot of the grid to help see this problem.
If the distortion is just due to severe straining, then you will have to use the "rezone" function that is internal to FLAC to re-create the mesh during the analysis and rezone the zones, which will replace the distorted elements with ones that have a "good" geometry. "Good" geometry, meaning something that looks like what you started with, more square-like and less parallelogram-like. There are a few problems with using the rezone command, one is that you as the user have to write the code to remesh the geometry, which can be tricky. The mapping of properties from the old mesh to the new mesh is what Itasaca's rezone function actually does, it doesn't recreate the mesh, you have to do that.
One other alternative that I don't recommend is to move the grid points "manually" by using the Alter Shape tool in giic. This will probably call into question the integrity of the results more than using Itasca's rezone functionality. I think there is something in the FLAC manuals stating that this is to not be performed; however, many practitioners have done it before.
The last option that I can think of is to just "call it good" when the bad geometry ratio occurs. Meaning, when it occurs that is the end of the simulation whether it is in equilibrium (complete) or not. You can get the information from that point and analyze as necessary, hoping it is good enough. It's not the best option either, but the realities of using FLAC sometimes dictate the results.
Navead
http://web.mst.edu/~norbert/ge5471/Assignments/Assign%201%20-%20FLAC%20I/FLAC%20manual/FLAC%20manual/f2d408.pdf
http://web.mst.edu/~norbert/ge5471/Assignments/Assign%201%20-%20FLAC%20I/FLAC%20manual/FLAC%20manual/f2d408.pdf
Large strain option does not prevent bad geometry (zero or near to zero area). Your main goal is to identify why bad geometry it's happening. Additionally to the rezoning procedure, you can try reducing the time step (remember that the explicit integration of FLAC makes it conditionally stable), or updating geometry terms more often to make calculation "more Lagrangian" (I think that the command is something like set geom_rep n and the default n is 10)