Dear Team,
I am aware that this issue has been raised by others in the links attached below, but I wish to share some of the methods that I have tried, hoping to solve the problem but to no avail.
https://communities.bentley.com/products/geotech-analysis/f/plaxis-soilvision-forum/205690/warning---very-small-embedded-beam-elements-generated
https://communities.bentley.com/products/geotech-analysis/f/plaxis-soilvision-forum/213574/warning-small-embedded-beam-elements-generated-messages-in-considerable-long-embedded-beam-piles
As shown in the figure below, I am currently modelling a piled-raft with 846 nos. of 500mm dia. spun piles, mostly spaced at 1.5m to each other. A bunch of warnings appear after meshing, and the calculation stops at Phase 1 due to incorrect solution and slow convergence.
Thinking that it may be due to low-quality meshes at the pile top and pile toe, I have created two volumes that envelope the pile top and the pile toe, and locally refine the mesh using the coarseness factor. The mesh could not be generated in this case.
Next, I have created volumes to envelope each and every pile. While this has significantly reduced the number of warning messages that I am getting, it creates a huge number of elements, which cause the entire model to be computationally expensive to run.
My last resort would be to model the piles as soil volumes with the material properties of concrete, but this is similar to the above and it will take forever to run as well.
My preference is to continue modelling the piles as embedded beams to keep the number of elements low and hence a shorter calculation time. Therefore, I wish to seek your advice on what else can be done to prevent me from getting Error code 204 (Picos solver: incorrect solution, slow convergence)?
Do let me know if you require further information on the model.
Thank you.
Regards,
Jia Sheng Lee
Your last resort may or may not solve your convergence problem, before you go that route I have a couple of suggestions you might try.
First, try computing a null step before you activate any piles. This will ease computation in the pile installation phase because the calculation will start from a solved equilibrium state.
Next try installing 1/2 the piles in one phase, and then add another phase to install the remaining piles. This will also ease the calculation by not presenting so many load bearing elements for calculation simultaneously.
Keep an eye on the global error, and global stiffness and if these are not erratic you may be able to make some progress (or at least speed up the calculation) by increasing the global error tolerance. I'm not sure if the Plaxis tech folks would agree with me, but I have a hunch that errors accumulate in small zones (like the very small embedded piles) with redundant structural elements (like closely spaced piles), or tight restricted spots where the soils can't shed load (corners and slots). All those little errors add up in big model, and may not influence the results you're interested in. For these types of problems I've found that increasing the error tolerance to 0.05 helps a lot. This is especially true when there are lots of interface elements which are in a plastic state.
I wonder if when there are a huge number of closely placed piles if a single unique solution even exits, I suspect there may be a huge number of possible force combinations in the piles which would be in equilibrium.
Martin
Dear Jia Sheng Lee,
The convergence issue is a complex calculation problem that needs some investigation. I see that Martin already gave quite a few pro tips on checking what can help continue with the analysis.
As I have explained in the most recent forum post, the meshing warnings about the embedded beam elements can be ignored as it is merely a message indicating that the soil elements are intersecting with the embedded beam elements. These specific meshing warnings should not create a big issue with convergence.
After trying a few things, if you still cannot understand why you get no convergence (or getting a divergence), feel free to reach out to our support team: https://apps.bentley.com/srmanager/ProductSupport