Dear Team,
I am modelling the diagonal compression of masonry assemblage, I am getting an error related to matrix solver, which I am not able to understand. Any Suggestions will be appreciated,
Regards,
Zeeshan
Error in Analysis
-
Zeeshanbhat
- Posts: 153
- Joined: Wed May 27, 2020 3:48 pm
Error in Analysis
- Attachments
-
- Capture.JPG (66.27 KiB) Viewed 1958 times
-
- ist.rar
- (12.1 MiB) Downloaded 189 times
Re: Error in Analysis
The error is saying that the linear solver is unable to do the symbolic analysis of your matrix. This happens either when your model is singular, or when your model is too large to fit a single processor.
Your model is fine. But the mesh is too small, in fact you have more than 1 million nodes.
Keep in mind that when you have such a large model, your system matrix, even if stored in a sparse matrix storage, may need indices (for the matrix entries) that can read 1 or 2 billions, or even more.
The umfpack (and other sequential solvers) have some limitations due to the fact that an integer (4 bytes) can only represent numbers up to 2 Billions.
You can either use your fine mesh with muti processors, or make the mesh coarser (which works fine anyway with a micromodel).
Furthermore I noticed that in your cad, to contiguous rows have the same masorny pattern (probably a mistake?).
Here is a working copy of your model (your original one works ok if you reduce the number of nodes)
Your model is fine. But the mesh is too small, in fact you have more than 1 million nodes.
Keep in mind that when you have such a large model, your system matrix, even if stored in a sparse matrix storage, may need indices (for the matrix entries) that can read 1 or 2 billions, or even more.
The umfpack (and other sequential solvers) have some limitations due to the fact that an integer (4 bytes) can only represent numbers up to 2 Billions.
You can either use your fine mesh with muti processors, or make the mesh coarser (which works fine anyway with a micromodel).
Furthermore I noticed that in your cad, to contiguous rows have the same masorny pattern (probably a mistake?).
Here is a working copy of your model (your original one works ok if you reduce the number of nodes)
-
Zeeshanbhat
- Posts: 153
- Joined: Wed May 27, 2020 3:48 pm
Re: Error in Analysis
Dear Team,
Thank you so much for helping me. I have few small queries.
1. How have you created these extra partitions (Figure attached) in my model?. Is there any particular command to do such partitions, because for me I would have created the whole model from beginning.
2. Which type of global seed have you used in this model?
3. Can we use more than one processor in case of openssesmp. because it gets enabled only when we select openseessp in STKO Solver command.
Regards,
Zeeshan
Thank you so much for helping me. I have few small queries.
1. How have you created these extra partitions (Figure attached) in my model?. Is there any particular command to do such partitions, because for me I would have created the whole model from beginning.
2. Which type of global seed have you used in this model?
3. Can we use more than one processor in case of openssesmp. because it gets enabled only when we select openseessp in STKO Solver command.
Regards,
Zeeshan
- Attachments
-
- Capture121.PNG (74.99 KiB) Viewed 1943 times
Re: Error in Analysis
The lines you highlighted in the picture are not geometric features, they are just the mesh edges. If you move from the Mesh context, you won't see them.1. How have you created these extra partitions (Figure attached) in my model?. Is there any particular command to do such partitions, because for me I would have created the whole model from beginning.
I used a global seed of 50 mm, if I remember properly. I think it is fine enough because with micro-modeling you don't really need very fine meshes.2. Which type of global seed have you used in this model?
The number of processors is enabled only with OpenSeesSP, because OpenSeesSP does the partition itself, based on the number of processors specified by the user.3. Can we use more than one processor in case of openssesmp. because it gets enabled only when we select openseessp in STKO Solver command.
On the other hand, if you want to use OpenSeesMP, after you do your mesh, you need to create the partition in STKO, using the partition command in the Mesh Tab. That's why you don't need to specify the number of processes, because STKO will take it from the number of partitions.
For example, if you want to run your simulation with 4 processes, you need to partition your mesh in 4 partitions.
Also note that with OpenSeesMP you need to use the MUMPS algorithm and the Parallel RCM numberer, otherwise it won't work