Solver Options

Activating the command: Setup Model Setup Parameters Advanced Other tab

The information on this page applies to the following analysis types except where indicated:

Mechanical Event Simulation (MES)

Static Stress with Nonlinear Material Models

MES Riks Analysis

The type of solver for a nonlinear analysis can be selected in the Type of solver drop-down box within the Other tab. The choices are as follows:

Automatic: This is the default option. If selected, the processor will choose between the Sparse and Iterative (AMG) solvers based on the model size (that is, on the number of degrees of freedom to solve). The sparse solver is typically optimal for small to mid-sized models. The iterative solver is generally optimal for large models and will require less RAM. Both of these solvers will take advantage of multiple threads/cores. You can specify how many of your available threads/cores that you want to use in the Number of threads/cores drop-down box. See also Types of Solvers Available for background information.
Sparse: Uses one of the sparse solvers available in the Type of sparse solver drop-down menu. The sparse solvers use multiple threads/cores when available (see the Sparse Solver Section topic below).
Iterative (AMG): Uses the Algebraic Multi-Grid iterative scheme to solve the system of equations. This solver uses multiple threads/cores when available. It is the default for models with a large number of degrees of freedom to solve and is recommended for models with thin walls, slender parts, plate/shell or beam elements, multipoint constraints (MPCs), and for ill-conditioned models.
Iterative (AMG-MF): This is a variant of the Iterative (AMG) solver, adapted from Autodesk Moldflow technology. It is best suited for models with relatively thick or fat part volumes and may provide a performance advantage, relative to the Iterative (AMG) solver, for such models. You must manually select this solver; the Automatic option will only use the Iterative (AMG) solver for large models. The Iterative (AMG-MF) solver uses multiple threads/cores when available.
When the Iterative (AMG-MF) solver is selected, an additional option becomes available within the Solution tab of the Analysis Parameters dialog box. Specifically, this solver can take advantage of GPU (Graphics Processing Unit) computing, in which the many small cores of the GPU assist with the computation-intensive task. Activate the option, Use the GPU version of the solver, to take advantage of this capability.

Tip: The sparse solver is recommended when the model contains plate or shell elements, even for large models. Select the sparse solver to prevent automatic selection of the iterative solver for large models.

As listed above, some of the solvers take advantage of multiple threads/cores if available on the computer. The drop-down Number of threads/cores control is enabled in this situation. You want to use all the threads/cores available for the fastest solution, but might choose to use fewer threads/cores if you need some computing power to run other applications at the same time as the analysis.

Iterative Solver Section:

If the iterative solver is chosen, then the Iterative Solver section will be enabled. The input for this section is as follows:

The Convergence tolerance field will determine how accurate of a solution is found to the matrix of equations. The smaller the tolerance, the more accurate the solution.
Maximum number of iterations will stop the analysis if the matrix of equations is not solved within this number of iterations.

Attention: The accuracy of the solution depends on the convergence tolerance; a smaller tolerance will result in a more accurate solution but may take more iterations. As with any iterative solution, the results should be checked to confirm that they meet the appropriate accuracy. In some cases, performing the analysis twice with a different convergence tolerance is the best way to confirm the accuracy.

Sparse Solver Section

If the sparse solver is chosen, then the Sparse Solver section will be enabled. The input for this section is as follows:

The Type of sparse solver drop-down box contains the sparse solvers currently available. If you choose a solver that is not available on an operating system, the processor will use the best one for the operating system and let you know that it was changed. The sparse solvers available are as follows:
- Default: use BCSLIB-EXT
- BCSLIB-EXT: (Windows and Linux) use the Boeing solver. For Windows only, note that the BCSLIB-EXT solver may write temporary files to the folder specified by the environment variable USERPROFILE. By default, this variable is set to the folder C:\Documents and Settings\Username where C: is the drive on which the operating system is installed. The error numbers -701 or -804 returned from the BCSLIB-EXT solver means that it ran out of hard disk space for storing the temporary files. If this occurs, change the USERPROFILE variable to a directory that can provide sufficient hard disk space. (See the Windows Help and Support for documentation on changing environment variables.)
- MUMPS: use the MUMPS distributed solver.
  Note: The distributed solver requires MPI to be installed on the computers. Refer to the MPICH2 on Windows page or the MPI Clusters on Linux page within the Simulation Mechanical Supplement of the Autodesk Installation and Licensing Guide for details.
The Solver memory allocation field sets the amount of memory to use during the sparse matrix solution for the BCSLIB-EXT solver. In general, allocating more memory should result in a faster analysis. Note that the other sparse solvers adjust the memory setting automatically; so no setting is required for them.

Options for Representing the Mass

When working with MES, you can choose between two mass representations in the Mass representation drop-down box: Lumped and Consistent. Lumped will place the mass of the element at the nodes; consistent will distribute the mass of the element throughout the volume. It should be noted that the computation of the effective load vector at each solution step is considerably longer when the consistent mass option is requested. The extra effort involved in a consistent (as opposed to the lumped) mass analysis may not be justified in many situations.