Since the semi-implicit method uses semi-implicit transient dynamics for part of the solution, every node in the model must have an associated mass. Therefore, each node must belong to at least one element that has material density defined.

Since the stable time increment is dependent upon the element length, material density, and the moduli, choosing the correct units and length scale for the problem is very important. For example, in an implicit static solution scheme with linear elastic material behavior, to model a uniaxial tension by modeling a block having a 1 cm length or a 100 cm length would take similar times and give similar results. However, when the same problem is solved in a semi-implicit solution, the stable time increment for the block of 1 cm length will be 100 times smaller (assuming identical meshes for the 1 cm and 100 cm cases).

To achieve a faster solution, you can scale the mass. Two methods of scaling are available. Specify a desired time increment, and the program automatically scales the mass to make the stable time increment at the first substep of the semi-implicit solution to be more than the desired time increment. Or scale the mass directly by giving a factor to achieve a faster solution. However, increasing mass increases inertia, and with more inertia the solution will be farther away from the quasi-static solution. Monitor the kinetic and potential energies that are printed from the solution to make sure the solution is close to quasi-static. The energies are also printed in the monitor file (file.mntr). A good rule of thumb is Kinetic energy < 1% of the potential energy for quasi-static problems.

If the solution fails during the semi-implicit solution, consider using a smaller factor of safety (for example, using the command SEMIIMPLICIT,SFAC,,0.1). Furthermore, a smaller mass scaling factor will also help to achieve a solution in the semi-implicit phase. However, reducing either of them would result in longer times to solve. It is also recommended that you use a smaller allowable increase in maximum displacement increment (for example. SEMIIMPLICIT,AUTS,DSPL,10) so that the semi-implicit solution bisects and provides a smoother solution.

For problems that include contact elements and fail during the semi-implicit solution, lowering the contact stiffness by a factor of 10 (real constant FKN = 0.1 for the contact elements) can be helpful. Lowering the pinball value may also help if the semi-implicit solution fails due to spurious contact detection.

For problems that include contact elements, a high mass scaling can result in large time increments and large penetration in the substep. The contact force to resist this penetration can sometimes be very high, causing instabilities. For a penalty enforcement of the contact constraint, this can cause the parts to separate and fly away; and for a Lagrange enforcement of the contact constraint, this may result in chattering and a non-smooth contact pressure distribution. Reducing the safety factor and/or mass scaling will improve this.

It is recommended that you specify a small value for the time to be spent in the semi-implicit phase. The intent of the semi-implicit method is to help bypass small periods of extreme nonlinearities where convergence in a pure implicit solution is difficult. However, if the problem has extreme nonlinearities throughout the load history, then it is better to just continue in the semi-implicit solution as each attempt to transition back to implicit is computationally expensive.

It is expected that the results obtained using the semi-implicit method can be different than the results obtained if the problem were to be solved entirely in the implicit solver. The magnitude of the difference depends upon many factors such as, but not limited to, the mass scaling used and thus the inertial loads in the body and the time spent in the semi-implicit solution phase.

You should not save results at every substep of the analysis when using the semi-implicit method. During the semi-implicit solution phase, there can be thousands of substeps. Therefore, saving results at each substep can make the result files extremely large.

If the implicit solution phase is a static analysis, the semi-implicit solution phase uses lumped mass by default. Therefore, if the static analysis has any load (gravity for example) that uses density, and therefore the mass, it is recommended that you explicitly define the mass lumping option you desire.

When transitioning to the semi-implicit solution phase from implicit, if the solution shows large oscillations, it is recommended that you define a mass proportional damping in the semi-implicit phase that will help in damping out the oscillations.

If gasket elements (INTER192 through INTER195) are included in the model, you must use the MP command to define density for these elements. The density will be used to calculate the mass matrix for the gasket elements during the semi-implicit solution phase. The mass matrix is not used for these elements in the implicit solution phase.