Application |
PLAXIS 2D PLAXIS 3D |

Version |
3DFoundation PLAXIS 2D |

Date created |
18 July 2011 |

Date modified |
18 July 2011 |

The iterative parameters *Desired minimum* and *Desired maximum* are primarily meant to determine when the calculation should take larger steps or smaller steps. If the calculation can solve a load step (hence converge) in fewer iterations than the *Desired minimum*, it starts using a load step that is twice as big. If, however, the calculation needs more iterations than the *Desired maximum* to converge, the calculation will decide to choose a calculation step only half the size.

For a Plastic analysis or Safety analysis, there is no influence on the results when changing the *Desired minimum* or *Desired maximum*. As long as the calculation converges every step it is unimportant if the calculation uses a lot of small steps with few iterations or a limited amount of larger steps with more iterations per step. For a Dynamics calculation, the step size is constant, determined by the time interval and the number of additional steps and dynamic substeps, and so *Desired minimum* and *Desired maximum* are not even used.

For a consolidation analysis, however, there is a small influence. As mentioned, the choice for taking larger steps or smaller steps is based on convergence of the calculation. In a consolidation analysis during a load step (hence during a time step) water flows out of the soil, and generally, the flow is slightly higher at the beginning of the step than at the end of the step. However, in one time step, the calculation will calculate with a constant flow, and typically that is the flow at the end of the time step. This means that on average the flow rate during a time step is slightly underestimated, leading to a slightly higher consolidation time. If the calculation takes larger steps this effect is bigger than when the calculation takes smaller steps.

In practice, when a consolidation analysis takes several hundreds of steps, the influence on the consolidation time will be very small. However, for very specific cases where consolidation analysis only takes a few steps, for instance when modelling laboratory tests on high permeable soils, there may be an influence of several percent. In that case, lowering the Desired minimum and Desired maximum in order to force the calculation to use smaller steps will improve the results.

A more common reason to change the Desired minimum and Desired maximum in a consolidation analysis is the calculation time. In a consolidation analysis, the stiffness matrix depends on the time step and so the time step must be know before composing the stiffness matrix. In case the calculation decides the step size can be scaled up or should be scaled down, the time step changes and a new stiffness matrix must be composed. Occasionally it happens that every other calculation step the calculation decides to scale up, and then the next step scale down again. This means that for almost any calculation step a new stiffness matrix has to be made, which is very time-consuming. In this case, slightly increasing the *Desired minimum* and *Desired maximum* can cause the calculation to change step size less frequent, which may give a significant improvement to the calculation speed.

In general, when decreasing the Desired Minimum will cause less scaling up, and increasing the Desired Maximum will reduce the number of downscales.

**Manually controlling load stepping**

A special option to control the load advancement scaling (size of the step) exists for setting these desired minimum and maximum values to specific values. When the *Desired minimum* is set to 2 and the *Desired maximum* to the maximum iterations, there will be no scaling up or down and you will see a constant load step for each step.

Note that with arc-length control switched on, the actual load step may be smaller. For consolidation analyses, the arc-length control is switched off always, but for plastic calculations with arc-length control switched on, the load steps may differ in size a bit.