|
General data: Implicit integration |
|
|
General data: Implicit integration |
The implicit integration scheme may use either a constant or a variable time step algorithm.
The default is to use a constant time step, and in general this is to be preferred. Variable time step schemes can introduce high frequency noise into a system which in turn may lead to noisy time histories, non-physical spikes, and generally inaccurate results. Note that this is not unique to OrcaFlex, but is a feature of all variable time step algorithms. For the majority of systems no problems arise with variable time steps, but, if you use them, then we do recommend that you check the quality of your results.
The variable time step algorithm adapts the time step according to the number of iterations taken for previous time steps. If a large number of iterations were required for previous time steps, then the current time step is reduced. Conversely, if a small number of iterations were required then the time step is increased. The results implicit solver iteration count and implicit solver time step can be used to track the performance of the variable time step algorithm.
If you are using a constant time step, then this data item determines that time step. OrcaFlex has a default value of 0.1s. However, different systems will require shorter time steps and indeed some systems can give perfectly accurate answers with longer time steps. If you wish to optimise run times then you may need to experiment with different values here.
If you are using a variable time step, then this data item limits the time step used by OrcaFlex and it will never exceed this value.
| Warning: | Implicit solvers can produce inaccurate results, even for stable simulations, if the time step is chosen to be too large. Because of this we strongly recommend that you carry out sensitivity studies on your time step to ensure accuracy. A very useful additional check is to compare results from the implicit solver and the explicit solver. |
The implicit integration scheme uses an iterative method to solve the dynamic equilibrium equations. This calculation is abandoned if convergence has not been achieved after this number of iterations.
If you are using a variable time step then this abandonment means that the time step is reduced and the calculation re-attempted, rather than the simulation being aborted. Accordingly, we recommend that a relatively small value is used: the default value is 20. This allows OrcaFlex to abandon the current iteration quite early and try again with a shorter time step.
If, however, you are using a constant time step, then abandonment in this case means that the whole simulation is aborted. Here, therefore, we recommend using a much larger value for the maximum number of iterations. The default value of 100 is usually a good choice.
A non-dimensional value which controls the accuracy of the solution. Increasing this value can result in reduced computation time since fewer iterations are needed to solve the dynamic equilibrium equations. However, doing so may also result in inaccurate results.
| Notes: | We recommend that you use the default tolerance value. Should you choose to increase it then we strongly recommend that you perform sensitivity studies to confirm the accuracy of your results. |
| For systems where the only responses are extremely slowly varying (e.g. drift motions, spar VIM) you might need to use a smaller tolerance than the default value to obtain accurate results. | |
| In addition to the tolerance, the accuracy of the analysis is also affected by the characteristic lengths and forces associated with the objects in the model. |