A multiphysics simulation or coupled simulation is always necessary if there is a dependency between different, normally separate physical fields and this must be taken into account accordingly. The term multiphysics simulation is a generic term for all simulations that meet this criterion. Which subject areas or which simulation methods are linked to each other is initially not important for a classification. From a numerical point of view, how this simulation task is implemented is much more important for differentiation.
Integrated and separated coupling
The term integrated multidisciplinary simulation is used when the various physical disciplines are solved with a single calculation model in one simulation run. On the other hand, there is the separate multiphysics simulation, in which several calculation models with different solvers contribute their share to the overall solution. The individual simulations must exchange result variables with each other in order to establish the relationship to each other. How often this exchange of information must take place depends on the type and strength of the coupling. If the dependence is given only in one direction, one speaks of a unilateral interaction. In this case it is only necessary to transfer the results once. If, on the other hand, there are interactions, this is a bi- or multilateral coupling. The stronger the interaction, the more often the individual systems of equations have to be solved and the results exchanged.
Pros and cons
At first glance, therefore, the integrated coupling appears to be a much simpler and more efficient approach, provided, of course, that the solver can and the existing license allows the disciplines to be combined. But even if this is the case, there are many applications in technical practice in which equally good or better results can be achieved in a shorter time with a separated multiphysics simulation.
There are many reasons for a separate multidisciplinary simulation. On the one hand, there is the significantly greater flexibility in modeling. In this way, for example, a static analysis can be coupled with a transient analysis, or different design states can be virtually married to each other. On the other hand, it can make sense to carry out the individual simulations with different time steps. In addition, there are often dissimilar meshing requirements of the various disciplines. A better result is achieved if it can be addressed in each case and does not have to be worked with a compromise. An additional aspect in favour of a separate approach is the sometimes unequal demands on hardware equipment and configuration. Furthermore, separated models that are processed sequentially or on different computers in parallel require fewer hardware resources. Instead of working on one large system, it is possible to work on several small systems. Finally, the calculation engineer can actively intervene in the solution finding process, which can considerably improve the numerical stability and thus the solution search. The disadvantage of all this is the effort involved in ensuring the exchange of information. If the whole thing is to run efficiently, an automated process is necessary, especially in the case of a strong bilateral coupling.