Intelligent use of computing power

Simulation technologies already cover an enormous part of real development and testing models. In order for this technology to be used pragmatically and economically, intelligent concepts for adequate calculation models are required despite the tremendous increase in computing power.

Simulations reflect the real world. The closer they can calculate to reality, the more accurate the results. Since the real world and all its framework conditions are relatively complex, corresponding calculation models are often costly. Thus, it is not uncommon for the simulation of flows or structures under certain conditions to take days or even weeks of time. Even though computing power has increased by a factor of 1 million over the last three decades. For a solid but also economical as well as pragmatic calculation, the intelligent definition of the calculation model to be used is therefore crucial.

To reduce computation time intelligently means to reduce the effort of the simulation and thus its costs. In engineering simulation calculations, it is the spatial discretization - i.e. selecting a finite number of (discrete) points on the basis of a large number of possible points (the continuum) - and the temporal discretization that serve as possible adjusting screws. Since computation time is quadratic to the number of points and linear to the time steps, spatial discretization is one of the most important steps.

Model simplifications play an important role. For example, thin structures can be described much more accurately using shell elements. Bar-like structures can be calculated efficiently via beams. In the flow domain, perforated plates can be simplified using porous media. For example, a model with several million volume elements can be described using a model with 500,000 elements. In terms of computing time, this means a reduction by a factor of 10-100. In addition, the meaningful decomposition of computational situations is often advantageous. Thus, depending on the task, one can use a 3D overall model, the 3D simulation of a section or a corresponding modification with the appropriate framework conditions for the calculation. If used intelligently, the same results can be achieved using all of the above methods. Whereby the computation times can differ strikingly. If this is now supplemented by logarithmic and non-linear time behavior, the memory requirement can be dramatically reduced.

"We have enormous computing capacities and performance at our disposal. In this respect, it seems tempting to be wasteful with it. However, in the area of simulation technologies, we still require enormous amounts of memory in many models. Therefore, the engineer has to deal intelligently with the available power in order to develop reasonable optimizations in a timely manner," says Dipl.-Ing. (TU) Stefan Merkle, managing director of Merkle & Partner GbR. "If you use the increased computing power and at the same time give less and less thought to how to reduce the required computing time, you waste resources and the performance is worse than it could be."