Buchholz Martin - Modeling and Simulation An Application-Oriented Introduction

(1)

(2)

(3)

The notion of simulation is quite ambiguous and requires clarification. In the context of this book, two of its interpretations are of particular relevance. In a broader sense, simulation is the complete process of the forecasting or replication of a certain scenario. Since such simulations are nowadays performed almost exclusively computer-based, we will notas it is oftentimes seen elsewhererefer to it as computer simulation. In a tighter sense (and in the title of this book), simulation only refers to the central part of this process, i.e., the actual computationa classical case of a pars pro toto. In the following, we will make use of both interpretations and only provide explicit clarification if the respective interpretation is not given implicitly.

In a broader scope, therefore, simulations are nothing other than virtual experiments on the computer. This remains unchanged by the fact that in most application areas served by simulation (for example, physics, chemistry or mechanics), the respective representatives of the computational guild are typically allocated to the theoreticians.

The attractiveness of these virtual experiments is obvious. For a multitude of cases, real experiments are simply impossible due to the underlying time and spatial scales, for example. To illustrate, one needs only to consider astrophysics: no matter how hard-working, it is impossible for any physicist to devote the necessary billions of years at the telescope to study the life cycle of a galaxy; or geophysicsit may be possible to create experimental earthquakes, i.e., artificial earthquakes in a James Bond production, but those are not practical in real life. Moreover, not all that is possible in principle is actually desirableone only needs to consider the testing of nuclear weapons, animal experiments, or genetic engineering. The former took their leave just at the time when the respective nations reached the ability to execute them in a completely virtual manner on the computer. The ethical componentnuclear bombs do not become friendlier if they are brought to perfection through simulationsmust not be left out here, but as well will not be discussed further. And even in the remaining set of the feasible and justifiable, the effort is often the limiting factor: The static of buildings, the vulnerability of the HIV virus, the evacuation of a fully filled soccer stadium, economical or military strategies, etc. etc.all these are not tested quickly, not even in the lab; not to mention the effort that fundamental experiments require in modern physics in the context of the Large Hadron Collider. Thus, there is no way to go without simulation and it is therefore worthwhile to take a closer look at its methodology. However, it is indisputable: Simulations complement theoretical analyses and experiments, they do not replace them.

The goals pursued by a simulation can be very diverse. Oftentimes, one wants to reconstruct a scenario which is well-known in principle in order to better understand it. This applies for example to catastrophies of a technical or natural kind. Why has an earthquake developed, why at this particular place, why at this particular instant in time? Why did one of the large traffic bridges across the Mississippi River in the US state of Minnesota collapse in August 2007? How could the tsunami in south-east Asia in late December 2004 develop such a devastating effect? The goal to predict unknown scenarios is also knowledge driven, but in general even more challenging. This applies not only to the catastrophies mentioned above (and for possible repetitions, resp.) as well as to urgent questions concerning climate change or the propagation of the world population, but also to many technical questions (properties of new alloys or composite materials). Besides discovery, another goal pertains to improvement, i.e., the optimization of a known scenario. Prominent examples include the (route) scheduling of airlines, the efficiency factor of chemical reactors and the efficiency of heat exchangers or the data throughput in a computer network.

Here, a simulation in the broader sense is not an intergral act, but rather a highly complex process consisting of a sequence of several steps which are traversed several times in various feedback loops.

To this end, the picture of a simulation pipeline has been established (see Fig.). We summarize the essential steps: