Simulation

Simulation, from the Latin simulare, means to "fake" or to "replicate." The Concise Oxford Dictionary of Current English defines simulation as a "means to imitate conditions of (situation etc.) with a model, for convenience or training." Sheldon Ross of the University of California, Berkeley, states less formally that "computer simulations let us analyze complicated systems that can't be analyzed mathematically. With an accurate computer model we can make changes and see how they affect a system." Simulation involves designing and building a model of a system and carrying out experiments to determine how the real system works, how the system can be improved, and how future changes will affect the system (called "what if" scenarios). Computer simulations of systems are effective when performing actual experimentation is expensive, dangerous, or impossible.

One of the principal benefits of using simulation to model a real-world system is that someone can begin with a simple approximation of the process and gradually refine it as his or her understanding of the system improves. This stepwise refinement enables good approximations of complex systems relatively quickly. Also, as refinements are added, the simulation results become more accurate.

The oldest form of simulation is the physical modeling of smaller, larger, or exact-scale replicas. Scaled-down (smaller) replicas include simulations of chemical plants and river–estuary systems. Scaled-up (larger) replicas include systems such as crystal and gene structures. Exact-scale replicas include an aircraft cockpit used for pilot training or a space shuttle simulator to train astronauts.

Simulation is central to the rise of digital computers, and the story starts, strangely enough, with the pipe organ. American inventor Edwin Link (1904–1981) received his inspiration for the first pilot training simulator while working for his father's piano and organ company in the 1930s. Link developed mechanical "trainers" that used a pneumatic system to simulate the movement of aircraft. During World War II (1939–1945), the Link Trainer proved the training value of flight simulation and convinced the U.S. Navy to ask that the Massachusetts Institute of Technology (MIT) develop a computer that would power a general-purpose flight simulator. This endeavor became Project Whirlwind and "evolved into the first real-time, general purpose digital computer… [which] made several important contributions in areas as diverse as computer graphics, time-sharing, digital communications, and ferrite-core memories," according to Thomas Hughes in his book Funding a Revolution.

In a society of limited resources and rapid technological change, training challenges are increasingly being addressed by the use of simulation-based training devices. Economic analysis supports the use of simulators as a sound investment, a flexible resource that provides a return for many years. Since the early 1960s, simulation has been one of many methods used to aid strategic decision-making in business and industry. As computer technology progresses and the cost of simulation for realistic training continues to decline, it is becoming increasingly possible to train simultaneously at different geographic locations and where training cannot be carried out in real life, as in shutting down a nuclear power plant after an earthquake.

Simulations can be classified as being discrete or analog. Discrete event simulation builds a software model to observe the time-based behavior of a system at discrete time intervals or after discrete events in time. For example, customers arrive at a bank at discrete intervals. Between two consecutive time intervals or events, nothing can occur. When the number of time intervals or events is finite, that simulation is called a "discrete event." The discrete event simulation software can be a high-level general-purpose programming language (C or C) or a specialized event/data driven application (a simulator).

However, in the real world, events can occur at any time, not just at discrete intervals. For example, the water level in a reservoir with given inflow and outflow may change all the time, and the level may be specified to an infinite number of decimal places. In such cases, continuous, or analog, simulation is more appropriate, although discrete event simulation could be used as an approximation. Some systems are neither completely discrete nor completely analog, resulting in the need for combined discrete–analog simulation.

A common way to simulate the random occurrence of events is to use Monte Carlo simulation. It is named for Monte Carlo, Monaco, where the primary attractions are casinos containing games of chance exhibiting random behavior, such as roulette wheels, dice, and slot machines. The random behavior in such games of chance is similar to how Monte Carlo simulation selects variable values at random to simulate a model. For example, when someone rolls a die, she knows that a 1, 2, 3, 4, 5, or 6 will come up, but she does not know which number will occur for any particular roll. This is the same as the variables used in computer simulations; these variables have a known range of values but an uncertain value for any particular time or event. A Monte Carlo simulation of a specific model randomly generates values for uncertain input variables over and over again (called "trials") in order to produce output results with statistical certainty, that is, providing a percentage chance that an actual output from the physical system will fall within the predicted range with virtual certainty.

Modeling is both an art and a science. It is an art to decide which features of the physical object need to be included in an abstract mathematical model. Any model must capture what is important and discard interesting features (uninteresting and irrelevant features are easy to discard). Complexity and processing performance also guide the art of deciding on a minimal set of features to be modeled from the physical object.

The science of simulation is the quantitative description of the relationship between features being modeled. These relationships dictate a model's transformation from one state to another state over time. Often when a mathematical model is eventually derived in a solvable form (closed form), it may or may not accurately represent the physical system. Computer simulation is preferable when the physical system cannot be mathematically modeled because of the complexity of variables and interacting components. Well-known examples of simulation are flight simulators and business games. However, there are a large number of potential areas for simulation, including service industries, transportation, environmental forecasting, entertainment, and manufacturing factories. For example, if a company wishes to build a new production line, the line can first be simulated to assess feasibility and efficiency.