Simulation Applications in Engineering Education

Simulation refers to a methodology to build usually complex mathematical or logical models to imitate the operations of real-life systems. These simulation models can be static or dynamic, deterministic or stochastic, and discrete or continuous, as argued by Averill Law.

Simulation can be used for several different purposes: For example, it is a technique that enables one to evaluate functions when mathematical forms of these functions are not available. Furthermore, simulation can be used to investigate different scenarios through the what-if analysis; that is, change one or more data in the simulation model and analyze the effects of these changes on the performance measures. Moreover, simulation can be used to find better working conditions for systems, which satisfy some limitations on the resources (i.e., optimization). However, simulation is time consuming. A further disadvantage of simulation is that if the models are stochastic, the performance measures are only estimated through some sophisticated statistical techniques up to certain measures of accuracy (i.e., variances). Therefore, analytical techniques are preferable to simulation, if they are available. Jack Kleijnen discusses sophisticated statistical tools for simulation and the various uses of simulation. This entry first discusses different types of simulation models and why simulation is important in engineering education. It then gives examples of simulation games used in engineering education.

An example of dynamic, stochastic, and discrete models is a queuing model. As an example of a queuing model, one can think of the simulation model for a supermarket, where customers arrive and do their shopping. For such systems, it is usually assumed that as soon as the customers finish shopping, they join the queue of a cashier and wait for their turn. Whenever the cashier finishes serving the customer currently in service, that customer leaves the supermarket, and in the simulation language, the cashier becomes idle. Then, instantaneously, the service of the first customer in queue, if there are any, starts, and the cashier becomes busy again; this is what is called the first-in-first-out queue discipline. Depending on the studied systems, the next customer to be served can be the last one in the queue, the one with the shortest processing time in the queue, and so forth. One could use a what-if analysis, changing the first-in-first-out queue discipline to last-in-first-out, and analyze its effects on the average delay in the queue.

For the supermarket model, one usually does not know the times between the arrival of successive customers or their shopping times; in fact, one can only observe and collect certain realizations of these times. Therefore, the interarrival and the shopping times form the stochastic components of the supermarket simulation model. Furthermore, building simulation models requires the definition of some state variables, which are the variables that describe the system at a particular point in time. For the supermarket model, if one is interested in estimating the expected delay time in queue and the expected utilizations of all cashiers, then possible [Page 645]state variables consist of the number of customers in the queue and indicators showing whether the cashiers are idle or busy at a particular time. Obviously, these state variables change dynamically in time, and the changes occur at discrete points in time by the occurrences of two types of events, namely the arrival and the departure of a customer. This finishes the description of one type of simulation, which is known as discrete-event dynamic simulation in the academic literature.

...