Use of Simulation Modeling
Modeling is a method of solving problems, in which the system under study is replaced by a simple object that describes the real system and/or its behavior and is called a model.
Simulation is used when conducting experiments on a real system would be impossible or impractical: for example, because of the high cost of prototyping and testing, or because the fragility of the system will not support extensive tests, or because of the duration of the experiment in real time is impractical.
Bits, not atoms. We need to distinguish between physical and mathematical modeling. An example of a physical model is a scale copy of an airplane in a wind tunnel. Simulations are a special class of computer-based mathematical models whose behavior is dictated by equations and algorithms, typically based on data, and represented by some type of computer user interface. These models mimic the behavior of some real-world system and develop theoretical outputs based on varying input data. This allows the simulation user to examine complex behavior and scenarios on a wide range of conditions far more quickly and inexpensively than with physical systems.
Example of a Simulation Model
How to apply simulation can be illustrated by looking at an example based on a bank’s customer service department. Let’s assume that we’re trying to determine the minimum staffing level needed to reach some established service level.
We develop a Quality of Service measure; average queue size customers cannot exceed N people. In order to solve this problem you need to have some basic knowledge about the system such as how many customers visit the bank, at what frequency, and how long it takes a teller to service an individual customer. And, since we’re trying to model a real world situation, we have to vary the customer arrival time (lunch hour rush) and length of service to account for simple and complex transactions.
This task only seems specialized; it is a one of a general case of similar queuing problems which arise in many fields involving human and technical resource utilization. Companies are always seeking to lower the cost of underutilized resources, whether technical experts or capital equipment. Determining optimal resource schedules and balancing systems costs to its workload reduces costs and increases profits.
The first step in solving this problem is to create a model that corresponds to the structure and business processes of the bank. The model only needs to consider those factors that impact the problem being analyzed. For example, the availability of office services for corporate accounts or credit department has no effect on services to individuals, because they are physically and functionally separate. Schematically, this model can be represented as a sequence of these actions.
The second step is to feed our model some raw data: the variability and peaks of customer arrivals, average customer service length, the amount of available tellers and service personnel. Based on this data, the model simulates or reproduces the work of the bank within a specified period of time.
The next step is to analyze the statistics collected and provided by the model. If the average queue size exceeds the specified N limit, the number of available staff should be increased and a new experiment should be done.
After several experiments the user will have determined optimal staffing for the expected customer profiles. Naturally, this type of experimentation has been automated and the user can quickly vary a wide range of parameter values or optimize for the best solutions rather than conduct multiple repetitive tests.
Advantages of Simulation
Experiments via simulation model have several important advantages versus physical experiments:
- Value. A simulation model can offer a dramatic return on investment. Suppose a company has revenue shortfall and they respond by cutting staff. But this also lowers customer service and response time, causing reduced sales and further revenue attrition. A simulation could be used to balance key parameters such as discounts, process changes, advertising, and many other factors needed to balance this redefined system.
- Time. In the real world evaluating the long-term impact of process or design changes can take months or years. A simulation model will inform your thinking in only minutes.
- Repeatability. Modern life requires organizations to quickly respond to changing market conditions. Analyses such as product demand forecasts have to be prepared quickly yet their results can be critical. A marketing team could use of a simulation model and vary parameters such as price and market segment for an unlimited number of experiments.
- Accuracy. Traditional computational mathematical methods require a high degree of abstraction and do not account for important details. Simulation modeling allows us to describe the structure of the system and its processes in a natural way, without resorting to the use of formulas and strict mathematical relationships.
- Visibility. A simulation model enables the visualization of the system over time; animations illustrate the system in operation and graphical outputs quantify the results. This allows us to visualize the resulting decision and dramatically simplifies the task of bringing these ideas to client and colleagues.
- Versatility. Simulation allows us to solve problems in any area: manufacturing, logistics, finance, health, and many others. In each case, the model simulates real life and allows for a wide range of experiments with no impact on real objects.