Date: 10/27/2016
Author: Descreye Solutions
Most simulation models are created manually, because their goal is to get insight into a specific problem. However, sometimes it becomes necessary to create a discrete-event simulation model programmatically. This could be for any of the following reasons:
In many cases these reasons make it worthwhile to make the effort to create the means whereby a simulation model can be created programmatically.
To create a simulation model automatically there are a few process steps. This image shows the various steps in model creation that must be defined to programmatically create and run a model.
When first creating a simulation programmatically it is often useful to first create the model manually to better understand how each step is accomplished manually before it is done through code. The following is a description of the importance of each step in the process:
The raw data definition is a critical part of creating a model programmatically. The raw data must contain all the information that is needed to completely build the model. For example, if the user was creating a production line model, then the raw data might need to contain the cycle times, flow, process step capacities, resource requirements, arrival times, etc. The raw data definition step is to make sure that all the data that is used in the model can be found in the raw data. Additionally, if any GUIs will be used for user input of various characteristics, then this would also be defined in the raw data definition.
The data translator plays the important role of transforming the raw data into data that the simulation software can use to build the model. For example, the raw data might have observations on the cycle time of a machine in the system. The data translator might fit those observations into a sampling distribution that the simulation software can use to estimate the variation occurring in that machine’s cycle time. The data translator step can be found within the simulation software or in an external program. Often Excel or Access are used to create the data translator, because they can create connections to external data sources. They also have analytical capabilities for summarizing the data in a way the simulation software can use it. Also, many simulation software packages can import the data from these programs.
The model creator is the step in the automatic simulation creation that is used to take the translated data and create the model within the simulation software. This step is usually done within the simulation software. Usually the software will import the translated data, if it wasn’t translated within the software, and then run a loop through the various items to create the objects and flow that is necessary for the model. Once this step is complete, then a executable simulation model should be built.
This is the result of the programmatic simulation builder that is made from the previous steps. Some automatic simulation building will only go to this step where the model is built. The user is then expected to run the experiments and export the experiment data to get insight. In some cases, the model will be built with GUIs that show the user some critical information. In those cases, it is basically a compressed version of the next couple steps.
In some cases the programming with include the actual experimentation of the scenarios that are created in the model. For instance, the model might be run for 100 replications to determine how it compares to a baseline that has been created previously.
When the modeler wants to make the analysis completely automated, then they would include a way to write the experiment data back to a file that the user can use to gain insight.
If the complete solution is created, then it could look like this:
Each of these steps will be discussed in detail in future blog posts.