For ease of readability, the term “model” will be utilized in this article to talk about what are called Discrete Event Simulation or Process Flow Models.
Purpose of Process Flow Models:
1) The most common purpose for a process flow model is to ensure that a new layout, process flow, and/or automation will work as desired. The model, in this case, is used to: A) ensure that the new process will deliver the desired results, B) identify how much of each critical resource is needed to achieve various levels of production demand, and C) identify bottleneck locations and other operational challenges that the new process will generate, and how to best deal with them.
2) The other use for a process flow model is to determine how to improve existing operations. In this case, the model is used to: A) identify the root cause of the process flow problems, and B) test out ideas for improving the operation.
Key Factors to Consider When Using Process Flow Models:
1) Level of Detail in Model:
Choosing the right level of detail to incorporate in a model is a big challenge. For people new to process flow modeling, their instinct is to put in as much detail as possible. This is always a bad idea since: A) it makes the effort for building the model much longer than it needs to be, and B) the resulting model takes too long to execute. You want your model to have just enough detail to answer the questions you are going to have the model resolve. Thus, more time spent up front clarifying the purpose of the model, is more critical than diving right away into the coding of the model.
2) Statistical Accuracy of Model:
Input Distributions: A big advantage for a process flow model over other analytic tools (such as spreadsheets) is that you can reflect the variability that exists in the real world operations, within the model. This is done through the use of statistical distributions to reflect the various process times, resource availability times, and customer/material/order arrival times. A key point here, is to take the time to collect and analyze data so as to select the appropriate statistical distribution for each of the aforementioned times.
Run-Time Statistics: Once you have built the model, but before you start experimenting with it, you need to define the following parameters:
Multiple Replications – is how many data points we need to collect for each experiment,
we run with the model, so as to ensure the accuracy of the results of the model.
Warm-up Period – is the length of time we need to run the model, before we ask the model
to collect results for us. This eliminates/minimizes the bias caused by the initial
operational conditions of the model.
Model Run Length – Since process flow models can execute a week’s worth of “real world”
time in a very short period of real time, it is necessary to identify, based on our objectives,
how long of a period (week, month, year, ???) the model needs to run so as to ensure
accurate useful results
3) Develop Confidence Intervals for Key Measures:
As you run your model, you need to keep in mind that the results provided are like a sample from a population, and the results from the individual replications of the model should be different. With this in mind, to present the most accurate results possible, we need to develop confidence intervals for all the key measures reported by the model.
The above three factors are probably the most significant ones when using process flow models. Additional factors to consider include: A) what software to use, B) do we want 2D or 3D animation, C) should the model have a spreadsheet/Access front end to improve model usability, etc.