Do You Make These Common Discrete-Event Simulation Mistakes?

Date: 9/20/2016
Author: Descreye Solutions

Creating simulation models can be difficult. Many people have spent hours creating models only to scrap them because they weren't getting any useful insight from them. The following 4 mistakes are common for people when they start creating simulation models. Avoiding them is an important part of becoming an experienced discrete-event simulation modeler.

  1. Forget to validate the model

    Many times someone has finished creating a current-state model and jumped right into testing various alternative scenarios. It is then surprising when the changes don't reflect anything close to the expectations. Often this is a result of not validating the original model.

    Validation is the step of simulation model building where the simulation model is tested against the real-life system to confirm that it accurately represents the system. The amount and rigidity of model validation depends on the model that is being created, but some model validation should be done in any situation. Another common mistake with model validation is waiting until the end of model creation to start validating the model. If one waits until the whole system is modeled, then if makes it difficult to identify which steps in the process is not representative of the real-life system. Validating the model during each step of model creation is an easier method. This is done by testing each process as it is added to the model to make sure it performs as expected. Once the model is complete the whole system can then be tested to make sure it is an accurate representation. Whatever the method, model validation is a necessary step to any successful simulation project.

  2. Use an average value

    Simulation thrives off of providing insights into systems that have a significant amount of variability. However, often people use average values as the input in their simulation models. This makes the model less accurate than if the data was fit to a distribution that could be used to sample random numbers. Simulation uses random number sampling distributions so that it can accurately reflect the variabiltity that is found in the real-life system. By accurately reflecting the variability it is then possible to get more accurate insight from the simulation model.

    Usually people use averages for the following reasons:

    1. Don't have the data
    2. If this is the case, then it is necessary to recognize that simulations follow the rule of data analysis; garbage in, garbage out. The accuracy of the model depends on the accuracy of the input data.
    3. Don't know how to fit data to distribution
    4. This is a common issue. There are hundreds of distributions that can be used to sample random numbers, and distribution fitting can be a complex mathmatical problem. Fortunately, most fully functional simulation software packages have a distribution fitting capability that will automatically suggest a distribution given a data set. Additionally, OPS provides the capability to do pseudo-distribution fitting. The user can use a comma-separated list of observations. This list will then be fitted to a histogramic-6 distribution. That distribution will then be used for sampling during the simulation run.

    Through developments in simulation software packages it has become much easier to do random number sampling, which makes using average values much less useful in simulating systems for process improvement.

  3. Model too much

    Trying to model every aspect of the system is a critical error in most simulations. Often people who are just starting to simulate systems try to add every aspect of the system to the model, because they want to see everything run as it does in real-life. This makes the modeling time grow and grow until the project inevitably fails.

    The following are a few questions an experienced simulation modeler will ask to ensure that they don't model too much:

    1. What is the goal of the simulation?
    2. An experienced model takes the goal of the model as the blueprint with which the model is built. If the goal of the model is to test how many resources are needed at one station in the plant, then it often isn't necessary to model the whole plant. In a simulation the inputs will be changed to determine how they affect various outputs. It is possible to narrow the scope of the model by focusing on the inputs and the outputs that will be used.
    3. Can we assume?
    4. Assumptions are crucial to narrowing the scope of a simulation project to a manageable size. Good simulation modelers make accurate and useful assumptions to avoid having to model additional details. Often a model has a process with a resource that is 100% dedicated to that process without breaks or downtime. In those cases there is rarely a need to model the resource, because the process can assume that it will have the resource available. By making assumptions, and documenting those assumptions, a simulation model builder can avoid unnecessary detail.

  4. Model too little

    Sometimes it becomes apparent that the simulation model doesn't have the details that are necessary for the intended analysis. This problem often becomes apparent during a meeting after the simulation is completed when someone semi-involved in the project asks, "What if we were to change x, how would that influence this scenario." If x wasn't in the original model or was an underlying assumption, then it is often difficult to add that into the model. However, the method for making sure that the model isn't too simple is to review the same questions that were asked for modeling too much. Those questions should be reviewed with every stakeholder in the project before the model is started in order to ensure that all the necessary detail will be included.

Back to blog

© Descreye Solutions, 2022