How Discrete Event Simulation Needs to Improve
Author: Descreye Solutions
Discrete event simulation has been around for a long time. It's roots go all the way back to the Georges-Louis Leclerc, Comte de Buffon. He used replications of an experiment to identify an approximate value of the constant associated with the "Needle Problem." However, with computer processing capability simulation has become a much more powerful tool at getting insight. Today there are many discrete event simulation software tools that allow people to completely replicate, design, or improve a system. In simulation software like FlexSim there are advanced 3D graphics that make it easy to build, visualize, and validate simulation models. The models that are created in these tools can be used with confidence in supporting improvement conversations. However, even with all these advances simulation still has some growing left to do. The following are some areas where discrete event simulation software must advance in order to reach further into system design and improvement.
Raw Data Translation
An area that simulation must become better at is transforming the raw data from a user into a simulation. Traditionally, the user has been responsible for this data translation.
For example, in most cases the user gets a set of observations or estimates from the real-life system. If done correctly, this data is then fit to a random number generating statistical distribution. This distribution is then used as input in the simulation software. The issue with this method is the reliance on the user, because it necessitates that the user has advanced statistical and mathematical knowledge. This drastically limits the userbase for discrete event simulation software.
Currently discrete event simulation software is reliant on the user for translating the data into a simulation, but as it continues to improve this will likely become a feature of the software itself.
One aspect of simulation that has been thoroughly discussed, but that is still inadequate in current software packages is system integration. In simulation the ideal would be to use a simulation that is integrated with the real-life system. The data would be fed to the simulation in real-time. The simulation could then be used to identify optimal routing, sequence, material planning, etc. for the system. It would become an essential part of system maintenance scheduling, material resource planning, and other critical events. While a few companies have created simulations that have some of these capabilities, they are incredibly complex. The complexity makes them very difficult to build and manage, and, as such, are not likely with the currently available software. The main reason for this is the lack of adaptability in simulation models.
Another aspect of simulation that must be addressed in future development is adaptability. Adaptability is defined as the ability of a simulation model to change when the real-life system changes. This could also be referred to as the maintenance problem of simulations. The maintenance problem is that a simulation model could be built to replicate a real-life system, but any changes to the system would have to be made in the model, as well. Because of the significant time investment in creating a simulation, there is a hope that simulations will be built once and then maintainable for a time. However, most find that the simulation is only good for the current project. This makes using simulation difficult, as the cost and time spent in simulating the system must be recouped in a single project.
There are a few ways that simulation can overcome the maintainability problem. The first is to make models incredibly easy to generate. By decreasing the cost and time involved in creating a model, the payoff becomes worthwhile for more projects. Another way it could be solved is by using system integration to auto-generate the model. By developing model generation APIs the data could be translated in a way that automatically generates the model. This would make updating a model instantaneous. Changes would then only need to be made if the change was so significant that it changed the API. This is currently sometimes done by using databases or excel files to auto-generate a model. This decreases the need to do maintenance on the model as it just requires updating the spreadsheet.
As models become better able to adapt to change, simulation will be able to be used in more and more situations to gain insight and solve problems.
← Back to blog