Model rescue
When a simulation model needs help
A lot of our work starts the same way: someone inherits a model they don't understand, or a model that takes too long to be useful, or a problem that has outgrown the tool it was built in. We have been fixing all three for more than 35 years.
- You inherited it. The modeler moved on, and no one can say what the model assumes or whether to trust it.
- It's too slow. A run takes hours, so it never makes it into the meeting where the decision is made.
- It outgrew the tool. The data, the logic or the number of users is more than the package was built for.
1 · Inherited
You inherited a model you don't understand
The person who built it left, the documentation is thin, and the results are being used anyway. We read the model itself, whether it's ExtendSim, Plant Simulation or another tool, and tell you plainly what it does.
- A logic map: what each part of the model represents, and the rules it follows.
- A data map: where every rate, failure and schedule comes from, and which ones are guesses. Often the data is best moved into tables (SDI DB) where people can see it.
- A validation check: the model's results against what the line or network actually did.
- A recommendation: keep it, fix it, migrate it or retire it, with the reasons.
What you get back
A model you can defend in a meeting, documentation your next modeler can pick up, and a trained owner on your team. No black box.
2 · Too slow
Your model takes too long to be useful
A model that runs for hours gets run once, and a decision needs dozens of scenarios. We find where the time goes and take it out:
- Model what matters at the right level. High-speed and bulk sections run far faster as rates than as individual items (discrete rate).
- Move data out of the logic. Table-driven models load and change faster than hand-set parameters.
- Run replications in parallel, and cut the warm-up and reporting that nobody uses.
- Use a faster engine where one fits, such as ReliaSim® for line reliability studies.
Example
6 hours to 20 minutes
VINLogic, a vehicle logistics model, took 6 hours per run, too slow for any business decision. Rebuilt, it ran in 20 minutes, so analysts could use it every week, and executives used it during Hurricane Katrina.
Read the VINLogic case3 · Outgrown
The problem outgrew the tool
Commercial simulation packages are excellent until the problem gets bigger than they were designed for: terabytes of live data, logic the blocks can't express, run times measured in hours, or hundreds of people who need answers without learning the tool.
When that happens we build a bespoke model: the same logic, validated against the original, written in .NET, C# or F# and wrapped in an app people can run themselves. It is still simulation, just no longer limited by a package.
Before you rebuild
A bespoke model isn't always the answer. Sometimes a custom library inside the tool you already own, or a move to a different tool, gets you there for less. We tell you which, and why.
Tell us your problem
Have a model that needs help?
Tell us which tool it's in, how long it takes to run and what it's supposed to answer. We'll tell you whether to read it, speed it up or rebuild it.
- A straight answer on whether simulation is the right tool
- Which tool fits, even if it isn't one we use every day
- A rough scope and timeline, before any commitment