All services
Surrogate or reduced-order model
Build a fast surrogate or reduced-order model of your simulation, validated against first principles.
From $3,000 for a typical scope
Discuss this service“Our simulation is too slow”
A rigorous simulation can be too slow for optimisation, sensitivity studies or what-if work. A surrogate or reduced-order model (ROM) reproduces its behaviour across a defined range and answers almost instantly. It stays an approximation, so we check it against first principles and write down its limits.
A good fit when
- You have a working simulation you trust
- You need many runs for optimisation or sensitivity work
- Each run takes too long to explore the question
How the work runs
Timing is a planning estimate, confirmed after the first stage.
| When | Work | You can review |
|---|---|---|
| Week 1 | Define inputs, outputs and range | Agreed specification |
| Weeks 2–3 | Design and run sampling of the full simulation | Training and test runs |
| Weeks 3–5 | Fit the surrogate or ROM, then test it on unseen runs and against conservation laws | Error map and validation report |
| Week 6 | Package the model and hand over | Model with documentation |
What we need
- Access to the simulation, or a set of runs
- The inputs and outputs that matter
- The operating range of interest
- Any existing validation against the plant
What you keep
- Surrogate model and source code
- Error map against the full simulation
- First-principles validation report, including mass and energy balance checks
- Documented valid range and out-of-range checks
Agreed before we start
- Maximum acceptable error on key outputs
- Maximum mass and energy balance error
- Required speed per evaluation
- Coverage of the agreed operating range
When we’d stop
A surrogate can only be as good as the simulation behind it. If the full model doesn’t match the plant, we recommend fixing that first.