SHD Sim Studies
One run answers a question. A sweep answers a design.
No new solver, no new mesh path, no new results reader. The whole family is a loop over the existing case model, which is why it is the cheapest capability on the roadmap and the one that changes how the product is used.
This module is available.
What it runs on
Established open-source solvers, bundled or provisioned — the same approach as the CFD module, where the case for paying is the hours the interface saves rather than a solver nobody else has.
Native
Sweeps, DoE and the run matrix. Needs no external tool and delivers most of the value
NLopt / pagmo
Derivative-free optimisation, in-process
Dakota
Uncertainty quantification and calibration, once the native loop exists
Analysis types
5 of them, in 3 groups, each with the backend that serves it. Naming the backend per row is deliberate: it is what tells you whether a capability is a setting away or a different solver entirely.
Sweeps and DoE
Vary values in the case model, run the matrix, tabulate the results. This is the part that needs nothing new at all.
| Analysis type | Backend | Notes |
|---|---|---|
| Parameter sweep | native | Vary case-model values, run the matrix, tabulate results |
| Design of experiments | native or Dakota | Latin hypercube, factorial |
Optimisation
The same sweep machinery, driven by an algorithm choosing the next point instead of a grid.
| Analysis type | Backend | Notes |
|---|---|---|
| Optimisation (derivative-free) | NLopt / pagmo | Genetic, particle swarm, Nelder-Mead over the sweep machinery |
Uncertainty
What the answer is worth, given that the inputs are not exact. The question every reviewer asks and few tools answer.
| Analysis type | Backend | Notes |
|---|---|---|
| Uncertainty quantification | Dakota | Propagation, reliability, Sobol indices |
| Calibration / parameter estimation | Dakota | Fit a model to measured data |
What you would use it for
Written as questions because that is how the work arrives. Nobody sets out to run a modal analysis; they set out to find whether the thing will resonate.
Which of these twelve designs is best?
A parameter sweep runs the matrix and tabulates the result, so the comparison is one table rather than twelve case folders and a spreadsheet.
Which input actually matters?
Design of experiments — Latin hypercube or factorial — separates the variables that move the answer from the ones that do not.
What is the best shape I can get from these knobs?
Derivative-free optimisation drives the same sweep machinery with an algorithm choosing the next point instead of a grid.
How confident should I be in this number?
Uncertainty quantification propagates input distributions to an output distribution, which is the question every reviewer asks.
Why does the model not match the rig?
Calibration fits model parameters to measured data instead of adjusting them until the picture looks right.
Who asks these questions
- Any sector already running CFD or FEA
- Automotive
- Aerospace and defence
- Energy and power
- Turbomachinery
- Consumer goods
- Academia and research
Sectors where this analysis is routine — not a claim that we have customers in them.
How to get it
Add Studies on the pricing page, choose named or floating seats, and the same licence unlocks local solving, paid-tier scale and eligible SHD Cloud workflows for this module.