Work with us
Simulation engineering
Computer modelling of flow, heat, combustion and structures, so a design can be tested before it is built.
Research
Computational lab
How SciHeart checks its simulations: grid independence, benchmark validation and turbulence-model calibration.
What the lab does
The computational lab is where SciHeart's simulation work is set up, solved and checked. Its CFD work uses ANSYS Fluent. A simulation always returns numbers, whether or not they are right. The lab's protocols exist to show that a result does not depend on the mesh, agrees with measurement and uses a turbulence model suited to the flow.
Engineers separate two questions. Verification asks whether the equations are being solved correctly. Validation asks whether they are the right equations for the real device. Grid independence speaks to the first, and benchmark validation to the second.
The protocols are maintained under the technical direction of the Chief Technology Officer. They are meant to support numerical reproducibility: the same case, with the same settings, should give the same answer whoever runs it.
Three protocols
Grid independence verification
The same case is solved on several meshes of increasing fineness. Richardson extrapolation across those solutions estimates the answer on an infinitely fine mesh, and the error that remains. When the result stops changing as the mesh is refined, it reflects the physics and not the grid.
Experimental benchmark validation
Model results are compared with sensor data recorded on real engine and heat-exchanger test rigs, ideally data that were not used to set the model up. Agreement within the combined experimental and numerical uncertainty is the evidence that a model describes the physical device and not only a tidy set of equations.
Turbulence-model calibration
The turbulence model is chosen for the flow, not by habit. The candidates are Realizable k-ε, k-ω SST and the Reynolds Stress Model, and the choice rests on wall detachment and on the curvature of swirling flow. As a rule of thumb, strong swirl calls for the Reynolds Stress Model, and separation at the wall for k-ω SST.
What this means for a client
The point of all three is traceability: a client should be able to ask where a number came from. If you already have measurements from your own engine or heat-exchanger rig, mention them when you get in touch: measured data is what a model is benchmarked against.