Method
Every assignment follows the same four-stage sequence. Each stage leaves a trace in the final report, so that the path from question to result can be reconstructed.
01Framing
The technical question is framed so that the answer can be checked, and the acceptance criteria and scope of the analysis are fixed. This stage also establishes whether detailed simulation is needed or a simplified model is enough.
02Modelling
The computational domain is built, the physical sub-models are selected and the boundary conditions are imposed, each with its source. The simplifications adopted are listed one by one, with their justification and expected effect on the result.
03Verification
Grid independence study with an estimate of the discretisation error, mass and energy balance checks, monitoring of the quantities of interest, and sensitivity analysis on the parameters that govern the outcome. These checks precede the delivery of every result.
04Delivery
The report separates results from their interpretation, compares computed quantities against acceptance criteria while stating the available margin, and closes with the limits of validity: the conditions under which the results hold and the changes that would require them to be revisited.
01Technical summary
Question, available data, scenarios, assumptions and acceptance criteria. Shared before the calculation starts, so that the assumptions are agreed.
02Results
Calculation outcomes with grid and sensitivity checks. Discussed before they become final.
03Final report
The complete document described below, with the summary for decision-makers.
Code verification and validation
A solver is credible for a class of phenomena, within a given parameter range, on the basis of published verification and validation documentation. For every assignment the code, its version and the validation references relevant to the phenomenon studied are stated.
Where reference experimental data exist, the model is validated against them before being applied to the real case. Where they are missing, the report says so.
Uncertainty and sensitivity
Three sources of uncertainty are distinguished and treated separately: input data, which depend on what is known about the real system; the model, which depends on the equations chosen; and the numerical solution, which depends on the discretisation. Keeping them apart makes it possible to estimate how much each one weighs on the result.
Sensitivity analysis separates a robust result, one that holds as the parameters vary within their plausible range, from one that depends on a particular combination of assumptions.
What we deliver
The technical report
A layout shared by every report: stated assumptions, input data with their sources, a dedicated chapter on sensitivity and uncertainty, limits of validity before the signature.
The decision brief
A short document answering the question in terms of available margin and consequences, written for decision-makers without a modelling background: a plant manager, a regulator, a judge.
The model files
Where the assignment provides for it, the complete model with its input files and the documentation needed to reproduce the results.
The visual output
Images and animations accompany the numerical results and are read by the same standards: scale, units and time are always stated.