1,138 checks, and what each one is worth
The validation report inside the app lists everything the software checks its own results against, and names each source so you can go and look at it yourself. It is not written by hand: every number in it comes from running the test suite on the build you are using.
Kept apart, because they are not equal
Comparing against an international standard is a stronger claim than agreeing with another program. A report that added them together would be hiding the difference, so this one keeps them in separate rows.
| Strength of evidence | Checks | Result |
|---|---|---|
| Analytic Answers you can verify by hand: closed-form results, balances that must close, a result that must agree with itself. | 293 | PASS |
| Published reference data IAPWS-IF97 for water and steam, NIST, NASA-7 coefficients, published worked examples. | 116 | PASS |
| Experimental unit-operation data Measurements taken on real equipment. | — | NOT RUN |
| Third-party simulator comparison The same case in DWSIM, COCO or Aspen HYSYS. This shows agreement, not correctness. | 147 | PASS |
Release 1.6.2: 1,138 checks and 218,489 individual comparisons, all passing. Many checks are internal contracts, such as saving and reopening a case or undo, and sit in none of the evidence rows; a few sit in two.
What it is checked against
Water and steam
Properties of water and steam against IAPWS-IF97, the international formulation engineers and regulators treat as definitive.
The component database
Pure-component properties against reference equations of state, and ideal-gas heat capacities against NASA-7 published coefficients.
A non-ideal azeotrope
Water–ethanol at 1 atm and 1.5 bar against CoolProp’s reference mixture model and the literature: composition within 0.008 and temperature within 0.7 K.
A published absorber
Ammonia absorbed from air into water: the absorber design exercise from The ChemSep Book, after Seader and Henley, reproduced end to end.
Columns, units and whole flowsheets
Mass and energy balances that must close, column profiles stage by stage, and whole cases that must solve to the same answer flat or folded into blocks.
Other simulators’ cases
Cases from DWSIM, COCO and Aspen HYSYS imported and solved, every stream compared with the source’s converged answer.
Why the experimental row is empty
Nothing in this software has yet been checked against measurements from real equipment. The row is shown rather than hidden, because a gap you can see is worth more than a gap you cannot. Filling it is what the validation campaign is for.
What the report does not claim
A check proves the software reproduces a reference, not that the reference is right, and a passing suite says nothing about whether your model represents your plant. Where coverage is light, the report says so by name.
How it stays honest
Every test file declares what it is measured against, or an audit refuses it. The report refuses to generate if it would leave out a test that ran. And it names the build it came from.
Read the full report in the app. Every download includes it. After installing AI Process Simulator, open it and choose Validations > Validation Cases to read the report inside the application, Validations > Open Full Report in Browser for a printable copy, or Validations > How the Validation Report Works for a guide to reading it.
Read the report in the app
Every download of version 1.6.2 includes it: Validations > Validation Cases.