The Accuracies Travelled Back. The Warning Did Not. K-Site, 1992–2025

What this note is: a reading of primary source documents. Every factual claim about a specific document is traceable to a passage in that document, and those documents are listed in Section A of the sources. (“Audit” here means documentary audit, as defined in Note 000.)

Sources are separated by how they were used. Section A lists documents I obtained and read for this note; where a value or quotation was verified against the page images rather than extracted text, the entry says so. Section B lists material cited at second hand or from general knowledge, which I have not read in the original. That distinction is the subject of this note, so it would be poor form to blur it in the note itself.

What it is not: a claim that any model is wrong, or that any engineer was careless. The reading suggests close to the opposite.

I. A report from 1992

In January 1992, three engineers at NASA Lewis Research Center and Analex Corporation presented the results of a self-pressurization test campaign on a 4.89 m³ flightweight liquid hydrogen tank — AIAA-92-0818, issued in parallel as NASA TM-105411. The tank was oblate spheroidal, 14.0 m² of surface area, 149 kg of structure, wrapped in high-performance multilayer insulation. It was held at two declared wall heat fluxes — 2.0 and 3.5 W/m², produced by a shroud at 294 and 350 K — and tested at 29 % and 49 % fill, complementing earlier tests at 83 %. Those flux values carry a caveat recovered by this audit’s dependency reading (Note 000, record A1d): they are total tank heat divided by internal area, with 13–17 % of that heat entering through discrete struts, plumbing and instrumentation penetrations rather than uniformly through the insulation — a distribution that matters for a stratification experiment, which the 1992 paper itself flags only in a one-line parenthesis (the ~85 %-uniform statement); the 13–17 % figure and the path enumeration live in Stochl–Knoll.

This is the K-Site dataset. The name comes from the test facility, and the paper itself uses it exactly once — thanking “the K-Site personnel” in the acknowledgements. It remains a live validation referent in the computational literature: four of the five documents examined below validate against it, and the fifth validates against the Multipurpose Hydrogen Test Bed at Marshall. I have not surveyed the field systematically enough to claim these are the foundations of the validation literature, and I do not.

The report states its instrument accuracies plainly, in a single paragraph that begins on page 2 and runs onto page 3. What follows is a compressed restatement, not a verbatim quotation — and because the ± symbols do not survive text extraction from this scanned document, every value below was verified against the page images of the report, not against its extracted text:

Liquid and vapor temperatures: ± 0.3 K, improved to ± 0.1 K by in-situ calibration against known saturation conditions. Wall temperatures: ± 0.6 K. Tank pressure: ± 0.01 kPa. Capacitance level probe: ± 1.9 cm, a maximum of ± 1.5 % fill at the 50 % level. Boil-off flow: ± 0.030 and ± 0.089 SCMH on the two meters.

That is a careful instrumentation statement. It is more than many contemporary papers provide.

Then the report does something better. On page 7, it reports a closure check that did not close:

“Liquid mass completely dominates the mass balance; absolute errors were less than 4 percent… The energy balance is also dominated by the liquid side and is most sensitive to the liquid temperature measurements… Energy balance errors were from 120 to 150 percent, increasing with fill level.”

An energy balance error of 120–150 % is not a rounding problem. And the authors did not leave it as an anomaly. They bounded the instrument explanation with their own numbers — insufficient, not excluded:

“…an exact energy balance would require a liquid level error of 2 to 3.5 times the measurement uncertainty. Therefore, the uniform radial temperature assumption must be questioned. Specifically, it appears that measured liquid temperatures are higher than a bulk temperature which would satisfy the energy balance requirement, thus indicating regions of lower liquid temperature somewhere away from the measurement locations.”

There is a detail in the report that makes this diagnosis concrete: the liquid and vapor temperature measurements “were limited to a single vertical axis near the tank centerline.” A single vertical rake cannot see radial nonuniformity. The authors understood exactly what their instrumentation could and could not resolve, and said so.

Read that carefully, because it is the most important sentence in this note.

The authors are telling the reader — in their own hedged wording, “it appears that” — that the liquid temperatures in their own dataset read higher than a bulk temperature would, that the evidence points to colder liquid away from the sensor locations, and that any analysis assuming radial uniformity should be questioned. The hedge is theirs and it should be preserved; the observation is unambiguous even so.

This is exemplary experimental practice. Quantified instruments. An honest closure failure. An explicit, physically motivated diagnosis pointing at unmeasured thermal stratification rather than at the equipment. The information a downstream user would most need is present, in the primary document, stated by the people who ran the test.

II. What a validation claim requires

A validation claim is a comparison between a simulated value S and a reference value D obtained from an experiment. The comparison error is:

E = S − D

That number, alone, means nothing. Its interpretation depends on the uncertainty attached to each side. The standard decomposition — used in this series — following the concept that ASME V&V 20, the verification-and-validation standard for computational fluid dynamics and heat transfer, formalizes (per its published scope, the accuracy inferred from a solution–measurement comparison is quantified from the uncertainties of both the simulation and the experimental data; the standard itself was not read for this note, see Sources B) — separates three contributions, written here in this series’ notation:

u_val² = u_num² + u_input² + u_D²

where u_num is the numerical (discretization and iteration) uncertainty of the simulation, u_input the uncertainty in parameters and boundary conditions, and u_D the experimental uncertainty of the referent. The quadrature form presumes independent contributions: where the same experimental data inform both the model’s boundary conditions and the compared result — the usual case in this literature — u_input and u_D are coupled and must be treated jointly. Assembling u_val is analysis, not lookup.

The practical consequence is blunt. If the observed simulation–experiment difference is 3 % while the applicable validation-uncertainty envelope — the combination above, dominated in this example by an experimental pressure uncertainty of ±10 % — is substantially larger, that comparison does not establish 3 % predictive accuracy. It establishes agreement finer than the comparison can discriminate, which is a different and weaker statement.

There is a second, subtler requirement. Validation compares a model to a referent — and the referent is not the raw sensor reading. It is the reading plus everything the experimenters knew about what it represented. In the K-Site case, the referent for a liquid temperature comparison is not “±0.1 K.” It is “±0.1 K, at this sensor location, in a tank the original authors state was not radially uniform, with colder liquid somewhere the sensors did not reach.”

These are two different problems, and the distinction matters. The instrument accuracy contributes to u_D — the uncertainty of the measurement as a measurement. The interpretive caveat is something else, and arguably worse: it says that the quantity measured — temperature at sensor locations on a single vertical rake — is not the quantity a bulk-energy comparison needs. That is not measurement uncertainty; it is a mismatch between the measured variable and the validation variable, and no error bar on the sensor repairs it. A thermometer can be perfectly accurate while the referent remains inadequate for the quantity being validated. Both pieces of information must travel with the data. The first is easy to carry forward. The second is where things go wrong.

III. What travelled — and what travelled back

I examined five computational documents that validate cryogenic tank models against these datasets. Three entered this note when it was first drafted; the two full-length versions of the 2025 work entered at v1.0, for a reason recorded honestly below.

Document	Type	Dataset	Referent instrument accuracies	Closure failure / single-axis caveat
Kartuzova, Kassemi, Agui & Moder (2015)	Conference paper, 19 pp.	MHTB	0	—
Yang, Patel & Williams (2022)	Conference paper, 19 pp.	K-Site	0	Absent
Kartuzova & Kassemi (2025)	Extended abstract, 7 pp.	K-Site	0 (weight-limited: an abstract cannot carry an uncertainty analysis)	Absent
Kartuzova, Kassemi & Hauser, AIAA SciTech (Jan 2025)	Full conference paper, 15 pp.	K-Site	Four of five restored, verbatim	Absent
Kartuzova, Kassemi & Hauser, Cryogenics 152 (Dec 2025)	Journal article, version of record	K-Site	Four of five restored	Absent

Method, stated precisely: the full extracted text of each document was searched for statements of instrument accuracy, measurement error, transducer accuracy, experimental uncertainty, and for the ± symbol applied to a measured quantity; the last column was searched as energy balance, closure, 120 to 150, single vertical, rake, centerline. Per-document counts and procedures are recorded in the verification log published with the companion dataset (doi:10.5281/zenodo.21895803).

The 2022 paper does discuss accuracy — the word appears, in one form or another, more than twenty times across its nineteen pages. In every instance it refers to the accuracy of the computational method: the ability to predict, the interface reconstruction, the gradient evaluation, the time-accurate resolution, the interfacial thermal layer. Not once to the accuracy of the data it is validating against.

How the two 2025 papers entered this note. The first deposited draft examined only the seven-page extended abstract and said plainly that the final paper had not been read; the verification log recorded the conference paper as paywalled, and a journal version as “not found … likely nonexistent.” Both records were wrong. The full conference paper had been publicly downloadable on NTRS the whole time (NTRS 20240016283, 15 pp.; SHA-256 2801c5695f2253d69d0f0be41b65fa9febf8f83ff01004b63642602f59fdef6d), and the journal article exists, open access, as Cryogenics 152 (2025) 104210 — existence verified against Crossref, text read in the version of record. Note 000, Erratum 16 records both failures. They are, uncomfortably and usefully, this note’s own thesis applied to itself: a negative claim about the literature is a timestamped search result, and it expires.

What the full-length 2025 papers carry. Four of K-Site’s five published instrument-accuracy categories, essentially verbatim — the ±1.9 cm capacitance probe (±1.5 % fill at the 50 % level), ±0.3 K fluid temperatures, ±0.6 K wall temperatures, ±0.01 kPa tank pressure: the quantities their comparisons use directly. The fifth category, the boil-off flowmeters (±0.030 and ±0.089 SCMH), does not reappear. The Stochl–Knoll thermal-performance report — the boundary-condition dependency this audit had to excavate — is cited as the source of the measured heat loads, and the journal version goes further: localized heat leaks through instrumentation penetrations are studied expressly, in the paper’s own highlights. The boundary-condition context whose loss Section I describes has, in this team’s newest work, travelled partially back: the Stochl–Knoll dependency is restored and the penetration heat leak is studied — but the base heat load remains, in the papers’ own words, “uniformly distributed along the tank wall”, and the recovered penetration heat is “applied uniformly across the top surface of the tank lid”. A recovered citation is not a reconstructed boundary condition: the original spatial information — 13–17 % of the heat entering through discrete penetrations located in both the upper and lower halves of the tank — has still not travelled whole: the 2025 recovery places the added heat at the lid alone, and the report’s own statement that the uniform-flux assumption is reasonable at the ~85 % level travels with neither. (The stronger spatial gloss this series itself first attached to that sentence is corrected in Erratum 25.)

What still has not travelled. Across the four K-Site downstream documents — the fifth document examined, the 2015 MHTB validation paper, cannot carry a K-Site caveat and is excluded from this count: zero occurrences of the 120–150 % energy-balance closure failure; zero occurrences of the single-vertical-axis limitation; and no assembled validation uncertainty — the restored accuracies are declared, not used to bound any comparison. The one sentence that determines what those accuracies are worth — the original authors’ statement that their measured liquid temperatures appear higher than a bulk temperature would be, with colder liquid beyond the sensors’ reach — appears in none of them. The accuracies travelled back. The warning did not.

These five documents are not five independent observations. Four share Kartuzova and Kassemi, and the three 2025 documents are one study’s own lineage — abstract, conference paper, journal article. What this section rests on is two research groups, and the recovery story rests on one group’s newest work. The non-independence of the original set was pointed out to me rather than noticed, and it is disclosed wherever it limits a claim.

On what this does and does not establish. Five documents are not a literature review, and these are competent pieces of work by researchers with far deeper standing in this field than mine; none claims to perform a formal uncertainty analysis, so the absences are not failures against a stated intention. After the 2025 reading, the observation is sharper than context vanishes: context can travel back — the accuracies and the boundary-condition dependency did — and the piece that still does not is the interpretive caveat that qualifies the data themselves. A comparison can now carry four of the five accuracy statements K-Site published and still not know that the temperatures those error bars attach to may not represent the bulk.

IV. A second case: the quantity with no data to compare against

The 2015 MHTB paper contains the same disclaimer five times — three times verbatim, twice in shortened form:

“no experimental data on the interfacial mass transfer was available for comparison.”

Interfacial mass transfer — evaporation and condensation at the liquid–vapor interface — is one of the central coupled mechanisms of self-pressurization: it changes vapor mass while carrying energy across the interface. It is a phenomenon the model exists to represent, and tank pressure is its measured consequence.

So the structure of the validation is: the mechanism is unmeasured, the consequence (tank pressure) is measured, and the model’s interfacial treatment is assessed by whether it reproduces the consequence. Several different physical treatments can reproduce one pressure curve. The paper is transparent about this; it states the limitation five times. But the limitation is a property of the referent, and it must accompany any subsequent use of that dataset that treats it as evidence for the fidelity of a model’s interfacial mass-transfer treatment — a narrower use, such as tank-pressure response under matched conditions, can remain valid with a correspondingly narrower claim.

A third, smaller example from the same family. The K-Site report contains no statement of the hydrogen’s ortho/para composition — a defined keyword search (ortho, para, purity, catalyst, converter) returned zero matches in its extracted text, and manual review of the apparatus and procedure sections found no statement of spin composition. The downstream analyses examined in this series work with parahydrogen properties — the 2025 K-Site journal paper states its parahydrogen property basis (NIST REFPROP) expressly — the SHIIVER final report, three decades later, tabulates parahydrogen property tables as its Appendix I — and the assumption is physically reasonable: the equilibrium composition at 20 K is overwhelmingly para, the ortho→para conversion is exothermic at a magnitude comparable to the latent heat of vaporization, and bulk liquid hydrogen is therefore produced catalytically converted for precisely that reason. (These are standard properties of hydrogen and are stated here from general knowledge, not from a source read for this note — see Section B.)

The assumption is physically well motivated for stored LH₂. It is also not traceable to anything in the experimental record. That distinction — between an assumption that is right and an assumption that is supported — is the whole subject of this note.

V. The counter-example

None of this is a story about old work being sloppy, and the clearest evidence is what modern practice looks like.

SHIIVER — the Structural Heat Intercept, Insulation and Vibration Evaluation Rig, a 4 m diameter tank tested with liquid hydrogen and liquid nitrogen at NASA Glenn between August 2019 and January 2020 — is documented in a 286-page final report, NASA/TP-20205008233, published in August 2021 by Johnson and seven co-authors. It carries 74 numbered tables (an earlier count of 52, corrected in Note 000 Erratum 14, missed two appendix series). Appendices B through G give skirt and vapor-cooling-line temperatures for each test. Appendix H describes the instrumentation and its locations. Appendix I tabulates parahydrogen and nitrogen properties with references, so the property basis of the analysis is explicit rather than inferred; the spin composition of the fluid actually loaded remains, as at K-Site, unstated — a distinction Note 000 §IV.4 records.

A separate 53-page test plan was published, setting sensor uncertainty requirements in a standalone plan (Revision I, partly post-test — Note 002 §II) — better than ±0.1 K on temperatures used for property evaluation, and better than ±0.02 psia on the pressure used for control. Setting uncertainty targets before a campaign, rather than reporting achieved uncertainty afterwards, is a meaningfully different discipline, and the right one.

And in a companion paper, the team reports that calibration of selected heat flux sensors — performed to ASTM C-1130 and C-1774 before installation — showed the published vendor sensitivities were usable at liquid hydrogen temperatures but carried uncertainties on the order of 50 %. The paper explains why: the sensors are thermopiles, the differential voltage generated at 20 K is small, and at low heat flux the uncertainty grows significantly. They publish the number anyway, and argue that the values and trends still agree with other calculation methods (the report’s own wording — it does not claim independence).

That is what good looks like. A 50 % uncertainty, stated plainly, is worth more to a downstream user than a clean-looking number with no uncertainty at all.

The practice improved. The problem this note describes is not located in the past.

VI. Plausible mechanisms by which context stops travelling

If the information exists at the source and the experimentalists are candid, where does it go? The audit establishes the documentary loss; it does not identify its cause. What follows are hypotheses consistent with the observed record, not findings.

Nothing dramatic. The mechanisms are ordinary:

A dataset is cited by name, not by revision. “The K-Site data” is treated as a stable object. In practice it is a set of figures in a document, read by different people, digitized differently, with a body of qualifications in surrounding paragraphs that a plot does not carry.

Page limits are real. Conference formats impose tight page budgets — the computational papers examined here run 7 to 19 pages. Re-deriving another team’s uncertainty statement costs half a page and adds nothing the reviewers demand.

A paper is a snapshot; a caveat needs a maintainer. The 1992 paper was issued in two containers — conference paper AIAA-92-0818 and NASA Technical Memorandum 105411, the same text under two identifiers — and, as far as I can determine, never revised after that. I found no maintained, structured mechanism in this lineage that keeps its limitations attached to the datasets extracted from it, and no one is identifiably responsible for that attachment.

One plausible incentive: publication rewards new modeling contributions more visibly than the republication of inherited limitations. There is professional credit for a new model and little for restating that a thirty-year-old dataset has a documented interpretive caveat.

And the number keeps working. This is the quiet part. The figure still plots. The comparison still runs. The agreement still looks reasonable. Nothing fails. What is lost is invisible, which is why it is lost.

The numbers survive, and — the 2025 papers show — can even travel back. The interpretive caveat, in the record examined here, has not.

VII. Why this matters now rather than in 1992

Two things changed.

The decisions got larger. In-orbit cryogenic propellant transfer has moved from a research topic toward a critical enabling capability: sustained lunar operations depend on transferring and storing cryogens in space over long durations, and NASA maintains an active cryogenic fluid management technology portfolio directed at exactly that problem.

The physics did not get easier. NASA’s SBIR program, in its 2021 Phase I solicitation — subtopic Z10.01, Cryogenic Fluid Management, led by Glenn Research Center — asked for a subgrid film-condensation model to be “validated against experimental data (with a target accuracy of 25%), with emphasis on cryogenic fluid-based condensation data.” That figure is worth sitting with. For that solicited modeling capability, the agency’s stated validation target was 25 % — a scale that says something about where at least one live cryogenic fluid management modeling problem stands. When the target is 25 %, the difference between a referent with a stated uncertainty and one without is not academic — it determines whether you can tell whether you hit the target.

A field with converged models and abundant data can afford loose bookkeeping about its referents. A modeling problem working to a target of this order, on a small number of large-scale experiments accumulated over decades, cannot.

VIII. What would fix it

Not a new experiment, and not a better model. A record that travels.

The proposal is unglamorous: every experimental referent used for validation should carry a structured record alongside the numbers, containing at minimum —

the primary source, by exact document and revision;
what was measured, with units and the qualifiers that fix meaning (absolute or gauge, static or total, at which sensor location);
the stated instrument accuracies;
the interpretive caveats the original authors published — closure failures, quantities known to be unrepresentative, mechanisms that were not measured;
the fluid state, including composition where it affects the physics;
whether the data are tabulated or must be digitized from a figure, and the uncertainty that digitization adds;
an explicit assessment of whether the validation can be reconstructed from the public record alone.

None of this is novel. It is standard provenance practice, applied to a class of object — the experimental referent — that is not represented as a maintained, structured record anywhere in the document lineages examined here. The concepts already exist in NASA-STD-7009, which treats input pedigree and results uncertainty as first-order credibility factors, and in ASME V&V 20, whose quantitative validation assessment cannot be completed without u_D.

In the lineages examined here, what is missing is not the standard. It is a maintained link between the numerical referent and the qualifications attached to it.

IX. What I am doing about it

I am auditing the public experimental evidence base for cryogenic propellant tank self-pressurization and boil-off, against six criteria fixed in advance:

Are the raw data published, or only plotted in figures?
Are the boundary conditions complete — heat flux magnitude and distribution, fill level, initial stratification, ullage composition?
Are instrument accuracies stated, and are the derived quantities’ uncertainties propagated?
Is the geometry fully specified?
Is the fluid state specified — for hydrogen, including the ortho/para composition?
Does the quality information the original authors published travel with the data into current use?

The candidate list — provisional and certainly incomplete — includes K-Site, MHTB, SHIIVER, the zero boil-off and boil-off reduction test series, the Tank Pressure Control Experiment, small-scale LH₂ and LN₂ dewar tests, and liquid nitrogen zero boil-off ground testing. For each, the target is the primary test report, not the papers citing it. That distinction has already proved to be the slow part of the work.

The criteria were fixed before the audit, for a reason. If they were set after seeing the results, they could be chosen to produce an interesting answer. Baselining them first removes that freedom, and it is the same standard the audit applies to everyone else.

That ordering is a claim about process, so it should be checkable rather than asserted. The criteria above were drafted in a method note, exercised unchanged in a three-referent pilot, and frozen before the full audit ran — the same chronology Note 000 §II records; the sequence, the search procedures behind every negative finding, and the SHA-256 hash of every source artifact are recorded in this series’ verification log and candidate list, published with the companion dataset (doi:10.5281/zenodo.21895803), and the artifact hashes in Note 000 §VIII. I would rather a reader audit that trail than take my word for the order.

The audit’s registry — each verdict tied to the document actually audited, with primary sources preferred and obtained for eight of ten referents — is published alongside this note as Technical Note 000, the foundational entry of this series, numbered so because every other note applies its criteria.

X. Corrections and additions wanted

If you worked on any of these test programmes, or have used their data, three things would be genuinely useful:

experiments missing from the list, or entries that do not belong on it;
quality information that exists but never reached the published report — calibration records, an uncertainty analysis cut for length, or simply a recollection of how a quantity was actually measured;
cases where I have this wrong.

The middle category is the most valuable and the least recoverable. Test reports are written under page limits; the knowledge is not always in them, and the people who hold it are still reachable.

Corrections will be credited. Where one changes a conclusion, it will be recorded as having changed it.

On verification. Before publication, every quotation and numerical value drawn from a Section A document was verified against the page images of that document — not only against extracted text, which corrupts symbols like ±, and not against search summaries, which paraphrase. Anything found wrong after publication produces a new version carrying a visible erratum — the superseded version remains preserved and citable — including when the correction weakens a conclusion already stated.

What follows this series

These notes are the documentary groundwork — published first so it can be checked first — for a committed quantitative study: reconstruct a defensible result-level experimental uncertainty (u_D) for one referent in this registry, combine it with the numerical and input uncertainties of a published comparison, covariances included, and report whether that comparison’s validation conclusion moves. In either direction: if nothing moves, that result publishes too. Until that study exists, everything here remains what Note 000 §IX declares — documentary findings whose engineering consequence is argued, not demonstrated. The study will appear as a new version under this series’ concept DOI, with these notes as its prior work.

Sources

A. Documents obtained and read for this note

Every quotation and every numerical claim about a specific document in this note comes from one of these. Most are public on the NASA Technical Reports Server; the exceptions state their source explicitly.

Van Dresar, N. T., Lin, C.-S., Hasan, M. M. — Self-Pressurization of a Flightweight Liquid Hydrogen Tank: Effects of Fill Level at Low Wall Heat Flux. NASA Lewis Research Center and Analex Corporation. AIAA-92-0818, also issued as NASA TM-105411 (E-6813). Prepared for the 30th Aerospace Sciences Meeting, Reno, Nevada, January 6–9, 1992. NTRS 19920009200. Author names, ± values and all quotations drawn from it verified against the page images. — https://ntrs.nasa.gov/citations/19920009200
Johnson, W. L., Balasubramaniam, R., Hibbs, R., Zimmerli, G. A., Asipauskas, M., Bittinger, S., Dardano, C., Koci, F. D. — Demonstration of Multilayer Insulation, Vapor Cooling of Structure, and Mass Gauging for Large-Scale Upper Stages: Structural Heat Intercept, Insulation, and Vibration Evaluation Rig (SHIIVER) Final Report. NASA/TP-20205008233, August 2021, 286 pp. — https://ntrs.nasa.gov/citations/20205008233
SHIIVER Test Plan, eCryo-PLN-0079 Rev. I, 2020. NTRS 20205003433, 53 pp. — https://ntrs.nasa.gov/api/citations/20205003433/downloads/eCryo-PLN-0079_SHIIVER_Test_Plan_RevI_2020-05-30_DLR%20(002).pdf
Results of Use of Heat Flux Sensors on Liquid Hydrogen Tanks. NTRS 20210019122. — https://ntrs.nasa.gov/api/citations/20210019122/downloads/CEC_SHIIVER_htflux_paper_rev3.pdf
Kartuzova, O., Kassemi, M., Agui, J., Moder, J. — Self-Pressurization and Spray Cooling Simulations of the Multipurpose Hydrogen Test Bed (MHTB) Ground-Based Experiment. Conference paper, 2015. NTRS 20150000249. — https://ntrs.nasa.gov/citations/20150000249
Yang, H. Q. (CFD Research Corporation), Patel, C. S. (Qualis Corporation), Williams, B. R. (NASA MSFC) — Jacobs Space Exploration Group at NASA Marshall Space Flight Center. Validation of Cryogenic Propellant Tank Self-Pressurization. 2022. NTRS 20220018548. — https://ntrs.nasa.gov/api/citations/20220018548/downloads/Validation_of_Cryogenic_Propellant_Tank_Self_Pressurization.pdf
Kartuzova, O., Kassemi, M. — Validation of a Two-Phase CFD Model for Predicting Tank Self-Pressurization in the Ground-Based K-Site Experiment. Extended abstract, AIAA SciTech Forum 2025. NTRS 20240006348. — https://ntrs.nasa.gov/api/citations/20240006348/downloads/AIAA-SciTech-2025-Kartuzova-Kassemi-Abstract.pdf
Kartuzova, O., Kassemi, M., Hauser, D. — Validation of a Two-Phase CFD Model for Predicting Tank Self-Pressurization in the Ground-Based K-Site Experiment. Full conference paper, AIAA SciTech Forum, Orlando, January 2025. NTRS 20240016283, 15 pp. SHA-256 2801c5695f2253d69d0f0be41b65fa9febf8f83ff01004b63642602f59fdef6d. Obtained and read for v1.0, after earlier drafts and the verification log wrongly recorded it as paywalled (Note 000, Erratum 16). — https://ntrs.nasa.gov/citations/20240016283
Kartuzova, O., Kassemi, M., Hauser, D. — CFD validation of k-site tank self-pressurization under varying fill levels and heat fluxes with different turbulence models. Cryogenics 152 (December 2025), 104210. doi:10.1016/j.cryogenics.2025.104210, open access (CC BY). Existence verified via Crossref; the version-of-record full text was captured from the publisher’s open-access page on 11 August 2026 and hashed in manifest.csv (companion dataset); the publisher’s PDF remains to be archived as an artifact. Read for v1.0, after the verification log wrongly recorded the article as not found (Note 000, Erratum 16).
NASA SBIR 2021 Phase I Solicitation, subtopic Z10.01 — Cryogenic Fluid Management. Lead center: Glenn Research Center. Source of the “target accuracy of 25%” sentence quoted in Section VII. The live page was unreachable at the time of writing; the text was read from the Internet Archive snapshot of 23 December 2023, which is the durable citation in any case. The subtopic appears to have recurred in the FY2022 solicitation with modified wording; the wording quoted here is the 2021 page’s. — https://web.archive.org/web/20231223153801/https://sbir.gsfc.nasa.gov/content/cryogenic-fluid-management-4 — snapshot retrieved, archived and hashed for this series at v1.0, quote concordance verified; SHA-256 in the companion verification log.

B. Cited from general knowledge — not read in the original

These are stated because they frame the argument, not because this note establishes them. A reader who wants to check the framing should go to the issued documents rather than to me. 11. ASME V&V 20 — Standard for Verification and Validation in Computational Fluid Dynamics and Heat Transfer (V&V 20-2009, R2021). Conceptual source of the comparison-error and validation-uncertainty framework of Section II. The standard was not read in the original; Section II therefore attributes the concept per ASME’s published scope and writes the decomposition in this series’ own notation. Obtaining the standard and verifying clause and notation against it is a recorded v1.1 action. 12. NASA-STD-7009 — Standard for Models and Simulations (current revision: NASA-STD-7009B, approved 5 March 2024), and NASA-HDBK-7009. Source of the treatment of input pedigree and results uncertainty as credibility factors, referenced in Section VIII. 13. Ortho/para hydrogen properties — equilibrium composition at 20 K, the exothermic ortho→para conversion, and catalytic conversion in bulk LH₂ production, referenced in Section IV. 14. NASA Cryogenic Fluid Management portfolio, referenced in Section VII for programmatic context only. — https://www.nasa.gov/space-technology-mission-directorate/tdm/cryogenic-fluid-management-cfm/

ElarionX CPMS develops and evaluates engineering models for cryogenic propulsion systems, and publishes what it finds — including when the finding is about its own work. Competing interest: declared in Note 000 §IX. This note is exploratory: it states an observation and a method, not a validated result. Corrections and correspondence: Luis.emc2@elarionx.com