E (deep-research, adversarially verified 24/25 claims 3-0 vs PRIMARY sources) Compiled 2026-06-20 to ground the nx_sim_credibility roadmap. This is a Nishi-internal research-synthesis artifact; the authoritative source URLs are listed per finding. Verification method: WebSearch fan-out -> fetch -> 3-vote adversarial verification (>=2/3 to kill a claim) -> synthesis. 1 claim refuted, scope-gaps flagged honestly. === NASA-STD-7009 "STANDARD FOR MODELS AND SIMULATIONS" (the M&S credibility backbone) === EIGHT credibility factors (2008 baseline / NASA-HDBK-7009 naming): A.Verification B.Validation C.Input Pedigree D.Results Uncertainty E.Results Robustness F.Use History G.M&S Management H.People Qualifications. 7009A (Change 1) renames to: Data Pedigree, Verification, Validation (M&S Development); Input Pedigree, Uncertainty Characterization, Results Robustness (M&S Use); M&S History, M&S Process/Product Management (Supporting Evidence). ('People Qualifications' does NOT appear in 7009A.) NASA-STD-7009B (approved 2024-03-05) is the CURRENT standard; it SPLITS the assessment into two: (1) M&S Capability Assessment: Data Pedigree, Verification, Validation, Development Technical Review, Development Process/Product Management; (2) M&S Results Assessment: Use Assessment, Input Pedigree, Uncertainty Characterization, Results Robustness, Use/Analysis Technical Review, Use Process/Product Management. SCALE: each factor graded on an ORDINAL 5-point scale, Level 0..4, a DISCRETE STEP FUNCTION with NO partial credit. Level 0 = "insufficient evidence to claim anything higher"; Level 4 = "the best possible assessment for that factor." "The factor assessments are ordinal only, and as such, should not be arithmetically manipulated." (=> do NOT average.) HARD CEILING (citable, load-bearing): "The only way to attain a Level 4 assessment for validation is by comparison with results from the actual RWS [real-world system]; therefore, any time before the mission, an assessment of Level 3 is the highest possible for validation." => experimental/real-system data is MANDATORY for top-tier validation credibility. V/V/UQ DEFINITIONS (NASA glossary, verbatim sense): VALIDATION = "the process of determining the degree to which a model or a simulation is an accurate representation of the real world from the perspective of the intended uses of the M&S." ("did we build the RIGHT model?") VERIFICATION (7009A/B) = "the process of determining the extent to which an M&S is compliant with its requirements and specifications." (HDBK phrasing: model accurately represents the underlying mathematical model and its solution.) ("did we build the model RIGHT?") 7009B notes TWO aspects: implementation/code verification vs. solution verification. UNCERTAINTY CHARACTERIZATION/QUANTIFICATION = identify all uncertainty sources (input variables/parameters, numerical errors) and PROPAGATE them to outputs. Level-4 requires "a statistical analysis (e.g., ANOVA) of the output uncertainty ... after rigorous and validated propagation of all known sources of uncertainty," results may include statistical moments, confidence intervals, sensitivity analysis. SOURCES (primary): 7009A https://standards.nasa.gov/sites/default/files/standards/NASA/w/CHANGE-1/1/nasa_std_7009a_change_1.pdf ; 7009B https://standards.nasa.gov/sites/default/files/standards/NASA/B/1/NASA-STD-7009B-Final-3-5-2024.pdf ; HDBK-7009 https://standards.nasa.gov/sites/default/files/standards/NASA/A/0/Historical/nasa-hdbk-7009.pdf === NASA NPR 7150.2 "NASA SOFTWARE ENGINEERING REQUIREMENTS" (risk-tailoring analogue) === Current = NPR 7150.2D (eff. 2022-03-08, OCE). SWE-020 requires the PM to classify every system/subsystem containing software into the highest applicable class. 7150.2D scheme: Classes A-F (F = CIO/business-IT). Earlier (A/B/C): A-H. Classification = PRE-TAILORING: it scales the number+rigor of applicable software-engineering/safety/assurance requirements to the class (not every requirement applies below Class A). Applicability = class COMBINED with safety-criticality via the Requirements Mapping/Compliance Matrix. This is the direct analogue of risk-informed credibility goals (rigor ~ consequence). Class A = highest rigor = "Human Space Flight Software Systems (ground and flight)..." Safety-critical => must be Class D or higher; Class E cannot be safety-critical. Safety-criticality is FORMALLY determined (PM + SMA org, per NASA-STD-8739.8 / 8719.13), not asserted. V&V are DISTINCT, separately planned (SWE-028/029) and tracked-to-closure (SWE-030/031). SOURCES (primary): https://nodis3.gsfc.nasa.gov/displayDir.cfm?Internal_ID=N_PR_7150_002D_&page_name=Preface ; https://swehb.nasa.gov/spaces/SWEHBVD/pages/102695405/SWE-020+-+Software+Classification === ASME V&V 20 (the verified V&V anchor for CFD/heat) === "V V 20 - 2009 (R2021) Standard for Verification and Validation in Computational Fluid Dynamics and Heat Transfer" (issued 2009, reaffirmed 2016 & 2021). Domain-specific V&V standard usable to grade a fluids/thermal sim, complementary to NASA-STD-7009. SOURCE: https://www.asme.org/codes-standards/find-codes-standards/standard-for-verification-and-validation-in-computational-fluid-dynamics-and-heat-transfer === UNIFIED CREDIBILITY AXES (deduplicated across the verified standards; 1-line def + regulator-expected evidence) === 1 CODE VERIFICATION -- code solves the intended equations; evidence: analytic/manufactured-solution comparison (MMS), observed-vs-theoretical order of accuracy. 2 SOLUTION/CALCULATION VERIFICATION -- numerical error in THIS run bounded; evidence: mesh/timestep convergence, discretization error estimate (Grid Convergence Index / Richardson), iterative+round-off bounds. 3 VALIDATION vs EXPERIMENT -- accurate vs the real world for intended use; evidence: quantitative output-vs-physical-measurement with a validation metric (NASA L4 needs the ACTUAL system). 4 UNCERTAINTY QUANTIFICATION/CHARACTERIZATION -- identify+characterize+propagate all uncertainty to outputs; evidence: source inventory, validated propagation, moments + confidence intervals (NASA L4 e.g. ANOVA). 5 SENSITIVITY ANALYSIS -- which inputs drive output variability; evidence: ranked input-output sensitivities. 6 CALIBRATION -- tuning vs data, kept SEPARATE from validation data; evidence: documented calibration data distinct from validation data. 7 INPUT/DATA PEDIGREE -- traceable provenance/quality of inputs+reference data; evidence: documented source/quality/applicability. 8 NUMERICAL-ERROR/CONVERGENCE ANALYSIS -- quantified discretization/iteration error + demonstrated convergence (overlaps 2). 9 REPRODUCIBILITY -- independent re-runs reproduce results; evidence: controlled inputs, versioned model, recorded environment. 10 DOCUMENTATION & TRACEABILITY -- every claim traces to evidence; evidence: V&V reports, requirement->test traceability (NASA track-to-closure SWE-030/031). 11 CONFIGURATION MANAGEMENT -- controlled versions of model/code/data; evidence: CM records, baselines. 12 RISK-INFORMED CREDIBILITY GOALS -- required rigor scaled to decision consequence/model risk; evidence: classification (NASA Class A-F) / context-of-use-driven credibility goals that pre-tailor requirements. 13 SOFTWARE QUALITY/LIFECYCLE -- disciplined engineering lifecycle; evidence: compliance with a SW-engineering standard (NASA NPR 7150.2) appropriate to class. 14 M&S MANAGEMENT & PEOPLE/REVIEW -- process/product management + technical review + qualified personnel; evidence: NASA Process/Product Management + Technical Review factors, formal SMA/Technical-Authority review. === HONEST SCOPE GAPS (requested but NOT verified this run -- research before grading against them) === ASME V&V 10 (computational solid mechanics); ASME V&V 40 (medical-device credibility: model risk, question-of-interest, context-of-use, credibility goals & gradation); FDA 2023 "Assessing the Credibility of Computational Modeling and Simulation in Medical Device Submissions" (credibility-evidence framework built on V&V 40); AIAA G-077 (CFD V&V guide); IEC 62304 (med-device SW lifecycle classes A/B/C); ISO 14971 (risk mgmt); ISO 13485 (QMS); DO-178C (airborne SW DAL A-E). Located source URLs for the next round: FDA https://www.fda.gov/regulatory-information/search-fda-guidance-documents/assessing-credibility-computational-modeling-and-simulation-medical-device-submissions ; ASME V&V40 https://www.asme.org/codes-standards/find-codes-standards/assessing-credibility-of-computational-modeling-through-verification-and-validation-application-to-medical-devices ; AIAA G-077 https://webstore.ansi.org/preview-pages/AIAA/preview_G-077-1998.pdf REFUTED (0-3): "NASA classes are canonically A-E plus subclasses" -- FALSE; real scheme is A-F (current) or A-H (older).