# Infinite Sandbox 0.0.13a_03 — Engineering V&V and Credibility Record

## Status

Infinite Sandbox 0.0.13a_03 is a **standards-informed research and educational
prototype**. It is not certified, accredited, or validated for professional
engineering design. Its output must not be used to approve a structure, choose
construction details, assess a real site, or replace a licensed engineer.

This distinction is required by the standards reviewed for this release. A
mathematically improved solver is not automatically an engineering-grade
product. Professional use also needs an explicitly bounded intended use,
traceable input data, code verification, calculation verification, experimental
validation, uncertainty quantification, configuration management, qualified
reviewers, and jurisdiction-specific design-code checks.

## Intended and prohibited uses

Permissible uses:

- interactive STEM education;
- qualitative experiments with load paths, material differences, failure, and
  uncertainty;
- software-verification research;
- development of future benchmark and validation plans.

Prohibited uses:

- real structural, civil, geotechnical, mechanical, or construction decisions;
- safety-critical analysis;
- certification or approval of any real component or structure;
- treating the included material constants as tested values for a real material,
  specimen, batch, or site.

## Standards and practices reviewed

The following public descriptions or documents formed the requirements
baseline. Listing a standard does not claim conformance or certification.

- [ASME V&V 10 and the ASME VVUQ program](https://www.asme.org/codes-standards/vvuq-standards): separates implementation/code verification, calculation verification, real-world validation, and uncertainty quantification for computational solid mechanics.
- [ASME V&V 20-2009 (R2021)](https://www.asme.org/codes-standards/find-codes-standards/standard-for-verification-and-validation-in-computational-fluid-dynamics-and-heat-transfer/2009): provides a framework for assessing computational-fluid and heat-transfer model accuracy against specified uses and validation data.
- [NASA-STD-7009B](https://standards.nasa.gov/standard/nasa/nasa-std-7009): requires lifecycle planning, traceable data and software, units and coordinate frames, assumptions, equations, limits, permissible uses, verification and validation domains, uncertainty, warnings, risk assessment, reporting, and technical review for credible model use.
- [NASA-HDBK-7009B](https://standards.nasa.gov/standard/NASA/NASA-HDBK-7009): implementation guidance for the NASA modeling-and-simulation lifecycle and credibility assessments.
- [NIST SI Units](https://www.nist.gov/pml/owm/metric-si/si-units): reference for the International System of Units used by the runtime and saved records.
- [USGS, Inside the Earth](https://pubs.usgs.gov/gip/dynamic/inside.html): public educational reference used only for rounded, non-rendered layer-boundary metadata.
- [ISO 2394:2015](https://www.iso.org/standard/58036.html): reliability- and consequence-informed foundations for decisions involving structures.
- [ISO 22111:2019](https://www.iso.org/standard/68335.html): general requirements for semi-probabilistic structural design within established domains of knowledge and practice.
- [ASTM D3080/D3080M-23](https://store.astm.org/d3080_d3080m-23.html): laboratory direct-shear characterization for drained soil strength.
- [ASTM D4767-11(2020)](https://store.astm.org/standards/d4767): triaxial compression data for cohesive-soil strength, deformation, and effective-stress interpretation.
- [ASTM D7012-23](https://store.astm.org/standards/d7012): compressive strength and elastic-modulus characterization of intact rock core specimens.

The full text of paid standards was not available in this project workspace.
Professional conformance work requires properly licensed copies and a competent
reviewer to create a clause-by-clause compliance matrix.

## Requirements traceability

| Credibility element | 0.0.13a_03 evidence | Status |
| --- | --- | --- |
| Intended and permissible use | This record and the saved `level.dat` explicitly limit use to education and research. | Implemented |
| SI units and coordinate frame | Meter–kilogram–second–newton–pascal–joule registry; right-handed frame with +Y upward; standard gravity recorded as 9.80665 m/s². | Implemented |
| Equations and conceptual model | Constitutive, slope, connection, fatigue, integration, shallow-fluid, moisture/burn, analytical-planet, and uncertainty methods are documented and versioned. | Implemented for the stated simplified models |
| Assumptions and limits | This record lists rigid-body, heightfield, isotropy, drainage, material, and validation limits. | Implemented |
| Data pedigree | Every default material records that it is illustrative and names the type of laboratory calibration still required. | Implemented disclosure; calibration not achieved |
| Code verification | Automated analytic tests cover elasticity, invariants, principal stresses, Mohr–Coulomb calculations, combined section demand, fatigue branches, deterministic UQ, fluid conservation, fluid material mobility, surface-state propagation, and analytical planet geometry. | Initial evidence |
| Calculation verification | Central-difference refinement and a velocity-Verlet oscillator timestep/energy study are automated. Runtime timestep loss is counted and reported. | Initial evidence, not a complete application suite |
| Validation against physical reality | No independent experiments or accredited laboratory data have been compared with game output. | **Not achieved** |
| Uncertainty characterization | Deterministic Latin-hypercube propagation, 95% intervals, and correlation sensitivity are implemented for scalar studies. | Kernel implemented; scenario-specific studies still required |
| Domain of verification/validation | Verified functions and their tests are identified. No physical validation domain exists yet. | Verification domain is narrow; validation domain absent |
| Warnings and result reporting | `level.dat` records credibility status, material pedigree, runtime solver diagnostics, warnings, stresses, utilization, factors of safety, and unresolved gaps. | Implemented |
| Configuration management | Application, level-format, cave, solver-model, and package versions are recorded; automated regression tests are packaged. | Partial |
| Independent technical review | No qualified independent reviewer has approved the models, code, test data, or claims. | **Not achieved** |
| Quality-management certification | No audited organizational quality-management system is in place. | **Not achieved** |

## Browser-console engineering record

Engineering details are intentionally kept out of the gameplay HUD. The HUD
contains only the version, measured frames per second, and measured chunk
updates. At startup and every ten simulated seconds, the browser developer
console receives a collapsed, structured engineering report containing:

- the SI unit registry, coordinate frame, application version, and solver-model
  version;
- material input pedigree and the illustrative material constants;
- separate connection tensile, compressive, shear, and equivalent stresses;
- separate terrain vertical, total normal, effective normal, driving shear,
  resisting shear, and bearing stresses;
- governing failure modes, utilization, and factors of safety;
- rigid-body, particle, and shallow-fluid timestep diagnostics, adaptive
  substep counts, volume residual, phase-change accounting, boundary exchange,
  and Courant number;
- non-finite-state event counts and the most recent quarantined state;
- assumptions, verification and validation status, warnings, and unresolved
  certification gaps.

Enter `InfiniteSandboxEngineering.report()` in the browser console to print the
complete report at any time. Enter
`InfiniteSandboxEngineering.snapshot()` to obtain the same information as a
plain machine-readable object without adding gameplay UI.

## Implemented mathematical model

### Units and numerical execution

All engineering-kernel inputs use SI base or derived units. The game uses a
120 Hz fixed rigid-body step and a nominal 20 Hz shallow-fluid update. Each
fluid update is split into deterministic adaptive substeps using a target
Courant number of 0.45 and a 32-substep safety ceiling. Runtime diagnostics
record clamped frame deltas, physics and fluid steps, maximum-substep events,
discarded simulation time, volume residual, boundary exchange, phase-change
sinks, Courant number, target exceedances, and non-finite states. A run that
discards time or exceeds the CFL target is explicitly identified as not
time-accurate.

Millimeter coordinate quantization is a representation choice; it is not proof
of millimeter result accuracy. Accuracy is determined by model form,
discretization, inputs, convergence, and validation together.

### Linear elasticity kernel

The kernel implements small-strain, homogeneous, isotropic Hooke elasticity.
Young's modulus and Poisson's ratio are converted to shear modulus, bulk
modulus, and the Lamé parameter. It computes the full symmetric stress state,
stress invariants, von Mises equivalent stress, and principal stresses.

This kernel is verified analytically but the game is not a continuum finite-
element solver. Rapier still advances rigid bodies and contacts.

### Terrain screening

Natural-terrain cells use a drained infinite-slope Mohr–Coulomb screening model:

`tau_resist = cohesion + effective_normal_stress * tan(friction_angle)`

The solver separates normal surcharge from slope-parallel driving stress and
computes a factor of safety. Bearing compression is checked separately. In
0.0.13a_03, water depth and the 0–100 moisture state create a bounded hydrostatic
pore-pressure approximation for exposed surface cells. This is not a
groundwater, seepage, consolidation, unsaturated-flow, or coupled drainage
model. Near-vertical faces are flagged because an infinite-slope approximation
is weak in that domain.

### Shallow fluids and surface exposure

Water and lava use a finite-volume-inspired, pairwise conservative depth update
on a bounded local grid. Grid resolution is selected from 65 to 129 samples
according to world size and recorded with the resulting cell spacing. The ocean
boundary is a fixed one-meter reservoir. Water and lava use different density
and mobility inputs; donor scaling prevents a cell from exporting more depth
than it owns. A gravity-wave-speed screen subdivides the nominal update toward
a Courant target of 0.45. Fluid diagnostics record water and lava volume,
reservoir exchange, water-lava phase-change removal, signed and absolute
numerical balance residuals, dry-cell clamps, adaptive substeps, Courant target
exceedances, runtime timestep loss, and non-finite states.

The rendered water surface now has deterministic repeating UV texture data and
animated texture coordinates. This is a visual feature only and is deliberately
excluded from the governing equations and validation claims.

The model is intentionally reduced-order. It does not resolve three-dimensional
velocity, pressure, turbulence, temperature, viscosity variation, free-surface
breaking, entrained air, sediment transport, groundwater, or thermochemical
solidification. Moisture and burn are bounded exposure indicators from 0 to
100, not calibrated moisture content or temperature. They adjust qualitative
material response and texture blending but are not fire-design calculations.

### Analytical planet and local coordinate frame

The Earth-sized world is not tessellated. A 6,371,000-meter mean-radius sphere
and rounded layer boundaries are stored analytically, while gameplay uses a
256-meter by 256-meter local East-Up-North tangent patch. The one-meter ocean
datum and spherical curvature are recorded in SI units. Crust, mantle, outer
core, and inner core entries are non-rendered metadata only; they do not enter
local stress or fluid equations. This prevents the metadata from implying a
site-specific geophysical model.

### Constructed-component connections

Every breakable connection records contact area, two contact dimensions, and a
local surface normal. The normal rotates with the attached body. Gravity and
relative motion are resolved into normal and tangential components. The section
check combines:

- axial tension or compression;
- transverse shear with a rectangular-section peak factor;
- bending stress from section modulus;
- torsional shear from an approximate rectangular contact section;
- an equivalent-stress screen.

Separate tensile, compressive, shear, and equivalent utilizations are computed;
the governing mode, peak utilization, minimum factor of safety, and stress
components are saved. This is more defensible than the earlier single scalar
bond strength, but it is still a reduced-order connection model rather than a
mesh-resolved joint analysis.

### Fatigue and fracture

Damage below the declared endurance ratio is zero in the simplified S–N model.
Above it, cycles are accumulated against a material reference curve. Static
overload uses a separate finite overload-failure time. Fracture debris energy
includes kinetic energy and a contact-area fracture-energy term.

The included S–N and fracture values are illustrative. Real use would require
tested curves, load spectra, mean-stress correction, environmental factors,
anisotropy, statistical scatter, and model validation.

### Uncertainty propagation

The engineering kernel provides deterministic Latin-hypercube sampling for
uniform and normal scalar inputs. It reports the mean, standard deviation,
median, minimum, maximum, 95% interval, and input/output correlation
sensitivities. A deterministic seed makes studies reproducible.

This capability does not automatically create a valid uncertainty model. A
qualified analyst must identify aleatory and epistemic sources, justify their
distributions and correlations, and validate the propagation model for each
proposed use.

## Automated acceptance evidence

Run `npm test` or `npm run test:engineering`. The engineering test suite checks:

1. required material fields, ranges, SI interpretation, and pedigree labels;
2. elastic constants against closed-form values;
3. uniaxial and pure-shear Hooke-law cases;
4. principal stresses and von Mises invariants;
5. effective-stress Mohr–Coulomb strength;
6. infinite-slope and bearing calculations;
7. second-order central-difference refinement behavior;
8. combined connection-section demand and governing modes;
9. endurance, finite-life fatigue, and overload branches;
10. oscillator timestep refinement and energy error;
11. reproducible uncertainty intervals and sensitivity signs;
12. mandatory non-certification and unresolved-gap disclosures;
13. analytical Earth radius, layer lookup, curvature, and serialization;
14. fluid conservation, boundary accounting, adaptive-CFL behavior, water/lava
    mobility contrast, water-lava interaction, wetting/burning, deterministic
    seeding, depth color, UV generation, and water-texture animation;
15. unbreakable natural terrain behavior for bedrock.

Passing these tests means the listed mathematical functions match their coded
specifications for the tested cases. It does not validate the game against the
real world.

## Major model limits

- rigid-body contacts are not a nonlinear finite-element or discrete-element
  continuum solution;
- natural terrain is a single-valued heightfield and cannot represent arbitrary
  three-dimensional stress tensors or internal strata;
- cave roofs use a triangle shell and do not yet solve surrounding rock-mass
  stress redistribution;
- pore pressure is only a shallow surface-water approximation; groundwater,
  seepage, consolidation, and unsaturated flow are absent;
- rock, soil, masonry, wood, and rooted soil defaults are uncalibrated;
- wood and sedimentary rock are treated as isotropic even though real materials
  may be strongly orthotropic or anisotropic;
- burn exposure is not a thermal/fire model; heat transfer, combustion,
  temperature-dependent properties, corrosion, creep, shrinkage, seismic
  spectra, cyclic soil degradation, liquefaction, and rate-dependent
  plasticity are absent;
- the shallow-fluid grid is four meters wide per cell and has no mesh/time-step
  convergence study, turbulence, sediment, waves, or experimental validation;
- the analytical mantle/core entries are metadata only and do not model
  geodynamics, self-gravity, pressure, temperature, or seismic response;
- no jurisdiction-specific load combinations, resistance factors, partial
  factors, detailing rules, or serviceability limits are implemented;
- no accredited experiments, round-robin comparisons, or independent review
  establish a domain of validation.

## What is required before an engineering-grade claim

1. Select one narrow proposed use, material family, geometry class, loading
   regime, jurisdiction, and required decision accuracy.
2. Obtain licensed current standards and have a qualified engineer create and
   approve the clause-level requirements matrix and acceptance thresholds.
3. Replace illustrative material defaults with traceable specimen/site data
   from suitable accredited tests, retaining uncertainty and environmental
   conditions.
4. Add the continuum, contact, plasticity, drainage, anisotropy, fracture, and
   boundary-condition models required by that use.
5. Complete method-of-manufactured-solutions or equivalent code verification,
   benchmark suites, mesh/timestep/iteration convergence, conservation checks,
   and independent reference-solver comparisons.
6. Validate every important response quantity against independent physical
   experiments across the claimed domain, including uncertainty and predictive
   error bounds.
7. Establish controlled requirements, reviews, defect tracking, releases,
   reproducible evidence, qualified users, and an audited quality process.
8. Obtain independent professional and, where applicable, regulatory approval.

Until those steps are completed, the correct description is
**engineering-inspired and standards-aligned educational simulation**, not
engineering-grade analysis software.