A thermal-protection research workbench for localized extreme-heat inserts. Version 0.2.0 turns the original layered TPS concept into executable thermal, resource, uncertainty and evidence workflows. It helps researchers reject weak stack assumptions, examine hotspots, compare candidate thicknesses and plan informative coupon measurements before interpreting physical performance.
Evidence: analytical software and explicitly synthetic reference cases. Physical maturity: no coupon, integrated-stack or flight validation. License: Apache 2.0. The optional externally installed AHIS package retains its own license.
The quality badge reflects GitHub's actual workflow state. The delivered release has local verification; GitHub must run the workflow after upload to earn its green check.
| Original 0.1 | Executable 0.2.0 capability |
|---|---|
| Architecture and thermal-budget prose | 1-D transient finite-volume response with radiation, contact resistance and heat accounting |
| Seam/hotspot risk register | 2-D lateral conduction and separate parallel-column defect screening |
| Optional active-relief concept | Finite-action predictive simulated relief with coolant inventory and sensor-fault denial |
| Qualitative margins | Layer/backface temperature limits, explicitly uncalibrated stress/recession indicators and eight uncertainty corners |
| Validation plans | Analytical numerical tests, refinement study, held-out calibration and measurement-selection planner |
| Evidence ladder | Full-input bundles, independent local/global conservation checks, numerical replay and content hashes |
| No build or CI | Installable Python package, one local quality gate and six-job GitHub Actions matrix |
Python 3.11 or later; NumPy is the sole solver dependency. No API key or paid service. Run inside the extracted project root. For Windows, avoid activation-policy issues by calling the virtual environment interpreter directly:
py -3.13 -m venv .venv
.\.venv\Scripts\python.exe -m pip install -e ".[dev]"
.\.venv\Scripts\python.exe scripts/check_green.py --lint
.\.venv\Scripts\python.exe -m hftazen simulate configs/max01-passive.json --out results/passive.bundle.json
.\.venv\Scripts\python.exe -m hftazen audit results/passive.bundle.json
.\.venv\Scripts\python.exe -m hftazen campaign --out results/campaignLinux/macOS: python3 -m venv .venv, then use .venv/bin/python for the same install,
gate and simulation commands. Complete commands for design, panel, calibration,
experiment planning and AHIS are in docs/USAGE.md.
Reference properties, dimensions, coefficients, thresholds, heat pulses and sensor records are synthetic numerical fixtures. They are not measured material data. The passive reference receives a peak absorbed flux of 5 MW/m², equivalent to 500 W/cm². Heating ends at 14 seconds; the run continues to 360 seconds to capture soak.
| Synthetic case | Peak surface | Peak backface | Screening limits |
|---|---|---|---|
| Passive reference | 2892.60 K | 644.85 K | Failed |
| Simulated local relief | 2571.64 K | 634.27 K | Failed |
| Damaged reference | 3244.13 K | 729.39 K | Failed |
| Thin carrier | 2893.86 K | 2042.53 K | Failed |
| 15 MW/m² exposure | 4111.96 K | 711.85 K | Failed |
The declared synthetic backface limit is 450 K. Cooling the surface does not establish
survival of deeper layers. The thickness study returns selected: null when all supplied
candidates fail. The code also flags a backface still warming at the end of a run.
See evidence/RELEASE_STATUS.md and the frozen JSON records for conditions, margins, conservation, refinement and negative controls.
The release has 56 passing unit tests, a passing eight-case evidence campaign and passing Ruff checks in Linux/Python 3.12.14. Numerical checks include analytical steady resistance, radiation equilibrium, exact heat storage, spatial and time convergence, 2-D/1-D agreement, a lateral eigenmode and energy redistribution. Adversarial checks reject mass/heat tampering, falsely upgraded claims, sensor faults and reused data IDs.
CI is configured for Ubuntu and Windows on Python 3.11, 3.12 and 3.13. Those remote jobs are not represented as already run. docs/GITHUB_PUSH.md gives the commands to update the existing repository and inspect the actual result.
The supplied original archive contains no AHIS references or executable donor code. Historical influence cannot be established from that archive alone. This release adds an independent native implementation and a small optional interface to externally installed AHIS coolant accounting, privately exercised against the supplied AHIS 4.0.2. No restricted AHIS implementation is copied or relicensed into this Apache project. Read docs/AHIS_INTEGRATION.md.
The solver uses constant properties and prescribed absorbed heating. It does not model ablation, reacting gas flow, changing oxide scales, moving interfaces, creep, fracture or flight aerodynamics. Stress and recession are uncoupled diagnostic surrogates whose coefficients need exact-coupon calibration. The 2-D panel represents rectangular solid conduction; gas penetration and actual seam mechanics remain unresolved.
The uncertainty screen enumerates corners without claiming a global bound. Design selection is limited to supplied candidates. Calibration and Fisher-information planning are aids for constructing experiments; neither establishes physical validation. No feature has device-dispatch authority. Recovery accounting supplies no TPS credit.
The original Hf-rich Hf-Ta-Zr-W-(C,N), HfSiCN transition, decoupling, carrier and monitoring architecture remains a candidate direction. No 4500..5000°C oxidizing-wall, reusable flight capability, superiority, manufacturing or certification claim is made.