Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 15 additions & 0 deletions CHANGELOG.md
Original file line number Diff line number Diff line change
Expand Up @@ -2,6 +2,21 @@

All notable machine-verified engineering milestones for `aska-physical/funi-frame`.

## [0.1.1] — 2026-11-17

Corrections from the cross-repository review in [`aska-physical/optifab`](/aska-physical/optifab) (`HANDOFF.md` §3). Prepared by an AI agent; awaiting human sign-off.

### Changed
- **Unlocked compute is reported in IT power, not utility power.** The `3.06 MW` of freed substation power carries **`2.85–2.96 MW`** of new IT load (`P_freed / PUE` if every overhead scales with load; `P_freed / (1 + 1/COP_loop)` if only the `ATES` loop does). `unlocked_annual_revenue_low_M_usd` falls from `76.62` to **`71.35`**, and `unlocked_annual_revenue_high_M_usd` (**`147.95`**) is now in the certificate. The low figure pairs the low IT load with `$25M/MW-yr` and the high figure pairs the high IT load with `$50M/MW-yr`; `unlocked_annual_revenue_pairing` says so, and two new fields give the range at each rate alone (`71.35–73.98` and `142.69–147.95`). `README.md` §6 retitled from "Freed Compute Power" to "Freed Utility Power". This assumes the `$25–$50M/MW-yr` value is per MW of IT load, which the owner has still to confirm.

### Added
- **Re-cooling temperature and its second-law limit (`Theorem 5`, `Must-fail #13`).** The certificate now carries `T_drive_supply_C = 58`, `T_chilled_water_C = 18`, `T_recooling_design_C = 30`, the reversible COP limit at design re-cooling (`2.051`) and at `45 °C` dry re-cooling (`0.423`), and `T_recooling_max_for_design_cop_C = 41.46`. The `30 °C` design value is a proposal made in this version with no source cited. `T_recooling_design_basis` records that it awaits the owner's confirmation.
- **`Must-fail #14`:** freed utility power may not be counted as IT power. It checks both IT figures by a power balance (new IT load plus its overhead equals the freed utility power, to within rounding) and both revenue figures against them, so a wrong formula that kept the ordering fails.
- **`declared_inputs`, `controls_on_declared_inputs_only` and `controls_resting_partly_on_declared_inputs`** in the certificate: every value that is stated from the literature or the design brief rather than computed, the one control (`#10`) that compares only such values, and the three (`#5`, `#11`, `#12`) that rest on them in part. This is disclosure, not a fix: `#10` still cannot fail, and a cost model for it is open work.

### Not changed
- No structural, acoustic, vibration, seismic or storage result. `PUE`, `WUE` and all `12` original controls are as in `v0.1.0`.

## [0.1.0] — 2026-11-16

### Added
Expand Down
16 changes: 11 additions & 5 deletions HANDOFF.md
Original file line number Diff line number Diff line change
@@ -1,8 +1,8 @@
# Inter-Agent Handoff — `aska-physical/funi-frame` (`v0.1.0`)
# Inter-Agent Handoff — `aska-physical/funi-frame` (`v0.1.1`)

Prepared per the Aska group ethos (*maximal autonomy and agency; capture every computation, decision, and negative finding so downstream agents build forward without re-deriving*).

## 1. What Was Learned, Proved, and Verified (`v0.1.0`)
## 1. What Was Learned, Proved, and Verified (`v0.1.0`–`v0.1.1`)

- **Theorem 1 — Hysteretic (`Fe-Mn-Si` / `CLD`) vs. Viscous High-Frequency Isolation Roll-Off (`SymPy + Must-fail #7`):**
- Classic Kelvin-Voigt viscous dampers suffer high-frequency damper lock-up: $\lim_{r\to\infty} r\,T_{\text{visc}}(r,\zeta) = 2\zeta > 0$ (`-20 dB/decade`), leaking `29.5–120+ Hz` chiller, pump, and genset harmonics into server cabinets (`0.137 g_rms` on conventional steel frames). Rate-independent martensitic (`Fe-Mn-Si`, `4.5 Hz`, $\eta_1 = 0.16$) and viscoelastic `±45°` CFRP Constrained-Layer Damping (`CLD`, `9.2 Hz`, $\eta_2 = 0.12$) follow $\lim_{r\to\infty} r^2 T_{\text{hyst}}(r,\eta) = \sqrt{1+\eta^2}$ (`-40 dB/decade`), holding white-space rack vibration to **`0.000090 g_rms <= 0.000500 g_rms`** (`17.91%` of `STRUCTURE Magazine` optical/director-switch ceiling, `-63.69 dB` improvement).
Expand All @@ -12,16 +12,22 @@ Prepared per the Aska group ethos (*maximal autonomy and agency; capture every c
- Plain chilled water over a `9 K` delta-T stores only `11.5 kWh_th/m³`, exhausting an `83.33 m³` mechanical buffer in `5.75 min`. Encapsulated `savE HS24` salt hydrate (`30.0 kWh_th/m³`, Osterman et al. 2023) delivers a full **`15.0 min` (`2.5 MWh_th`)** ride-through in `-61.67%` less tank volume, while the outer `~5.2 mm` active Biot penetration skin (`eta_biot = 0.10`) of `28 °C` `mPCM`-doped Poraver wall/ceiling liners passively buffers cold-aisle air `<= 35 °C` (`ASHRAE A2`) for **`28.30 min`** vs. `1.78 min` for unbuffered room air.
- **Theorem 4 — Waste-Heat Sorption Self-Cooling Threshold (`SymPy + Must-fail #11`):**
- Critical liquid-cooling capture fraction $\alpha_{\text{liq}}^* = 1/(1 + \text{COP}_{\text{ads}}) = 61.73\%$ at silica-gel $\text{COP}_{\text{ads}} = 0.62$. Capturing `85%` of IT heat at `58 °C` (`ASHRAE W45`) self-generates `100%` of the `15–18 °C` chilled water needed for the `1.50 MW_th` air/RDHx load using `2.419 MW_th` of server reject heat with **zero mechanical chiller compressors**, dropping campus `PUE` from `1.38` to **`1.074`** at **`WUE = 0.000 L/kWh`**.
- **Theorem 5 — Sorption Re-Cooling Limit (`SymPy + Must-fail #13`, added in `v0.1.1`):**
- A heat-driven chiller between a drive at $T_h$, re-cooling at $T_a$ and chilled water at $T_c$ obeys $\text{COP}_{\text{ads}} \le \frac{T_c}{T_a - T_c}\cdot\frac{T_h - T_a}{T_h}$, so $T_a \le \frac{T_c T_h (1 + \text{COP}_{\text{ads}})}{\text{COP}_{\text{ads}} T_h + T_c}$. At `58 °C` drive, `18 °C` chilled water and `COP_ads = 0.62` the re-cooling temperature must stay below **`41.46 °C`**. The design value is `30 °C` (reversible limit `2.051`, so the machine runs at `30%` of the reversible limit). The certificate now carries the three temperatures.
- **Identity 6 — IT Power From Freed Utility Power (`SymPy + Must-fail #14`, added in `v0.1.1`):**
- Freed utility power $P_{\text{freed}}$ carries $P_{\text{freed}}/\text{PUE}$ of new IT load if every overhead scales with load and $P_{\text{freed}}/(1 + 1/\text{COP}_{\text{loop}})$ if only the `ATES` loop does: **`2.85–2.96 MW`** for the `3.06 MW` freed here. `v0.1.0` multiplied the full `3.06 MW` by the compute value per MW; `v0.1.1` uses the IT figure (`$71.35M–$147.95M/yr`). The low end pairs the low IT load with `$25M/MW-yr` and the high end pairs the high IT load with `$50M/MW-yr`; at one rate the range is `$71.35M–$73.98M` (`$25M`) or `$142.69M–$147.95M` (`$50M`). This assumes the `$25–$50M/MW-yr` value is per MW of IT load, which the owner has still to confirm.
- `Must-fail #14` checks both figures by a power balance: the new IT load plus the overhead it brings must use up the freed utility power to within the certificate's rounding, and counting the freed power itself as IT load must leave the balance open. The two formulas are also proved as SymPy identities.
- **Negative Findings Captured (Do Not Re-Derive):**
1. **Shallow `L/f > 30` (`f = 0.68 m`) on `24 m` heavy-plant roof vaults over-stresses the crown:** Keeping `L/f <= 18` (`f = 1.35 m`, `L/f = 17.78`, `h_vault = 1,480 mm`) cuts horizontal arch thrust in half (`2,181.65 kN`, `f_c = 9.60 <= 24.75 MPa`) while still saving `920 mm` against `2,400 mm` deep heavy roof trusses.
2. **Stage-2 rib isolation above `15 Hz` amplifies `29.5 Hz` screw chillers past `0.0005 g_rms`:** Both isolation stages must sit below `f_min / 3` (`4.5 Hz` plinth + `9.2 Hz` `CLD` cassette seat) to prevent `29.5 Hz` chiller bleed-through.
3. **Air-cooled server exhaust (`32 °C`) and `ASHRAE W17` (`25.3 °C`) cannot drive sorption chillers without heat-pump lift (`Must-fail #11`):** Direct heat-pump-free regeneration requires `>= 55 °C` (`W45`/`W+`) for silica-gel or `>= 70–75 °C` (`HTHP`-boosted, ORNL OSTI 3740567) for single-effect `LiBr-H2O`.
4. **Li-ion battery arbitrage (`$0.068/kWh` annualized pack capex) is `4.86x` costlier than crushed basalt (`$0.014/kWh_th`, `Must-fail #10`).**
4. **Li-ion battery arbitrage (`$0.068/kWh` annualized pack capex) is `4.86x` costlier than crushed basalt (`$0.014/kWh_th`, `Must-fail #10`).** Both figures are declared inputs: `Must-fail #10` compares two stated constants and cannot fail (listed under `controls_on_declared_inputs_only`). Listing it is disclosure, not a fix. A cost model that derives the two figures is open work, and until it exists `#10` guards nothing. `Must-fail #5`, `#11` and `#12` rest partly on declared inputs (`0.72` absorption, written twice in `verify.py`; `COP = 0.0` at `32 °C` exhaust; `WUE = 0.0`) and are listed under `controls_resting_partly_on_declared_inputs`.
5. **`45 °C` dry re-cooling cannot reach the `0.62` adsorption COP (`Must-fail #13`):** the reversible limit there is `0.423`. Zero-water operation at this COP needs the `ATES` / energy-pile / sky sinks.

## 2. Repository Artifacts (`main`)

- [`verify.py`](verify.py) (`v0.1.0`) — 4 SymPy theorems + 6 coupled thermo-structural solvers + **12/12 must-fail controls**.
- [`results/certificate.json`](results/certificate.json) (`v0.1.0`) — Canonical machine-checked outputs.
- [`verify.py`](verify.py) (`v0.1.1`) — 6 SymPy identities + 6 coupled thermo-structural solvers + **14/14 must-fail controls**.
- [`results/certificate.json`](results/certificate.json) (`v0.1.1`) — Canonical machine-checked outputs, with `declared_inputs` naming every value that is stated rather than computed.
- [`docs/ENGINEERING_STUDIES.md`](docs/ENGINEERING_STUDIES.md) — Complete archival monograph of the Whole-Building Vibro-Thermal Shell Study and the Three-Tier Waste-Heat TES & Closed-Loop Sorption Cooling Study.
- [`README.md`](README.md) & [`CHANGELOG.md`](CHANGELOG.md) — System specification and version history.

Expand Down
9 changes: 5 additions & 4 deletions README.md
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
# funi-frame

**Monolithic Vibro-Thermal Funicular Data-Center Shell, Three-Tier Waste-Heat TES & Zero-Water Sorption Cooling (`v0.1.0`).**
**Monolithic Vibro-Thermal Funicular Data-Center Shell, Three-Tier Waste-Heat TES & Zero-Water Sorption Cooling (`v0.1.1`).**

`funi-frame` unifies the [`aska-physical`](/aska-physical) hardware stack—[`funi-floors`](/aska-physical/funi-floors), [`cfrp-truss`](/aska-physical/cfrp-truss), [`fibral-rigidity`](/aska-physical/fibral-rigidity), [`venturi-collars`](/aska-physical/venturi-collars), and [`impedence-analyzers`](/aska-physical/impedence-analyzers)—into a **complete, machine-verified AI data-center building exoskeleton and closed-loop thermal plant** engineered for `80–250 kW/rack` GPU halls that generate high-grade liquid heat (`50–60 °C`) under strict optical/switch vibration limits (`<= 0.0005 g_rms`).

Expand All @@ -21,16 +21,17 @@
- Cuts whole-shell structural dead mass by **`64.83%`** (`1,431.4` vs. `4,069.8 tonnes`), saves **`2.02 m` of building height**, reduces seismic base shear by **`44.23%`**, holds top-down loaded story drift to **`0.563% <= 1.00%`** Immediate Occupancy limit (unbraced heavy frames hit `4.359%`), and seals overhead concrete inside stay-in-place composite pans (`0` spalling dust).
5. **Three-Tier Waste-Heat Thermal Energy Storage (`PCM` + Crushed Basalt + `ATES`/`RTES` + Energy Piles):**
- Pairs **`83.33 m³` salt-hydrate `HS24` `PCM` skids** (`30.0` vs. `11.5 kWh_th/m³` water = `-61.67%` tank volume for `15-min` `2.5 MWh_th` ride-through), **`88.68 m³` (`20 MWh_th`) `500 K` crushed-basalt packed beds** (`225.5 kWh_th/m³`, `$14,046` mineral cost at **`$0.014/kWh_th`** annualized capex—`4.86x` lower than Li-ion), **`200` thermo-active foundation piles (`250 kW_th`)**, and **coastal `ATES` / deep brackish `RTES` wellfields (`COP 10–54`)**.
6. **Closed-Loop Sorption Cooling (`PUE = 1.074`, `WUE = 0.000 L/kWh`, `+3.06 MW` Freed Compute Power):**
- SymPy proves self-cooling above $\alpha_{\text{liq}}^* = 1/(1 + \text{COP}_{\text{ads}}) = 61.73\%$: `85%` `ASHRAE W45` (`58 °C`) cold-plate return directly regenerates **solid-sorbent silica-gel adsorption chillers (`2.419 MW_th` drive -> `1.50 MW_th` RDHx cooling, zero heat pump)**, cutting annual cooling electricity by **`85.05%`**, saving **`157.68 million L/yr` of potable water**, and freeing **`3.06 MW` of utility power (`$76.62M–$153.2M/yr` unlocked AI compute revenue)**.
6. **Closed-Loop Sorption Cooling (`PUE = 1.074`, `WUE = 0.000 L/kWh`, `+3.06 MW` Freed Utility Power = `+2.85–2.96 MW` IT):**
- SymPy proves self-cooling above $\alpha_{\text{liq}}^* = 1/(1 + \text{COP}_{\text{ads}}) = 61.73\%$: `85%` `ASHRAE W45` (`58 °C`) cold-plate return directly regenerates **solid-sorbent silica-gel adsorption chillers (`2.419 MW_th` drive -> `1.50 MW_th` RDHx cooling, zero heat pump)**, cutting annual cooling electricity by **`85.05%`**, saving **`157.68 million L/yr` of potable water**, and freeing **`3.06 MW` of utility power, which carries `2.85–2.96 MW` of new IT load on the same substation (`$71.35M–$147.95M/yr` unlocked AI compute revenue)**. That range varies two things at once: the low end is the low IT load (`2.85 MW`) at `$25M` per MW-yr of IT load and the high end is the high IT load (`2.96 MW`) at `$50M`. At one rate the range is narrow: `$71.35M–$73.98M` at `$25M` and `$142.69M–$147.95M` at `$50M`. New racks need their own cooling and overhead, so freed utility power is not all IT power (`Must-fail #14`).
- **Re-cooling limit (`Must-fail #13`):** the `0.62` adsorption COP at a `58 °C` drive and `18 °C` chilled water is thermodynamically possible only below **`41.46 °C`** re-cooling. The design re-cools at `30 °C` into the `ATES` / energy-pile / sky sinks (reversible limit `2.051`); `45 °C` dry re-cooling cannot reach it (reversible limit `0.423`). The zero-water claim and the COP claim both depend on those sinks.

## Reproduce & Verify

```bash
python3 verify.py
```

Verifies four SymPy symbolic theorems and **12/12 must-fail controls** against [`results/certificate.json`](results/certificate.json). Full engineering monographs, inter-agent handoff, negative findings, and roadmap live in [`docs/ENGINEERING_STUDIES.md`](docs/ENGINEERING_STUDIES.md), [`HANDOFF.md`](HANDOFF.md), and [`CHANGELOG.md`](CHANGELOG.md).
Verifies six SymPy symbolic identities and **14/14 must-fail controls** against [`results/certificate.json`](results/certificate.json). The certificate names the inputs that are declared from the literature or the design brief rather than computed (`declared_inputs`), the one control that rests on declared inputs only (`controls_on_declared_inputs_only`), and the three that rest on them in part (`controls_resting_partly_on_declared_inputs`). Naming them is disclosure, not a repair. Full engineering monographs, inter-agent handoff, negative findings, and roadmap live in [`docs/ENGINEERING_STUDIES.md`](docs/ENGINEERING_STUDIES.md), [`HANDOFF.md`](HANDOFF.md), and [`CHANGELOG.md`](CHANGELOG.md).

## Attribution & Ethos

Expand Down
2 changes: 2 additions & 0 deletions docs/ENGINEERING_STUDIES.md
Original file line number Diff line number Diff line change
@@ -1,6 +1,8 @@
# Engineering Studies Monograph — `aska-physical/funi-frame` (`v0.1.0`)

Complete archival capture of both session engineering studies for [`aska-physical/funi-frame`](/aska-physical/funi-frame), machine-verified by [`verify.py`](../verify.py) against [`results/certificate.json`](../results/certificate.json) (`12/12` must-fail controls):

> **Correction note, `v0.1.1` (2026-11-17).** This monograph is the archive of `v0.1.0` and its text is unchanged. Two things in it were corrected later. The unlocked-compute figure in II.3 (`$76.62M–$153.24M/yr`) counted `3.06 MW` of freed utility power as IT power; the corrected figure is `2.85–2.96 MW` of new IT load. And the `0.62` adsorption COP has a re-cooling limit of `41.46 °C` that this text does not state. See [`CHANGELOG.md`](../CHANGELOG.md) and [`HANDOFF.md`](../HANDOFF.md) for `v0.1.1`.
1. **Study I:** *Whole-Building Monolithic Vibro-Thermal Data-Center Exoskeleton (Funicular Radiative Roof, Zero-Bridge Poraver-mPCM Sandwich Shear Walls, Three-Stage Hysteretic Vibration Cascade & Top-Down Seismic Cellular Box)*
2. **Study II:** *Capturing AI Data-Center Waste Heat, Three-Tier High-Density Thermal Energy Storage (PCM, Mineral Packed Beds & Subsurface ATES/RTES/Energy Piles), and Closed-Loop Sorption Cooling*

Expand Down
Loading