A serial-path, single-node cooling-stack model for sizing the liner cooling system of small gas-turbine combustors. You compose a cooling scheme from independent modules (smooth backside convection, ribbed channel, impingement jets, film rows), and the tool closes the energy balance on the liner wall — including coolant heat-up along the path, a pressure-drop budget, and an automatic advice layer that tells you where the design stands and which lever is worth pulling next.
The tool reports suggestions; it never changes your parameters automatically — parameter changes remain a human decision. All advice thresholds are explicit in one function and can be edited.
- Serial-path topology: coolant flows through the scheme's module
list in order, absorbing heat and accumulating ΔP segment by segment.
The path-end coolant state feeds the film rows (film supply
temperature
T_c_film). - Two module kinds:
kind='cold'(backside convection / ribbed channel / impingement — change h and absorb heat) andkind='film'(hot-side film row — changes the driving temperatureT_ad). Wall layers (TBC / metal) are pure thermal resistances handled by the solver. Modules are zero-coupled: add, remove, or reorder freely. - Solver: bisection on the single unknown
Tw_cold(cold-side wall temperature) with second-law clamping (no module outlet may exceed its supply temperature). Bisection is used deliberately: with serial coolant heating, fixed-point schemes stop contracting at high NTU (impingement NTU ≈ 3.8) and diverge. - Single-node, design-point, lumped (evaluation station defaults to the hottest station). Station-resolved axial extension is a known future item — see Limitations.
Four built-in schemes at the reference-engine cruise design point (provenance below):
| Scheme | Configuration | T_metal [K] | T_ad [K] | ΔP [kPa] | Coolant rise [K] |
|---|---|---|---|---|---|
| A0 | smooth backside only, film declared but disabled | 1158.7 | 1334.0 | 0.00 | 44.4 |
| A1 | full-length ribbed channel + 1 film row (cross-check) | 760.9 | 837.1 | 0.31 | 20.6 |
| B | impingement (dome) + smooth aft + 1 film row | 662.1 | 740.9 | 5.00 | 168.0 |
| C | B + second film row | 544.5 | 740.9 | 5.65 | 50.4 |
Axial layout of scheme B — colored blocks above/below the wall show the axial coverage of each method; wall layers drawn x8 (not to scale):
Coolant temperature rise and per-module ΔP accumulation along the path (scheme C, two film rows fed by the path end):
After every solve, a rule-based diagnosis prints [ADVICE] lines (and
returns them in res['advice'] for programmatic use). Thresholds are
explicit in cooling_solver.diagnose() — change them there:
| Check | Trigger | Advice |
|---|---|---|
| Wall-temperature margin | T_metal vs creep allowable (30 MPa / 1000 h): < 0 K over / < 100 K margin |
add cooling or lower T_gas; recheck stress level |
| Cold-side saturation | coolant rise ≥ 0.9 × (T_metal − T2) and ΔT > 50 K | lengthening/strengthening cold-side modules is ineffective; bottleneck = coolant heat capacity → more coolant / film row / lower T_gas |
| ΔP budget | > 100 % of budget / > 80 % | enlarge holes / fewer holes / lower impingement ΔP_j |
| Missing film | η = 0 and margin < 150 K | try enabling a FilmRow (set span > 0) |
| Film ineffectiveness | film supply temperature > 0.8 × T_gas | add coolant or shorten upstream absorption length |
Worked example of the saturation rule: in scheme B, doubling the
impingement jet length (40 → 80 mm) leaves T_metal unchanged at
662.1 K — the cold side is saturated and the real levers are coolant
mass flow or film supply.
span = 0 (or span < MIN_SPAN, default 1 mm) = the module does not
exist. One semantics for all module types; disabled modules are
listed as [SWITCH] ... disabled in the console output. Declaring a
film row with span 0 keeps it visible in the scheme definition as a
reminder.
git clone https://github.com/yybelinda/cooling_stack
cd cooling_stack
pip install -r requirements.txt
python cooling_main.py # full run: 4 schemes + figures
pytest tests/ -v # regression suite (14 tests)No compiled dependencies beyond numpy and matplotlib. Figures are
written next to the script (runtime outputs; committed screenshots
live in docs/figures/).
from cooling_solver import solve_scheme, diagnose
from cooling_params import build_bc, scheme_B
bc = build_bc()
res = solve_scheme(scheme_B(), bc, verbose=True)
advice = diagnose(res, bc) # list of suggestion strings
print(res['T_metal'], res['dP_total'])# in cooling_params.py
def scheme_D():
from cooling_modules import ImpingementJet, BacksideConvection, FilmRow
return [
ImpingementJet('impingement (dome)', 0.0, 0.040, L_j=0.050),
BacksideConvection('smooth backside (aft)', 0.040, L_TOTAL),
FilmRow('film row 1', 0.0244, 0.030, d=2e-3, n_rows=12, n_per_row=20),
]
SCHEMES.append(('D_custom', scheme_D))Module list order = coolant flow order. Composition rules (what to
stack, what never to stack) are documented at the top of
cooling_modules.py.
- Baseline, no ribs, no film (A0). 1158.7 K — only 39 K below the 1198 K creep allowable of Hastelloy X (30 MPa / 1000 h). This number is the quantitative case for impingement or film cooling at this T_gas level.
- Enable one film row in A0. Set the FilmRow span (0.0244–0.030 m): T_metal 1158.7 → 760.9 K for 0.31 kPa — the cheapest ΔP in the whole menu, and the anchor for the advice layer's film suggestion.
- Read the saturation rule. Scheme B: raise impingement length 40 → 80 mm; nothing changes (662.1 K). The advice layer names the bottleneck (coolant heat capacity) instead of letting you chase a dead lever.
Scheme A1 is a deliberate cross-check configuration: it reproduces an independent multi-station reference calculation of the same wall stack (ribbed channel + film, frozen coolant temperature) at 750 K. The single-node model gives 760.9 K; the +10.9 K difference is dominated by the coolant temperature rise (+20.6 K), which the reference model freezes at T2 — a convention difference, not an error. The regression suite pins these anchors (±1 K) plus energy residuals (<2 %), the A1 pressure drop, advice-layer triggers, and four boundary cases (all-cold-disabled raises, span filtering, zone overlap detection).
The default design point corresponds to the reference engine published
in the companion technical note
(DOI: 10.5281/zenodo.22784135),
which documents the parent 1D combustor design method and its design
point boundary values. cooling_stack is a standalone cooling-module;
it shares the gas-property formulations and the material data with the
parent suite released as
microgt-combustor-1d,
and is developed as a modified/extended version of that codebase.
- Single-node, design-point, lumped model; station-resolved axial extension is future work (the span mechanism is already in place).
- The impingement correlation (Nu = 0.72·Re^0.75·(H/d)^−0.1) is on the optimistic side and awaits test calibration; the default H/d = 3.6 lies below the correlation's stated validity range (set by the small annulus width).
- Film blowing ratio M is reported on two bases (annulus and jet); the primary basis awaits 3D geometry freeze.
- Multiple film rows share one full-flow m_film (η superposition basis); per-row splitting is not yet modeled.
- Cold-side saturation (NTU ≫ 1) makes cold-side modules insensitive — this is physics, not a bug; re-evaluate per station once the axial extension exists.
Code: MIT. Documentation and figures: additionally CC BY 4.0.
See CITATION.cff or cite as:
Yang, Y. (2026). cooling_stack: a modular 1D cooling-stack design tool for small gas-turbine combustor liners (v1.2.0). GitHub. https://github.com/yybelinda/cooling_stack
Author: Yang Yang — github.com/yybelinda ([email protected] / [email protected])