heatsink_natural_mlp

PINNeAPPle ModelRegistry 'bench_mlp' (4 x 128, SiLU) on log-scaled geometry, material and operating variables, trained on 5000 solves of the HeatSink Sizer physics engine in natural convection (Bar-Cohen & Rohsenow composite correlation, 3-D finite-volume base conduction) with monotonicity constraints: a hinge penalty keeps d log R / d log k and d log R / d log(source width) negative on the training designs and on random in-range designs (pinneapple_design.thermal.heatsink.monotone). Returns the resistance and a split-conformal 95 % upper bound.

heatsink_natural_mlp v1.0.0 (revision v1.0.0.post1) · part of the PINNeAPPle zoo · training data PINNeAPPle-Labs/heatsink-natural-convection

Try it without installing anything: PINNeAPPle zoo demo (pick heatsink_natural_mlp: its predictions along each input with their band and where its physics checks warn or fail, and your own inputs evaluated in the browser).

Task: Plate-fin heat sink in natural convection (vertical fins, free air): base-to-air thermal resistance

Inputs

Outputs

Validity range

• base_width_mm: [30.0, 200.0]
• base_depth_mm: [30.0, 200.0]
• base_thickness_mm: [2.0, 15.0]
• fin_thickness_mm: [0.6, 4.0]
• fin_height_mm: [8.0, 80.0]
• k_w_mk: [150.0, 400.0]
• source_frac_w: [0.15, 1.0]
• source_frac_d: [0.15, 1.0]
• t_ambient_c: [0.0, 50.0]
• fin_gap_mm: [2.2, 30.6]
• fin_gap_note: gap sampled in 2.5-20 mm, then the fin count is rounded; the realised gaps of the training designs span 2.2-30.6 mm
• power_w: [1.0, 80.0]
• regime: natural convection in free air, vertical fins, no radiation

Known limitations

• Natural convection without radiation: at low power, radiation can carry a comparable share of the heat for black-anodised sinks; the reference does not model it.
• Correlation-based reference (typical +-15 % against experiments); the surrogate's own error is against the engine, not against a measurement.
• The engine's resistance rises with power on 0.5 % of designs (non-monotone correlation blending); the power law is checked on the model but not enforced in training.
• Screening tool: re-verify a chosen design with the physics engine.

Architecture

• bench_mlp — {"in_dim": 12, "out_dim": 1, "width": 128, "depth": 4, "act": "silu", "targets": ["log R"], "monotone": {"log_k": -1, "log_fw": -1}}

Physics

• Bar-Cohen & Rohsenow (1984) composite natural-convection correlation for vertical parallel plates
• 3-D finite-volume conduction in the base (spreading resistance)

Validation

Reference: 750 held-out designs solved by the HeatSink Sizer physics engine (pinneapple_design.thermal.heatsink.engine). The engine is checked against independent methods: natural h vs Elenbaas (1942) 2 %, Lee et al. spreading (1-5 %), FVM grid convergence < 2 %.

Trust

Decision: APPROVED · score 1.00 · coverage 100%

Training data

• dataset: heatsink_natural_convection
• source: apps/heatsink_sizer/artifacts/dataset_natural.json.gz

License: apache-2.0

Citation

PINNeAPPle contributors (2026). HeatSink Sizer natural-convection surrogate (monotone). PINNeAPPle model zoo, https://github.com/PINNeAPPle-Labs/PINNeAPPle

Physics validation

pass: every applicable check passed (15/15 applicable checks pass; level standard; generated by pp.validate with PINNeAPPle 0.7.0, schema validation-report/v1).

Model SHA-256 a93b5a6de4cb56d956b1fc9d5e7216d0f19e9c2c170d7b91ba8a4b0eb1e23115.

How to use

pip install "pinneapple @ git+https://github.com/PINNeAPPle-Labs/PINNeAPPle"

import pinneapple as pp

model = pp.zoo.load("heatsink_natural_mlp")          # weights checked against their SHA-256
inputs = {"base_width_mm": 85.6305, "base_depth_mm": 87.183, "base_thickness_mm": 4.99737, "n_fins": 15.0, "fin_thickness_mm": 1.38136, "fin_height_mm": 12.9024, "k_w_mk": 281.774, "source_width_mm": 28.218, "source_depth_mm": 40.7197, "t_ambient_c": 11.241, "power_w": 39.4607}
out = model.predict(inputs)                 # warns for inputs outside the validity range
print({k: round(float(v[0]), 6) for k, v in out.items()})
# {'r_base_k_w': 3.854359, 'r_base_upper95_k_w': 3.906871}
print(pp.zoo.query_checks(model, inputs)[0]["verdict"]["status"])   # per-query trust checks
# pass
print(pp.zoo.validate(model).decisao)        # card metrics reproduced + physics checks
# APPROVED

The inputs are the first held-out test case of the training dataset; its reference value is r_base_k_w = 3.87935. pp.zoo.evaluate("heatsink_natural_mlp") recomputes every metric of the card on the stored reference.

Hosted inference

No Hugging Face inference provider can run this model, so the Hub widget is off (inference: false). It is a small PyTorch network, but its prediction is more than a forward pass: the inputs are encoded and normalised, the outputs denormalised (and, for ensembles, averaged with their spread), and every query is checked against the validity range and the model's physics checks. The pinneapple library does all of that (below); the ONNX graph carries the pre- and post-processing but not the checks; the zoo demo Space shows the model along a sweep of each input and runs it, with its checks, in your browser.

ONNX

model.onnx (opset 17) maps x, shape (n, 11) float32 with the columns base_width_mm, base_depth_mm, base_thickness_mm, n_fins, fin_thickness_mm, fin_height_mm, k_w_mk, source_width_mm, source_depth_mm, t_ambient_c, power_w, to the outputs y: ['r_base_k_w', 'r_base_upper95_k_w']. Pre- and post-processing are in the graph (largest difference from PyTorch on the reference inputs: 9.3e-07); the validity-range and physics checks are not, so check the range yourself.

import numpy as np, onnxruntime as ort
from huggingface_hub import hf_hub_download

path = hf_hub_download("PINNeAPPle-Labs/heatsink-natural-mlp", "model.onnx", revision="v1.0.0.post1")
session = ort.InferenceSession(path)
x = np.array([[85.6305, 87.183, 4.99737, 15.0, 1.38136, 12.9024, 281.774, 28.218, 40.7197, 11.241, 39.4607]], np.float32)
print(dict(zip([o.name for o in session.get_outputs()], session.run(None, {"x": x}))))

Files

The SHA-256 are those of the package's manifest: pp.zoo.load refuses any other file.

Cite

If you use this model, please cite PINNeAPPle (software, Zenodo DOI 10.5281/zenodo.23343062; see CITATION.cff):

@software{pinneapple,
  author = {Barros, Yan and Schiassi, Enrico},
  title  = {PINNeAPPle},
  doi    = {10.5281/zenodo.23343062},
  url    = {https://github.com/PINNeAPPle-Labs/PINNeAPPle}
}