{"success":true,"data":{"author":"Curl and Hu, University of Arkansas NED3 Lab","bucketAccessType":"public","bucketName":"UARK-NED3/ColdPlate-CFD-2000sims","bucketPrefix":"","category":"Physics","cloudProvider":"huggingface","createdAt":"2026-11-16T02:38:29.576Z","description":"2,000 steady-state CFD simulations of a liquid-cooled, dual-chip, pin-fin cold plate for electronics thermal management, deposited on Dryad (10.5061/dryad.k0p2ngfp5) by Curl and Hu (University of Arkansas) with the POD-NN surrogate described in AI Thermal Fluids (2026). ANSYS Fluent solves the steady incompressible Navier-Stokes and energy equations with Menter's k-omega SST turbulence model on a roughly 50,000-element 3D grid; a Latin Hypercube design varies six boundary conditions: the two chip heat fluxes and the mass flow rate and inlet temperature of two coolant inlets. Each simulation is one NumPy archive (1_Dataset/sim_0001.npz to sim_2000.npz) with temperature fields and node coordinates on the yz mid-plane (11,110 nodes), the zx mid-plane and the bottom cooling face, plus scalar chip mean and max temperatures, outlet temperature and pressure drop per circuit; model_setup.json holds the design of experiments. The repository also carries the trained proper orthogonal decomposition plus neural network surrogate (4 modes capture 99.3 percent of variance) and scripts that reproduce every figure. About 1.4 GB, readable with NumPy. CC0-1.0 (Dryad); cite Curl and Hu, AI Thermal Fluids 2026, doi 10.1016/j.aitf.2026.100040.","displayTitle":"Cold Plate CFD: 2,000 Steady Conjugate Heat Transfer Simulations of a Dual-Chip Pin-Fin Liquid Cold Plate","doi":"10.5061/dryad.k0p2ngfp5","id":"1f0e8762-41cc-5a41-b2f7-17c30c9253d3","license":"CC0-1.0","licenseEvidenceUrl":"https://datadryad.org/api/v2/datasets/doi%3A10.5061%2Fdryad.k0p2ngfp5","objectCountEstimate":2023,"probeStatus":"ok","referenceUrl":"https://doi.org/10.1016/j.aitf.2026.100040","region":"us","schemaCard":{"accessHints":[{"language":"python-datasets","snippet":"from datasets import load_dataset\nds = load_dataset(\"UARK-NED3/ColdPlate-CFD-2000sims\")"},{"language":"python-hub","snippet":"from huggingface_hub import snapshot_download\nsnapshot_download(repo_id=\"UARK-NED3/ColdPlate-CFD-2000sims\", repo_type=\"dataset\")"}],"caveats":["HuggingFace metadata: author=UARK-NED3, license:cc0-1.0, size_categories:1K<n<10K, downloads=787"],"format":"mixed","generatedAt":"2026-11-16T02:38:32.794Z","layout":{"example":"1_Dataset/sim_0001.npz","pattern":"{repoId}/{filename}"},"mountPath":"/mnt/archive","sampleHeaders":[{"bytes":"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","encoding":"hex","key":"1_Dataset/sim_0001.npz","truncatedAt":512}],"sampleObjects":["1_Dataset/model_setup.json","1_Dataset/sim_0001.npz","1_Dataset/sim_0002.npz","1_Dataset/sim_0003.npz","1_Dataset/sim_0004.npz"],"sidecars":[{"bytes":"---\nlicense: cc0-1.0\nlanguage:\n- en\ntask_categories:\n- tabular-regression\ntags:\n- computational-fluid-dynamics\n- surrogate-modeling\n- proper-orthogonal-decomposition\n- electronics-cooling\n- thermal-management\n- reduced-order-model\n- ansys-fluent\n- digital-twin\n- cfdtwin\nsize_categories:\n- 1K<n<10K\npretty_name: Cold Plate CFD (2000 simulations)\nviewer: false\n---\n\n# Cold Plate CFD: 2000 Simulations\n\n2,000 steady-state CFD simulations of a liquid-cooled, dual-chip, pin-fin cold\nplate, plus the POD-NN surrogate trained on them and the code that reproduces\nevery figure in the paper.\n\n**Dataset DOI:** [10.5061/dryad.k0p2ngfp5](https://doi.org/10.5061/dryad.k0p2ngfp5)\n· **Primary article:** *AI Thermal Fluids*, [10.1016/j.aitf.2026.100040](https://doi.org/10.1016/j.aitf.2026.100040)\n· **Model:** [UARK-NED3/PODNN-ColdPlate](https://huggingface.co/UARK-NED3/PODNN-ColdPlate)\n· **License:** CC0 1.0 (public domain dedication)\n\nMirror of the Dryad deposit published 29 April 2026.\n\n## What was simulated\n\nThermal management sets the ceiling on power density in modern electronics,\nand CFD is accurate enough to design against but far too slow to sit inside an\noptimization loop or a live digital twin. This dataset was built to train and\nbenchmark surrogates that close that speed gap.\n\nThe CFD model solves the steady, incompressible Navier–Stokes equations\ncoupled with the energy equation using Menter's k-ω SST turbulence model on a\nroughly 50,000-element three-dimensional grid in ANSYS Fluent. The design of\nexperiments varies six boundary conditions across the cold plate's operational\nspace by Latin Hypercube Sampling:\n\n| Parameter | Boundary |\n| --- | --- |\n| `chip1.thermal.heat_flux` | chip 1 |\n| `chip2.thermal.heat_flux` | chip 2 |\n| `inlet1.momentum.mass_flow_rate` | inlet 1 |\n| `inlet1.thermal.total_temperature` | inlet 1 |\n| `inlet2.momentum.mass_flow_rate` | inlet 2 |\n| `inlet2.thermal.total_temperature` | inlet 2 |\n\nThe published surrogate uses only the central yz mid-plane (11,110 nodes).\nThe other stored surfaces are included for reuse but are not exercised by the\npaper.\n\n## Headline results from the paper\n\n- POD with 4 retained modes captures **99.32%** of the temperature-field\n  variance.\n- The trained POD-NN predicts a full 2D temperature field **412,500× faster**\n  than the underlying CFD solve.\n- A 100-sample training set is enough to reach the error floor; the\n  iteration-cost crossover with CFD occurs at 100.5 design iterations\n  (about 2.76 hours).\n\n## Contents\n\n### `1_Dataset/`\n\n`sim_0001.npz` through `sim_2000.npz` — one NumPy archive per simulation. The\ni-th file corresponds to the i-th boundary-condition row in\n`model_setup.json`. Each holds 14 float-array keys:\n\n| Key | Shape | Meaning |\n| --- | --- | --- |\n| `yz-mid\\|temperature` | (11110,) | Temperature (K) on the central yz mid-plane — the field used in the paper |\n| `yz-mid\\|coordinates` | (11110, 3) | X/Y/Z node coordinates (m) |\n| `zx-mid\\|temperature` | (40939,) | Temperature (K) on the zx mid-plane |\n| `zx-mid\\|coordinates` | (40939, 3) | X/Y/Z node coordinates (m) |\n| `bottom\\|temperature` | (3193,) | Temperature (K) on the bottom-face cooling boundary |\n| `bottom\\|coordinates` | (3193, 3) | X/Y/Z node coordinates (m) |\n| `chip1_tavg\\|temperature`, `chip1_tmax\\|temperature` | (1,) | Chip 1 mean / max temperature |\n| `chip2_tavg\\|temperature`, `chip2_tmax\\|temperature` | (1,) | Chip 2 mean / max temperature |\n| `outlet_tavg\\|temperature`, `outlet_tmax\\|temperature` | (1,) | Outlet mean / max temperature |\n| `pdrop_1\\|temperature`, `pdrop_2\\|temperature` | (1,) | Pressure drop, circuits 1 and 2 |\n\nTwo notes on the scalar channels, both verified against the released files:\n\n- The `pdrop_*` keys carry **pressure drop**, not temperature. The\n  `|temperature` suffix is an artifact of the original export naming — every\n  scalar report was written with it. Sampled values span roughly 4.2 to 12.0\n  while temperatures sit at 300–480 K. `model_setup.json` lists `pdrop_1` and\n  `pdrop_2` as their own Fluent Report Definitions. **The pressure unit is not\n","encoding":"utf8","key":"README.md","name":"README.md","size":12464,"truncated":true}],"sourceObjectKey":"1_Dataset/sim_0001.npz","starterCell":"from huggingface_hub import hf_hub_download\npath = hf_hub_download(repo_id=\"UARK-NED3/ColdPlate-CFD-2000sims\", filename=\"1_Dataset/sim_0001.npz\", repo_type=\"dataset\")\nprint(path)","starterSamplePath":"1_Dataset/sim_0001.npz","totalObjects":2023},"summary":"2,000 steady-state CFD simulations of a liquid-cooled, dual-chip, pin-fin cold plate for electronics thermal management, deposited on Dryad (10.5061/dryad.k0p2ngfp5) by Curl and…","title":"Cold Plate CFD: 2,000 Steady Conjugate Heat Transfer Simulations of a Dual-Chip Pin-Fin Liquid Cold Plate","updatedAt":"2026-11-16T15:46:16.547Z","visibility":"public","websiteUrl":"https://huggingface.co/datasets/UARK-NED3/ColdPlate-CFD-2000sims"}}