UARK-NED3/BoilingBench-SeqReg

BoilingBench-SeqReg

BoilingBench-SeqReg is the self-contained model-and-test-data package distributed for SeqReg, an open sequence-regression package for experimental pool-boiling heat-flux prediction from hydrophone, AE-hit, and optical-image inputs.

This Hugging Face Dataset preserves the supplied release tree exactly: 79 files totaling approximately 3.12 GiB. It intentionally includes the same four model artifacts hosted at UARK-NED3/SeqReg, so users can obtain the documented model-and-test bundle from one release.

Contents

Models/ contains the pretrained artifacts:

• —HydReg.joblib — hydrophone sequence-regression model.
• —Hit2Flux_weights.h5 — AE-hit sequence-regression weights.
• —ImgReg.hdf5 and ImagePCA.pkl — optical-image sequence-regression artifacts.

TestData/ contains the supplied evaluation materials:

• —1D-HydReg_TestData/testdataset.csv
• —2D-Hit2Flux_TestData/ — 66 numbered text files
• —RawData/Sound.lvm and RawData/Temperature.lvm
• —ZippedData/ — packaged 1D HydReg, 2D Hit2Flux, 3D ImgReg, image, raw-AE, and raw-image test-data archives

The directory includes both directly accessible files and supplied ZIP archives. They are retained as released; do not treat similarly named files as independent experimental samples without inspecting the SeqReg loading and preprocessing workflow.

Intended use

Use this package with the canonical SeqReg repository, which provides dependencies, data-layout expectations, preprocessing, model loading, and tutorials. The appropriate pretrained artifact and configuration depend on modality:

• —HydReg: hydrophone data; documented with FFT preprocessing and SeqLen=4000.
• —Hit2Flux: AE-hit data; documented with FFT preprocessing, SeqLen=25, and sequence output.
• —ImgReg: optical images; documented with PCAnpy, pcskeep=40, and SeqLen=200.

Provenance, use limits, and safety

The canonical public release record cited by the SeqReg documentation is the SeqReg OSF project. This Hugging Face Dataset is a distribution mirror, not a replacement for its source record or the canonical software repository.

The artifacts were supplied as a release package and their model-file SHA-256 hashes are documented in the companion SeqReg model card. No new training, inference, or performance evaluation was performed during this hosting release. Evaluate model transfer before applying the artifacts to a different facility, fluid, surface, sensor chain, operating regime, or transient protocol.

HydReg.joblib and ImagePCA.pkl are serialized Python artifacts. Load only files obtained from this verified release and in a controlled environment.

License and attribution

SeqReg and the linked OSF project are released under Apache-2.0. Cite the applicable original publication and use the canonical repository for software issues and updates.

• —C. Dunlap, H. Pandey, E. Weems, and H. Hu, Nonintrusive Heat Flux Quantification Using Acoustic Emissions During Pool Boiling, Applied Thermal Engineering, 2023. https://doi.org/10.1016/j.applthermaleng.2023.120558
• —C. Dunlap, C. Li, H. Pandey, and H. Hu, Hit2Flux: A Machine Learning Framework for Boiling Heat Flux Prediction Using Hit-Based Acoustic Emission Sensing, AI in Thermal Fluids, 2025. https://doi.org/10.1016/j.aitf.2025.100002
• —C. Dunlap, C. Li, H. Pandey, Y. Sun, and H. Hu, A Temporal-Spatial Framework for Efficient Heat Flux Monitoring of Transient Boiling, IEEE Transactions on Instrumentation and Measurement, 2024. https://doi.org/10.1109/TIM.2024.3460944

Start here

What this resource supports. BoilingBench-SeqReg provides released pretrained artifacts and test data for sequence regression of pool-boiling heat flux from hydrophone, AE-hit, and optical-image inputs. First five minutes. Use the canonical SeqReg repository for installation, loading, preprocessing, and tutorials: https://github.com/cldunlap73/SeqReg . Then select the artifact matched to the modality and retain its documented preprocessing and sequence configuration: HydReg uses hydrophone data with FFT preprocessing and SeqLen=4000; Hit2Flux uses AE-hit data with FFT preprocessing and SeqLen=25; ImgReg uses optical images with PCAnpy, pcskeep=40, and SeqLen=200. Use with care. This Hub dataset is a distribution mirror; no new training, inference, or performance evaluation was performed during hosting. Assess transfer before using an artifact with a different facility, fluid, surface, sensor chain, operating regime, or transient protocol. Load the serialized Python artifacts only from this verified release and in a controlled environment. Continue. Model card: https://huggingface.co/UARK-NED3/SeqReg NED³ software catalog: https://ned3.uark.edu/software/