Dataset for boiling acoustic emissions: A tool for data driven boiling regime prediction.

Area of Science:

• Thermal Management
• Fluid Dynamics
• Acoustics

Background:

• Boiling is crucial for thermal management but susceptible to catastrophic failures at boiling crisis.
• Machine learning offers potential for in-situ monitoring and preemptive control of boiling crisis.
• High-quality, labeled datasets are essential for data-driven boiling monitoring.

Purpose of the Study:

• To present a comprehensive dataset of boiling acoustics for machine learning applications.
• To facilitate the development of models for predicting boiling regimes and crisis.
• To support preemptive control strategies for boiling-based thermal management systems.

Main Methods:

• Collected acoustic signals from near-saturated pool boiling experiments under controlled conditions.
• Utilized a hydrophone, pre-amplifier, and data acquisition unit for reliable acoustic data capture.
• Categorized audio files into four boiling regimes: background (BKG), nucleate boiling (NB), pre-critical heat flux (Pre-CHF), and transition boiling (TB).

Main Results:

• Compiled a dataset with 2056 (BKG), 13367 (NB), 399 (Pre-CHF), and 460 (TB) audio files.
• Included acoustic emission data from transient pool boiling experiments with varied parameters.
• Provided MATLAB® codes for audio file processing and classification.

Conclusions:

• The presented dataset is foundational for training and evaluating deep learning models to predict boiling regimes.
• The dataset enables the development of robust data-driven models for enhanced thermal management.
• This work supports advancements in the preemptive control of boiling crisis in high-energy-density systems.