Consolidated Physics-Driven Machine Learning Strategies to Advance Phase-Change Thermal Management Systems
Abstract
Power, propulsion, and energy are significant drivers for many future naval systems. This includes the need for developing a new generation of directed energy weapons and sensors, and an increase in energy concentration in future electrified naval ships, vehicles, and aircraft systems. The main challenge to developing and deploying these systems is the multifold increase in heat rejection loads, which, if not properly managed, would significantly affect their safety and reliability and limit their performance. As identified in the Operational Endurance Priority in the Naval Research and Development Framework, this is a critical research need for the Navy.Most current thermal management systems across naval applications have relied on single-phase cooling methods, which can no longer meet the high energy dissipation requirements. Two-phase thermal management systems utilizing boiling and condensation configurations provide orders of magnitude enhancement in heat transfer coefficients directly increasing the amount of energy dissipation, reducing system weights and volumes, reducing flow and pumping power requirements, and providing more uniformity in system temperatures. However, good design tools that can accurately predict thermal performance in two-phase configurations have only seen limited success. Two-phase flows are extremely complex as flow and thermal transport behaviors change drastically even with small geometric and operating parametric variations, making most traditional predictive-type design tool development futile and only applicable to a very small region of multidimensional parametric space or having low accuracy. Descriptive-type data sciences-driven tools have the capability to capture the effects of multiple phase interactions observed in phase-change configurations, making them a viable alternative to traditional tools. Unfortunately, there are very limited systematic approaches in the field to apply data science to thermal-fluidic devices. The PI proposes research that synergizes multi-faceted efforts, including novel data sciences-driven machine learning (ML) methods, experimental testing, ML vision, and adaptive data-driven computational fluid dynamics (CFD) simulations, as well as model development and validation. This project will have a transformative impact on the fundamental science and engineering principles of two-phase flow thermal management systems by enhancing the state-of-the-art knowledge in the following three aspects. First, consolidated databases will be amassed from worldwide sources on flow boiling and flow condensation performance parameters, namely, heat transfer coefficient, pressure drop, and critical heat flux, and be utilized to develop data-driven, physics-informed MLmodeling tools based on available theories in the literature. The ability to validate individual physical models with comprehensivedatabases would help reveal what mechanistic models or correlations best capture the performance behaviors. Second, with information based on exploratory data analysis and data-driven modeling tests, critical experiments will be performed to obtain heat transfer,critical heat flux, pressure drop, and high-resolution flow visualization data. This new data would help us understand data trends that reveal effects of important flow input parameters on performance. And third, ML-enabled vision of image data and adaptive ML-based mass transfer modeling in CFD will be utilized to extract multi-level liquid, vapor, and interfacial features. These feature-captures will reveal information critical to performance parameter predictions. This project will provide a pathway to develop accuratemodeling tools in phase-change configurations that can be used by DoD engineers and scientists for designing future power, propulsion, and energy thermal management systems that, while capable of handling ultra-high heat fluxes, are also compact, light-weight, and energy-efficient.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Dec 15, 2023
- Source ID
- N000142412039
Entities
People
- Chirag R Kharangate
Organizations
- Case Western Reserve University
- Office of Naval Research
- United States Navy