AI-Enabled Physics-Based Inference from Multimodal Measurements for Two-Phase Flow and Heat Transfer Learning
Project Description
Thermal management systems that leverage liquid-to-vapor phase change are critical to applications ranging from large-scale AI computing and power electronics to electrified transportation. Yet their design and optimization remain constrained by long-standing challenges in validating flow-boiling simulations and establishing their generality across operating conditions and flow regimes. Addressing these challenges requires a multidisciplinary convergence of computer vision, fluid physics, and interfacial transport. Experiments can provide sparse and noisy multimodal field measurements, including observations of liquid-vapor interfaces and two-dimensional slices of flow velocity and temperature, while leaving much of the evolving flow inaccessible. The Lillian Gilbreth Postdoctoral Fellow would develop a distinct, yet complementary research direction aligned with the METHODS (Machine Learning Enabled Two-Phase Flow Metrologies, Models, and Optimized Designs) MURI program led by Purdue. The Fellow would explore AI-enabled inference approaches that combine these measurements with governing physics to infer otherwise inaccessible full-field dynamics under measurement uncertainty. The resulting measurement- and physics-informed field estimates would enhance fundamental understanding of coupled flow and interfacial transport, provide richer constraints for simulation validation and model refinement, and advance generalized prediction of phase change and heat transfer during two-phase flows.
Start Date
The project has a multi-year horizon. We will accommodate the availability of the best candidates.
Postdoc Qualifications
Background in mechanical/electrical engineering or computer science with expertise in physics-informed machine learning, computer vision, or modeling/metrology expertise related to phase change processes, interfacial transport phenomena, or flow boiling heat transfer.
Co-advisors
Justin A Weibel, jaweibel@purdue.edu Professor, School of Mechanical Engineering; Director, Cooling Technologies Research Center: https://engineering.purdue.edu/CTRC/research/index.php
Fengqing Maggie Zhu, zhu0@purdue.edu Associate Professor, Elmore Family School of Electrical and Computer Engineering; https://engineering.purdue.edu/~zhu0/
Bibliography
• Khodakarami et al., Neural Netw. 193, 2026. Mitigating spectral bias in neural operators via high-frequency scaling for physical systems
• Sakrikar et al., Int. J. Heat Mass Transf. 259, 2026. Physics-informed Neural Networks to Predict Mass Transfer at Interfaces in Liquid-Vapor Phase Change Problems
• Naderi et al., IEEE ITherm 2025. Enhancing Thermal Management through Deep Learning-Based Analysis of Bubble Dynamics in Flow Boiling
• Khodakarami et al., Comput. Method Appl. M. Spectral bias in physics-informed and neural operator learning: analysis and mitigation guidelines