LANS Seminar

Seminar Title: From Prediction to Forecasting: Machine Learning for Boiling Dynamics

Speaker: Sheikh Md Shakeel Hassan, Ph.D. Student, Department of Electrical Engineering and Computer Science, University of California, Irvine

Date: Thursday, June 19, 2025

Time: 2:30 PM-3:30 PM (In-Person)

Location: Hybrid, Bldg. 240, Conference Room 1405

Description: Modeling boiling — an inherently chaotic, multiphase process central to energy and thermal systems— remains a formidable challenge for computational simulations, often demanding days on petascale supercomputers to simulate only seconds of real world time. This has motivated the development of data-driven machine learning (ML) surrogates that promise orders-of-magnitude acceleration, enabling new capabilities in real-time forecasting, parametric studies, and design-space exploration. However a lack of accessible and diverse datasets suitable for ML training poses a significant challenge. To address this, we created 2 datasets, BubbleML and subsequently BubbleML 2.0. Together, these datasets serve as a large-scale open-source dataset of about 240 high-fidelity boiling simulations, a collection covering multiple fluids (cryogens, refrigerants, dielectrics), boiling regimes (bubbly, slug, annular), configurations (pool and flow), and different gravity and boundary conditions. Benchmarks on downstream temperature and velocity prediction tasks reveal three key limitations of existing state-of-the-art neural PDE solvers: they depend on simulations for future bubble positions during an autoregressive inference, cannot capture discontinuous bubble nucleation events, and struggle with velocity fields in flow boiling. To overcome these limitations, we designed Bubbleformer, a transformer‐based neural PDE solver capable of autonomously forecasting long‐range boiling dynamics (including nucleation, interface evolution, and heat transfer) independent of the simulation inputs. Bubbleformer generalizes across geometries and fluids, and sets new benchmarks for both prediction and forecasting tasks.

Bio: Sheikh Md Shakeel Hassan is a Ph.D. student in the Department of Electrical Engineering and Computer Science at the University of California, Irvine, where he works in the High Performance Computing Lab under Dr. Aparna Chandramowlishwaran. His research focuses on developing scientific machine learning methods and neural PDE solvers to enable an ML-driven understanding of multiphase flow and improve the stability of these models during long autoregressive rollouts. Prior to joining UCI, he completed his Bachelor of Technology in Electronics and Communication Engineering (with a minor in Computer Science) from the National Institute of Technology, Trichy, India. He also worked at an Indian company as a Machine Learning Engineer, developing computer vision and NLP models for AI-enabled mobile applications that saw large-scale user adoption.

Please note that the meeting URL for this event can be seen on the cels-seminars website which requires an Argonne login.