AI for hidden physical systems
Haein Jung
정 해 인

Most physical systems cannot be fully observed. Their performance and failure are governed by hidden physical states such as temperature fields, stress distributions, and interfacial dynamics.

My work develops AI models that infer these hidden states from limited observations, enabling real-time prediction of performance, failure risk, and safe operating limits.

The goal is to make complex physical systems understandable, predictable, and controllable in real time.

profile
position Postdoctoral Researcher
education Ph.D., Mechanical Engineering
GIST (Gwangju Institute of Science and Technology)
field Thermal Management
Scientific Machine Learning
Physical Hidden-State Inference
papers 7 SCIE papers
5 first-author
focus Boiling · Laser
§ 01

Research
Program

This research program is organized around one question: how can hidden physical states be inferred from partial observations, and used to predict performance, failure, and safe operating limits in thermo-fluid systems?

00 core problem
Hidden-State Inference in Physical Systems

Most engineering systems are only partially observable. However, critical outcomes—performance degradation, fracture, and burnout—are governed by internal states that cannot be directly measured. My work aims to infer these hidden states using physics-informed learning, inverse modeling, and operator-based methods.

inverse problem latent state inference physics-informed learning operator learning
01 active
Use Case I — Boiling and Two-Phase Systems

Experimental and numerical investigation of subcooled and saturated flow boiling in microchannels, concentric annuli, and finned geometries. Current focus includes onset of nucleate boiling, bubble dynamics, and critical heat flux in phase-change systems.

boiling ONB instability
02 active
Use Case II — High-Power Laser Systems

Thermo-optical analysis of liquid-immersion-cooled solid-state laser systems under high heat flux. Focus on hidden states such as internal temperature fields, thermal stress, refractive index variation, and their effects on beam quality degradation, safe operating limits, and fracture risk.

laser power thermo-optics beam quality failure
03 in development
Physics-Informed AI for Real-Time Physical Decision Making

Ongoing work develops physics-informed neural operators and inverse learning frameworks to predict hidden states, performance, and failure conditions from sparse observations. The long-term goal is real-time physical decision making for design, operation, and control.

PINN DeepONet neural operator surrogate modeling real-time inference
§ 02

Selected
Work

Selected publications in thermal-fluid dynamics and physical systems modeling. Physics-informed AI work is currently in preparation.

2025
Comparison of Subcooled Flow Boiling in Concentric Annuli with Bare and Finned Surfaces
Applied Thermal Engineering · H. Noh, H. Jung, M. Kwon, S. Lee
SCIE
2024
Experimental and Photographic Investigation into Horizontal Subcooled Flow Boiling in Concentric Annuli for Cooling System of Ultra-Fast Electric Vehicle Charging Cables
International Communications in Heat and Mass Transfer · H. Jung, H. Noh, S. Lee
SCIE
2024
Experimental Investigation into Thermal Characteristics of Subcooled Flow Boiling in Horizontal Concentric Annuli for Cooling Ultra-Fast Electric Vehicle Charging Cables
Case Studies in Thermal Engineering · H. Jung, S. Lee
SCIE
2023
Mechanistic Model to Predict Oscillating Frequency of Flow Boiling in Large Length-to-Diameter Ratio Micro-Channel Heat Sinks
International Journal of Heat and Mass Transfer · H. Jung, S. Lee
SCIE
2020
Fouling Mitigation in Crossflow Filtration Using Chaotic Advection: A Numerical Study
AIChE Journal · S.Y. Jung, H.I. Jung, T.G. Kang, K.H. Ahn
SCIE
2020
Flow and Mixing Characteristics of a Groove-Embedded Partitioned Pipe Mixer
Korea-Australia Rheology Journal · H.I. Jung, J.E. Park, S.Y. Jung, T.G. Kang, K.H. Ahn
SCIE
2018
Numerical Study on the Mixing in a Barrier-Embedded Partitioned Pipe Mixer (BPPM) for Non-Creeping Flow Conditions
Korea-Australia Rheology Journal · H.I. Jung, S.Y. Jung, T.G. Kang, K.H. Ahn
SCIE
in prep.
Experimental investigation of subcooled density-wave oscillations in concentric annulus
International Journal of Heat and Mass Transfer · H. Jung, S. Lee
in prep.
§ 03

HiddenField

AI for revealing hidden physical states and predicting performance and failure in real-world systems.

HiddenField explores AI models that infer unobservable internal states in physical systems from sparse observations. Instead of replacing physics, the goal is to combine data, governing equations, and physical constraints to make real-time prediction, diagnosis, and decision making possible.

input
System Conditions
Geometry, operating conditions, and sparse sensor or imaging data.
output
Hidden State Inference
Temperature fields, stress distributions, interfacial dynamics, and other latent physical variables.
output
Performance & Failure Prediction
Output limits, instability onset, fracture risk, and safe operating ranges.
current use cases
Boiling systems are the first target domain for hidden-state inference in multiphase transport, including ONB, bubble growth, instability, and CHF dynamics.

High-power laser systems are the second target domain, where internal thermo-optical states govern beam quality, output limits, and fracture risk.

The long-term direction is broader: AI that can understand and predict hidden states across real world physical systems.
Status: active development · preprints in preparation
Focus: hidden-state inference · performance prediction · failure-aware AI
2
initial use cases
boiling · laser
active development
7
SCIE papers
validation base
5 first-author
future domains
physical systems
long-term direction
§ 04

Curriculum
Vitae

education & positions
2025–
present
Postdoctoral Researcher
Thermal-Fluid & Physics-Informed AI Research
Republic of Korea
~2025
Ph.D., Mechanical Engineering
GIST (Gwangju Institute of Science and Technology)
Thermal-fluid dynamics · two-phase heat transfer
current
HiddenField (in development)
AI for hidden-state inference in physical systems
Independent research direction
selected experience & collaboration
ongoing
Collaborative Research — GIST APRI / LIG Nex1
High-power laser system analysis
Defence photonics · thermo-optical coupling
2019–
2025
Research — Flow Boiling & Heat Transfer
Microchannels · concentric annuli · EV charging
current
Research Direction — Scientific Machine Learning
Inverse problems · operator learning · physical hidden-state inference
Get in
touch

I welcome enquiries regarding research collaboration, scientific machine learning for physical systems, or HiddenField pilot partnerships.

email haeinjung@gm.gist.ac.kr
location Republic of Korea