Developing a Machine Learning-Based High-Throughput Adaptive Photothermal Method for Heat Transfer Characterization Using a Digital Micromirror Device or Laser Scanning
Project Description
Characterizing thermal properties rapidly and non-destructively is critical for materials screening, coatings qualification, and device diagnostics. This project develops an adaptive photothermal platform using a Digital Micromirror Device (DMD) or laser scanning to project programmable heating patterns while a camera captures spatiotemporal temperature responses. ML-driven adaptive illumination will automatically select patterns that maximize information gain, enabling high-throughput extraction of thermal conductivity, diffusivity, and interfacial thermal conductance across samples and locations. The goal is a turnkey instrument that outperforms point-by-point pump–probe methods in speed and coverage, with automated calibration and robust inference.
Supervisor
ZHENG, Qiye
Quota
2
Course type
UROP1000
UROP1100
UROP2100
UROP3100
UROP3200
UROP4100
Applicant's Roles
• Assemble and align the DMD-based optical setup; integrate camera triggering and synchronization.
• Implement adaptive pattern selection (e.g., Bayesian optimization/active learning) and fast thermal parameter inference.
• Perform calibration and benchmark against known standards; document protocols and contribute to writing figures/captions.
Applicant's Learning Objectives
• Gain practical experience in thermal imaging analysis, signal processing, and deep learning model development.
• Learn uncertainty quantification, cross-validation, and reproducible ML practices.
• Understand structure–property–performance links in microelectronics/composites and how ML can guide process control.
• Developing a Machine Learning-Based High-Throughput Adaptive Photothermal Method for Heat Transfer Characterization Using a Digital Micromirror Device
Complexity of the project
Moderate