A General Super-Resolution Approach Integrating Physical Information for Temperature Field Measurement

Area of Science:

• Engineering
• Thermal Science
• Computational Science

Background:

• Industrial temperature measurement often uses thermocouples and spectroscopy, yielding low-resolution data.
• Existing methods lack precision for detailed thermal analysis in real-world scenarios.

Purpose of the Study:

• To develop a general super-resolution approach for enhancing temperature field measurement accuracy.
• To leverage low-resolution (LR) data from sensor arrays for high-resolution (HR) temperature prediction.

Main Methods:

• A novel super-resolution model incorporating skip connections and multi-path learning.
• Integration of physical information loss to improve predictive accuracy.
• Validation through simulations of silicon chip heating, water mixing, and convective heat transfer.

Main Results:

• The developed learning model accurately predicts high-resolution (HR) temperature fields.
• Simulations demonstrate effectiveness across diverse thermal scenarios.
• The approach successfully generates HR temperature data from LR inputs.

Conclusions:

• The proposed method offers an accurate and efficient alternative to traditional, time-consuming simulations.
• Simplified deployment via a fixed model and lightweight physical loss function.
• Facilitates applications in computational fluid dynamics (CFD), engineering measurements, and beyond.