A new phase of sensing phase change materials

Phase change materials (PCM), wax-like substances that absorb heat or cold as they transition between phases, are a promising technology for power plants, electric vehicle batteries, computer chips, and even heated blankets. Sensing a PCM’s melt-fraction level — and therefore how much heat it can still absorb — historically involves temperature sensors that are tedious to calibrate, costing time and labor.

Parankusam et al. developed a convolutional neural network that uses smartphone infrared camera images to predict PCM melt-fraction levels with 96% test accuracy. The method could replace conventional physical models that require high-accuracy data from finicky temperature sensors.

“You don’t have to mount thermocouples. You don’t have to calibrate them. Just take your cell phone, take a picture, and it immediately tells you how much energy is lost or gained in your thermal battery,” said author Debjyoti Banerjee.

At first, the researchers only collected melt-fraction level data from mounting thermocouples inside and outside PCM vessels, using infrared camera images as temperature validation. Then, the team realized that the image data was “a gold mine that we were sitting on,” said Banerjee. So, they coded and trained a neural network that used the infrared images to predict PCM melt-fraction levels. Finally, the team validated their model using new infrared images, achieving 96% accuracy.

In the future, the researchers aim to make their model predictions more granular, as well as test the model on PCMs of different compositions. While Banerjee says there are currently no commercial power plants that use PCMs, he hopes to develop the model into an industry-ready product that can decrease plants’ water demands from energy-intensive cooling procedures.

Source: “Convolutional neural network-based visual classification of melt-fraction levels in phase change materials,” by Sai Manasi Parankusam, Nishit Pachpande, Anusree Sen, and Debjyoti Banerjee, Journal of Vacuum Science & Technology: A (2026). The article can be accessed at https://doi.org/10.1116/6.0005202 .

This paper is part of the Phase Change Materials: Fundamentals, Innovations, and Applications Collection, learn more here .