Cooking is not just for heating dinner. It is also for cooling things. Converting liquid to gas removes energy from hot surfaces and keeps everything from nuclear power plants to powerful computer chips from overheating. But when surfaces get too hot, they can experience what is called a boiling crisis.
In a boiling crisis, bubbles quickly form, and before they are released from the heated surface, they cling together to form a vapor layer that insulates the surface from the coolant above. Temperatures rise even faster and can cause disaster. Operators want to predict such errors, and new research provides insight into the phenomenon using high-speed infrared cameras and machine learning.
Matteo Bucci, Norman C. Rasmussen Assistant Professor of Nuclear Science and Engineering at MIT, led the new work, published June 23 in Physical letters used. In previous studies, his team spent nearly five years developing a technique where machine learning could streamline relevant image processing. In the experimental setup for both projects, a transparent heater sits 2 cm across under a bath of water. An infrared camera sits under the heater and points upwards and records at 2,500 frames per second with a resolution of approx. 0.1
The goal was to estimate how close the water was to a boiling crisis. The system looked at 17 factors provided by image processing AI: the “density of the core development site” (the number of sites per unit area where bubbles regularly grow on the heated surface), as well as for each video frame the average infrared radiation at these sites and 15 other statistics on the distribution of radiation around these sites, including how they change over time. Manually finding a formula that properly weighs all of these factors would pose a daunting challenge. But “artificial intelligence is not limited by our brain’s speed or data handling capacity,” Bucci says. Moreover, “machine learning is not biased” by our preconceived hypotheses about cooking.
To collect data, they boiled water on a surface of indium tin oxide, alone or with one of three coatings: copper oxide nanoleaves, zinc oxide nanowires or layers of silica nanoparticles. They trained a neural network on 85 percent of the data from the first three surfaces and then tested it on 15 percent of the data under these conditions plus the data from the fourth surface to see how well it could generalize to new conditions. According to one measurement, it was 96 percent accurate, even though it had not been trained on all surfaces. “Our model didn’t just remember features,” Bucci says. “This is a typical problem in machine learning. We are able to extrapolate predictions to another surface.”
The team also found that all 17 factors contributed significantly to prediction accuracy (though some more than others). Instead of treating the model as a black box using 17 factors in unknown ways, they identified the three intermediate factors that explained the phenomenon: nucleation site density, bubble size (calculated from eight of the 17 factors), and the product of growth time and bubble exit rate. (which was calculated from 12 of the 17 factors). Bucci says that models in the literature often use only one factor, but this work shows that we need to consider many and their interactions. “This is a big deal.”
“This is amazing,” said Rishi Raj, an associate professor at the Indian Institute of Technology in Patna, who was not involved in the work. “Cooking has such complicated physics.” It involves at least two phases of substance and many factors that contribute to a chaotic system. “Despite at least 50 years of extensive research on this topic, it has been almost impossible to develop a predictive model,” says Raj. “It makes a lot of sense to us the new tools for machine learning.”
Researchers have discussed the mechanisms behind the boiling crisis. Does it stem solely from phenomena at the heating surface or also from distant fluid dynamics? This work suggests that surface phenomena are sufficient to predict the event.
Predicting proximity to the boiling crisis not only increases security. It also improves efficiency. By monitoring the conditions in real time, a system could push chips or reactors to their limits without reducing them or building up unnecessary cooling hardware. It’s like a Ferrari on a track, Bucci says, “You want to unleash the power of the engine.”
Meanwhile, Bucci hopes to integrate its diagnostic system into a feedback loop that can control heat transfer, thereby automating future experiments so the system can test hypotheses and collect new data. “The idea is really to push the button and get back to the lab when the experiment is over.” Is he worried about losing his job to a machine? “We just spend more time thinking and not performing operations that can be automated,” he says. In any case, “It’s about raising the bar. It’s not about losing the job.”
New understanding of heat transfer in boiling water can lead to efficiency improvements in power plants
Madhumitha Ravichandran et al., Decrypting the boiling crisis through data-driven exploration of high-resolution infrared thermometry measurements, Physical letters used (2021). DOI: 10.1063 / 5.0048391
Provided by the Massachusetts Institute of Technology
This story has been republished with permission from MIT News (web.mit.edu/newsoffice/), a popular site covering news about MIT research, innovation and teaching.
Citation: Infrared cameras and artificial intelligence uncover the physics of cooking (2021, July 8) retrieved July 8, 2021 from https://phys.org/news/2021-07-infrared-cameras-artificial-intelligence-uncover.html
This document is subject to copyright. Except for any fair trade for the purpose of private investigation or research, no parts may be reproduced without written permission. The content is provided for informational purposes only.