A team of researchers has taken inspiration from the ancient Chinese board game Go to train an AI designed to provide optimal cooling strategies. 

The team set out to analyse and predict the most effective method of spray cooling, which could keep electrical grids and data centres operating amid surging demand.

The engineers also claim their research could lead to more effective ways of keeping engines, individual computers, and turbines from overheating. 

Taking inspiration from AlphaGo

One of the scientists behind the research, associate professor Jiangtao Cheng at Virginia Tech, had played Go since high school. Invented more than 2,500 years ago, the two-player strategy board game requires the winner to 'control' the most territory on the board. 

From 2014, Google’s AlphaGo allowed human players to take on an AI-powered competitor. Using machine learning, AlphaGo was able to refine its approach the more it played, allowing it to beat a professional human player within a year. Since then, it has taken on and beaten the world’s top players.

Cheng decided to take on AlphaGo himself, a press statement has revealed. Though he lost continuously, the experience, surprisingly, gave him the idea to build a strategy for cooling hot machinery.

“Go is a game of interconnected dynamics, just as a spray cooling system is a network of interacting parameters,” said Cheng. “Success – whether winning the game or optimising the system –requires a holistic understanding of the network and careful management of its interactions, a task that can be greatly enhanced with AI by analysing complex patterns, predicting outcomes, and guiding optimal strategies.” 

Using AI to analyse water droplets

The team, which published their findings in a paper in the journal Artificial Intelligence Review, set out to present a comprehensive analysis of the effectiveness of spray cooling.

The key is water droplets. When they hit a hot object’s surface, each tiny droplet evaporates, carrying away some of the heat, helping to regulate the surface temperature and cool down the object. 

“The way water changes as it encounters heat is different with droplets,” said Lori, the study’s first author. “Droplets pick the heat up more quickly and carry it away because they boil and evaporate so quickly. Because of this fast turnaround cycle, droplets allow a much more effective approach to temperature control.”

However, analysing individual droplets is a tricky task that poses many questions. For instance, what is the best droplet size to effectively tackle heat? What type of spray nozzle is most effective at producing these droplet sizes? Should alternatives to water – solvents, lubricants, or engineered mixes – be considered?

Inspired by AlphaGo, the researchers used machine learning to analyse publicly shared data from 25 existing studies on water droplets. This allowed them to evaluate the basic properties of liquids, how these form droplets, and how they absorb heat. 

“Even though AI always wins on the Go board, I never felt frustrated but learnt to take advantage of AI to tackle challenges and dilemmas in real life, such as thermal management of high power-density electronics,” said Cheng.

“By bridging thermo-fluid science with AI, we’re not just improving spray cooling,” he continued. “We’re actually redefining how we understand and design the thermal systems of the future.”