It’s one of the most common scientific processes in the world — put a liquid in a heated vessel and it generates vapour bubbles as it evaporates. Boiling forms the basis of cooking, heating and most industrial activity.
Pretty much everyone boils something every day, largely without event. Yet, in the scientific world, the science of boiling remains a deep mystery.
To take a simple example, milk will suddenly form a layer that traps some of the vapour, which eventually overwhelms the layer to cause a frothing spill. It is difficult to predict exactly when this will occur.
At the industrial level, the results of this lack of predictability can be disastrous. Thousands of boilers explode around the world, causing an estimated 2,300 accidents worldwide each year. At least a third of the resultant deaths occur in India.
Some of these accidents are a result of lapses in maintenance and safety standards, but some occur simply because there is still no way to tell when a boiler is being overwhelmed.
Rishi Raj, 38, an associate professor with the department of mechanical engineering at the Indian Institute of Technology – Patna (IIT-P), believes that bubbles could hold the key to this mystery. For 16 years, starting with doctoral research at the University of Maryland, he has been working to decode the precise whys and hows of bubbling.
At IIT-P, he and four other researchers recently demonstrated that an artificial intelligence (AI) and machine learning (ML) algorithm could be used to try and predict when vapour volumes in a container will reach danger levels. And they did this, by listening to the bubbles.
Their findings have been published in the journal Cell Reports Physical Science. Their research is supported by the Swarnajayanti Fellowship of the union government’s department of science and technology.
Here’s the larger problem Raj and his colleagues are seeking to address: Currently, boiler safety strategies and protocols depend largely on estimated correlations between pressure, temperature and the build-up of vapour. But the pressure and temperature sensors inside boilers cannot predict when a sudden surge or variation might occur, and by the time the surge has begun — as with an avalanche — it can be too late to get out of the way.
“If we had a really sound understanding of the science of boiling, maintaining protocol and avoiding accidents would not be too difficult,” says Raj. “But the science of boiling is actually in its infancy. Basically, we do not really know how boiling works.”
A lot could be learnt, Raj believes, by listening better. “In the lab, we could tell when a liquid was on the verge of dangerous expansion from the sound it made. So we decided to focus on sound instead of temperature and pressure alone,” Raj says.
It turns out that larger bubbles have a different sound signature and, in their own quiet way, these signatures have been acting as a sort of alarm all along.
Raj and Atul Thakur, also from the IIT-P department of mechanical engineering, are now using AI and ML to build an algorithm that can tell different bubble sound signatures apart. This will create a new layer of information — one that is not reflected in thermal data.
Over time, as more data is gathered, the theory goes, sound signatures could help predict when a particular substance is nearing danger levels in a boiler.
The thing they still can’t figure out, is why. “We don’t know how this is happening. So now we are also working on understanding the physics behind boiling and sound patterns,” Raj says.