New Delhi: Researchers at the BML Munjal University have developed a safer way to transport large LPG tankers. Their system combines AI, weather data, GPS, and other sensors to detect gas leaks and estimate the size of the danger zone based on the leak.
Lead researcher Amiya Dash from BML Munjal University (BMU) got the idea from Odisha gas leaks. He found cases where tankers would topple, and the authorities were not sure of how large the danger zone really was.
Dash said that in most cases, assessments of the damage took quite a while. His solution was to create a leak-detection system to provide emergency responders with critical information within minutes of an accident.
“If you are underestimating the danger, then that is a disaster. If you are overestimating it, then you are wasting the resources. With AI, the disaster management team can immediately know the risk in the least possible time,” Dash told ThePrint.
Tested and ready to scale
At a time when several accidents are reported due to LPG leaks and tanker explosions, Dash said that preventive technology is essential. Recently, in March, an LPG tanker overturned on NH66 near Udupi. Traffic was diverted due to the gas leak, and it took several hours for the authorities to get the situation under control. Earlier, in 2022, traffic along the Pune-Bengaluru National Highway was also halted due to an overturned LPG cylinder. The police had to evacuate the area and move several people to safer locations.
Dash’s detection system has now been successfully tested at an Indian Oil Corporation LPG bottling plant in Malda, West Bengal. The lead researcher is waiting for a patent and interested companies to scale up production.
How the technology works
The leak detection device is like a small box that can be fitted onto different parts of the LPG tanker. In the case of a malfunction, a sensor detects the unusually high number of gas particles in the air, which triggers a warning. The device then records the location of the tanker, wind speeds, temperature, humidity, and the AI model analyses this data in real time. This information is then sent to emergency services that can cordon off the required area and act accordingly.
To test it out and train the AI model, the team created several accident scenarios virtually with simulations of different leak sizes and weather conditions. Even though Dash had worked with AI for risk assessment before, he was sceptical about how efficient it could be.
“In AI, we have hundreds of mathematical models and algorithms. From these, it is difficult to find which one is most suitable to the kind of task you want it to perform,” said Dash.
He explained that the research team combed through these models to see what kind of parameters they dealt with, whether many of these were linear or non-linear, static or dynamic, and ultimately they found a few to test out.
“We were not sure that the AI model would make predictions that would be accurate or in line with our theoretical safety analysis. But within a few tries, we found that it was nearly 95 per cent to 97 per cent accurate,” said Dash.
At Malda, to test out the final product, the team made small holes in the most vulnerable parts of an LPG tanker, the ones that are usually impacted during accidents or leaks. The AI system successfully collected the relevant data and predicted a risk zone. Dash added that it can also be adapted to detect ammonia, chlorine, and other hazardous chemicals.
Also read: Haryana judge fell for a Tinder scam. Lost Rs 52 lakh
Research to industry
Even though such technology may make the transportation of chemicals safer, in Dash’s experience, taking such an innovation from academia to the market is never easy.
“I have around six patents in my name. But when I go to the industry for commercialisation, I find that people are not generous enough to give funds to researchers,” said Dash. What is offered instead is nearly one-tenth of the cost of building such a product, and many such innovations remain on the shelves of academia.