Imagine being able to hear the flaws in a weld, catching defects before they even reach the inspection stage. Sounds like something out of a sci-fi movie, right? Well, that's precisely what acoustic sensing and AI are now making possible, and it's poised to revolutionize the world of MIG welding.
Remember the 1992 movie "Sneakers"? There's a scene where Whistler, the blind audio expert, pinpoints the location of a hidden device simply by listening to surveillance footage. He wasn't seeing the problem, he was hearing it. Now, fast forward to today, and a company called Sonibel Instruments is applying that same principle to welding. Instead of cracking codes, they’re cracking the code to perfect welds.
The story goes that the idea sparked from a simple observation: a job shop owner who could identify bad welds just by listening to them. CEO Sophia Millar recalls the owner joking about how he could hear the mistakes from his office and had to correct the new hires.
So, how does it work? Sonibel has developed a system that uses an aluminum-cased sensor mounted directly on the welding torch. Unlike vision-based systems that rely on cameras, this system focuses on the sound of the molten metal droplets as they hit the weld puddle. Chief Technical Officer George Hallo explains that the variations in the frequency of these vibrations are the key. These frequencies act as a sonic fingerprint, revealing whether the weld is solid or riddled with defects like perforations or lack of fusion.
The system then uses a proprietary algorithm, developed by Hallo and Chief Product Officer Hooman Piroux, to analyze the sound feedback, interpret it, and display the results on a monitor. "We take small chunks of audio at a time and run it through our model to get the output," Hallo explains. "That output gives us a classification saying whether it’s a defect or it’s a good weld."
And this is the part most people miss: it's not just about detecting a defect; it's about locating it. Currently, the system indicates whether a weld is "good" or "defective" after it's complete, and if there is a defect, the welder knows where in the weld bead the problem occurred. But here's where it gets controversial... While the current system provides a general 'defect' warning and location, the next iteration promises something even more powerful: specific defect identification. Hallo says the next version will pinpoint specific types of defects like porosity or lack of penetration. Imagine the time and material savings!
Millar notes that the system can immediately identify obviously bad welds with porosity. However, she emphasizes that the true value lies in detecting subsurface defects. "Sometimes it’s very clear that this is a terrible, porosity-filled weld," Millar says. "But other times, it’s subsurface, and that’s where our tool would help most.”
The accuracy of Sonibel's system hinges on its extensive database of welds and inspection results. Hallo emphasizes the importance of capturing a wide range of welding settings to ensure the algorithm can accurately identify even subtle subsurface porosity. "A lot of work has been going into getting that variability, covering all the settings, making sure… it doesn’t only catch the really bad defects, but when we have audio that just has a little bit of subsurface porosity, we’re still able to extract that and verify that this sound correlates to this. And with our processing, we’re able to find the certain characteristics that point to porosity.”
Sonibel is still in its early stages, with only a few units in the field for testing. However, the feedback from these tests is crucial for refining the AI software. The company even had a waitlist of partners eager to buy the product before it was fully ready, offering invaluable feedback and data in exchange for beta testing.
"Right now, a big part of what we’re doing is going back to that wait list and seeing who wants to convert, who wants to actually buy the unit," Millar says. "We’re still pretty limited in the manufacturing, but right now we’re doing demos, and then we’ll do the pilots. Because it’s such a new technology, a lot of people don’t know that it works—they just want to test it. So, we’ll give it to them for two or three weeks, and then if they like it, they’ll buy it."
The potential impact of Sonibel's technology is significant. By catching defects early, manufacturers can reduce waste, improve quality, and ultimately save money. It also opens up the possibility of real-time weld process control, where adjustments can be made on the fly to prevent defects from occurring in the first place.
But here's a thought-provoking question: Could this technology eventually replace skilled welders? While the system can identify defects, it still requires a skilled operator to interpret the data and make necessary adjustments. Or does it? What happens when the AI gets so good, it can not only identify defects, but also prescribe the corrective actions? Is this a tool to empower welders, or a step towards automation that could displace them? What are your thoughts? Share your opinions in the comments below!
