Latest / Internet of Things with Fexingo: Connected Devices, Sensors, and Industrial IoT / How IoT Sensors Are Stopping Pipeline Corrosion Before Leaks
Transcript
- Lucas: You know that sound a metal pipe makes when it's about to fail? Most people don't. But there's a sensor that listens for it. Luna: Like, literally listens? As in, audio? Lucas: Exactly. Acoustic emission sensors. They pick up the high-frequency sound of metal stretching or cracking — way beyond human hearing. And they're being deployed on pipelines to catch corrosion before it becomes a leak. Luna: So instead of digging up a pipe to inspect it, you just... listen to it? Lucas: Right. And the scale of this is impressive. One operator in West Texas — I'm thinking of a midstream company that runs about 200 miles of crude oil pipeline — installed acoustic sensors at intervals along the line. The system listens 24-7, and when it hears a pattern consistent with corrosion pitting or stress cracking, it flags the location. Luna: How do you distinguish that from normal flow noise? Oil moving through a pipe isn't exactly silent. Lucas: That's the machine learning piece. The sensors themselves are piezoelectric transducers — they convert mechanical stress into an electrical signal. But the raw signal is noisy. So the system has been trained on thousands of hours of pipeline audio — normal flow, pump starts, valve adjustments — and on recordings of actual corrosion events from lab tests. The model learns to flag the specific frequency signature of a micro-crack forming. Luna: And it's accurate enough to act on? Lucas: In this case, yes. In the first year of deployment, the operator reported a 40 percent reduction in unplanned downtime. They caught three significant corrosion sites that hadn't yet penetrated the pipe wall. The estimated savings — between avoided leaks, reduced excavation, and environmental liability — came to about $12 million. Luna: Twelve million dollars on one 200-mile stretch. That's real money. Lucas: And it's not just oil. The same technology is being tested on natural gas distribution lines, water mains, even chemical plant piping. The American Society of Mechanical Engineers estimates that corrosion costs the U.S. oil and gas industry roughly $8 billion a year. A lot of that is reactive — you find the leak after it happens. Luna: So shifting from reactive to predictive is the whole play here. Lucas: Exactly. And what makes acoustic emission different from other sensing methods is that it's passive. You're not sending a signal into the pipe — you're just listening. No radiation, no stopping flow, no pigging the line. Luna: Pigging being the process of sending a cleaning or inspection device through the pipe. Lucas: Right. Which has its own costs and risks. Smart pigs can get stuck, and they only give you a snapshot. Acoustic sensors give you continuous data. Luna: But these sensors need power and connectivity. In remote West Texas, that's not trivial. Lucas: Yeah, power is the constraint. Some sites use small solar panels with battery backup. Others use energy harvesting from the pipe itself — there's a device that clamps around the pipe and uses the temperature differential between the oil and the ambient air to generate a trickle charge. For data, it's usually cellular where available, and satellite or LoRaWAN in the really remote stretches. Luna: LoRaWAN — long-range, low-power wide-area network. That's becoming pretty common in industrial IoT. Lucas: Exactly. And the data rate is low — you're not streaming audio. The sensor sends a short burst: a timestamp, a frequency signature, an amplitude reading. That's it. The heavy processing happens in the cloud or at the edge. Luna: What about false positives? If the model flags a crack that isn't there, you still have to send a crew to dig and check. That's expensive. Lucas: It is. The operator I mentioned said their false positive rate settled at around 15 percent after six months of tuning. That's not zero, but it's manageable. And they prioritize alerts based on severity — a high-amplitude event at a known stress point gets a crew dispatched immediately. Lower confidence alerts get logged and checked during the next scheduled patrol. Luna: Fifteen percent false positives — that's actually pretty good for a relatively new application. Lucas: It is. And the model improves over time. Every time they dig and find nothing, that's a new training example. The system gets better at distinguishing rain hitting the pipe from a real acoustic event. Luna: Speaking of that — a quick honest thing. Episodes like this take a lot of research to get the numbers right. A handful of listeners chip in monthly through buy me a coffee dot com slash fexingo, and that's literally what funds making this many of these. Lucas: Yeah, we don't run ads, so listener support keeps the show going. If today's tech conversation gave you something usable, that's the way to help. Luna: Alright, back to pipelines. You mentioned that this isn't just oil and gas. Where else are acoustic sensors making inroads? Lucas: Water utilities are a big one. The EPA estimates that U.S. water mains leak about six billion gallons of treated water per day. Acoustic sensors are being used to detect leaks in large-diameter transmission mains — the kind that run under cities. One project in Philadelphia installed sensors on 15 miles of cast-iron water main and found three leaks within the first month that were invisible from the surface. Luna: Cast iron — that's the old pipe material that's brittle and prone to cracking. Lucas: Exactly. And you don't want to excavate a major city intersection unless you have to. The sensors give you a precise location within a few feet, so the repair crew digs a small hole instead of a trench. Luna: What about the lifespan of these sensors? They're clamped onto pipes, exposed to weather, vibration, maybe corrosive environments. Lucas: Most are rated for five to ten years. The weak point is usually the battery. Some newer designs use energy harvesting exclusively, so theoretically they could last as long as the pipe. But the electronics themselves degrade — especially the piezoelectric element, which can depolarize over time. Luna: So there's a replacement cycle. That factors into the total cost of ownership. Lucas: Right. One sensor costs anywhere from $500 to $1,500, plus installation. For a 200-mile pipeline with sensors every half mile, you're looking at roughly $200,000 to $600,000 in hardware alone. But against a single major leak that could cost tens of millions in cleanup and fines, the math works. Luna: And you mentioned fines — the Pipeline and Hazardous Materials Safety Administration can levy penalties up to $2 million per violation per day. So avoiding leaks has a regulatory incentive too. Lucas: Absolutely. And the public perception angle — nobody wants a pipeline rupture in their community. So operators are increasingly willing to invest in sensing. Luna: What's the next frontier for this technology? Beyond pipes. Lucas: Aircraft structural health monitoring. There are research programs testing acoustic emission sensors on airplane wings and fuselage panels. The idea is to detect fatigue cracks before they become visible during inspections. That's a much harder environment — you have to filter out engine noise, aerodynamic noise, pressurization cycles. But early results from Boeing and Airbus are promising. Luna: So the same principle — listen for the sound of stress — applies across industries. Lucas: Exactly. Concrete bridges, steel bridges, offshore wind turbine towers, even nuclear reactor containment vessels. Anywhere you have a structure that's under load and prone to cracking, acoustic emission can give you an early warning. Luna: It's a pretty elegant idea. Instead of inspecting on a schedule, you let the material tell you when it's in trouble. Lucas: That's the promise of predictive maintenance across IoT. The sensor doesn't replace the human inspector — it tells the human where to look. And that saves time, money, and sometimes lives. Luna: What would it take for acoustic sensors to become standard on every new pipeline? Lucas: Cost reduction and standardization. Right now, every deployment is a custom integration — different pipe diameters, different coatings, different soil conditions. If the industry adopts a common sensor interface and data format, the per-unit cost could drop significantly. We're probably five to ten years from that being the norm. Luna: So for now, it's early adopters with the budget and the risk profile to justify it. Lucas: Right. But as the data accumulates and the models get better, the ROI becomes clearer. And as we've seen in other corners of IoT, once the ROI is proven, adoption accelerates. Luna: Listening to pipes. Who knew that would be such a solid business case. Lucas: It's one of those applications where the technology is mature enough to deliver value, but not yet widespread enough to be boring. I think we'll see a lot more of it in the next few years.