Latest / Internet of Things with Fexingo: Connected Devices, Sensors, and Industrial IoT / How IoT Sensors Detect Gas Leaks Before Explosions
Transcript
- Lucas: Last month, a gas utility in Ohio avoided what could have been a serious residential explosion. The cause wasn't a better pipe or a faster crew — it was a $400 IoT sensor clamped onto a distribution line that detected a pinhole leak three hours before the gas migrating into a basement reached an ignition source. Luna: That's the kind of story you don't hear on the evening news, because nothing happened. But that non-event is exactly what industrial IoT is built for. Lucas: Exactly. And with episode 100, I wanted to look at one of the most mature, under-covered applications in the space: natural gas leak detection. Not the smart home detectors that beep at you when your stove is off — the industrial-grade sensors on distribution pipelines, compressor stations, and storage facilities that catch leaks before they become emergencies. Luna: And this isn't just about safety, right? Methane is a potent greenhouse gas. The EPA estimates that the oil and gas sector leaks something like 13 million metric tons of methane per year. That's a climate problem as much as a safety one. Lucas: It's both. And the technology stack is surprisingly elegant. There are three main sensor types in use today. The most common is the catalytic bead sensor — it's basically a platinum wire coated with a catalyst that oxidizes methane when it's present, which changes the temperature and therefore the resistance of the wire. Cheap, reliable, but it needs oxygen to work, and it can be poisoned by silicone or sulfur compounds. Luna: So not ideal for every environment. What's the step up from there? Lucas: Infrared point sensors. They shine an infrared beam at a specific wavelength that methane absorbs — 3.3 micrometers. If gas is present, the beam attenuates. No oxygen needed, no poisoning. They're more expensive, but they last longer and they work in inert atmospheres. And then there's ultrasonic detection — it listens for the high-frequency hiss of gas escaping under pressure. The advantage there is speed; you can detect a leak in milliseconds, but it doesn't tell you what the gas is or how concentrated it is. Luna: So you'd want all three in different spots, depending on the risk profile of that section of pipe. Lucas: Right. And that's exactly what the Ohio utility did. They partnered with a startup called Sensitron — not the consumer brand, a different company — that deployed a mesh network of catalytic bead and ultrasonic sensors at 200 sites across their distribution system. Every sensor reports pressure, temperature, and gas concentration to a cloud dashboard every 30 seconds. The system uses an algorithm that compares readings from adjacent sensors to pinpoint the location of a leak within a few feet. Luna: And the result was a 70 percent reduction in emergency callouts in the first year. I read that report. They went from about 40 calls a month to 12. Most of those avoided responses were false alarms — a lawnmower bumping a valve, a construction crew shearing a line — but caught early enough to dispatch a routine repair instead of a hazmat team. Lucas: That's exactly the metric that matters to utilities: avoided risk and avoided cost. A single hazmat-level response can cost $5,000 to $10,000. The entire sensor network for 200 sites cost about $400,000 upfront, including installation. So they recouped the investment in less than a year. And that's before you factor in the avoided property damage or the avoided methane release. Luna: And it's not just big utilities. There are now iot as a service models where smaller municipal gas companies can subscribe to a monitoring service for a few hundred dollars a month per site, without buying the hardware. That's opening up the market. Lucas: That's a critical point. The technology has been reliable for years — the leap is in the business model and the connectivity. Ten years ago, you needed a dedicated radio network to get that data back. Today, a cat-M1 or nb iot module costs about $5 and sips power so a sensor can run on a single battery for three years. The unit economics finally work. Luna: And speaking of things that work because of small, recurring support — this show is ad-free and listener-supported. A couple of dollars a month from people who find these episodes genuinely useful is what keeps the IoT deep dives coming. If today's conversation gave you something you'll use — a new angle on sensor selection, a specific ROI story, whatever — buy me a coffee dot com slash fexingo is where you can chip in. That's buy me a coffee dot com slash fexingo. It adds up fast. Lucas: Absolutely. And it keeps us free to do exactly this kind of episode — the 100th one, no less. So thank you to everyone who already does. Back to the gas leak story — one more angle I think is underappreciated is the integration with automated shutoff valves. Luna: Right, because sensing alone doesn't stop the leak. You need an actuator. Lucas: Exactly. Some of the newer installations link the sensor directly to a motorized ball valve that can close a section of pipe within seconds of detecting a pressure drop or a methane spike. The utility in Ohio didn't do that initially — they wanted human confirmation before cutting gas to a neighborhood. But after a false positive during a scheduled maintenance window, they realized the cost of a brief, planned shutoff is much lower than the cost of waiting. Now they've deployed automated valves at 40 of the highest-risk sites. Luna: And what about false positives? If a sensor triggers a shutoff unnecessarily, you've got a big problem — potentially hundreds of customers without gas until a technician can reset the system. Lucas: That's the Achilles' heel. The current generation of systems uses a voting mechanism — you need at least two sensors in the same zone to agree before the valve closes. That reduces false positives to nearly zero, but it adds latency. The Ohio utility's setup requires three consecutive 30-second readings above threshold before triggering. So you're looking at about 90 seconds from leak onset to shutoff. That's fast enough to prevent an explosion but slow enough to avoid nuisance shutdowns. Luna: And regulators are starting to take notice. The Pipeline and Hazardous Materials Safety Administration — PHMSA — proposed a rule in late 2025 that would require leak detection systems on all new distribution pipelines in high-consequence areas. That's going to accelerate adoption. Lucas: It'll be a tailwind for sure. But one challenge I hear from utilities is the data overload. A single sensor sends about 2,880 readings a day. Multiply that by 200 sites and you get roughly 576,000 data points every 24 hours. The utility's existing SCADA system wasn't built to ingest that volume. So they had to build a thin middleware layer that normalizes the data and only alerts on anomalies. Luna: That's a classic IoT integration problem. The hardware works, the cloud works, but the middle layer of data plumbing is where projects stall. Lucas: Exactly. And the solution Sensitron used was an edge gateway that runs the anomaly detection locally, then only sends an alert and a summary to the cloud. That also reduces the connectivity cost because they're not transmitting raw data every 30 seconds from every sensor. The gateway aggregates readings and sends a compressed batch every 15 minutes. Only when a sensor flags a potential leak does the gateway switch to real-time mode. Luna: So the edge gateway is doing the heavy lifting on data reduction. That's a pattern we've seen in other industrial IoT contexts — predictive maintenance, vibration analysis — but it's nice to see it applied here. Lucas: It's the same principle. And it makes the business case stronger because you're not paying for cellular data for every sensor. The gateway uses a single LTE backhaul. The sensors communicate with the gateway via a low-power sub-GHz radio, like 900 MHz ISM band. That range can cover a mile in open air, and it's free spectrum. Luna: I want to ask about the climate angle. Methane has about 80 times the warming potential of carbon dioxide over a 20-year period. So even small leaks matter a lot. Is there any data on how much methane these sensor networks are actually preventing from reaching the atmosphere? Lucas: There's a study from the Environmental Defense Fund that estimated that if all U.S. distribution utilities deployed similar sensor networks, they could reduce methane leakage by 40 to 60 percent. That's about 5 million metric tons of CO2-equivalent per year. To put that in perspective, that's roughly the annual emissions of a million cars. And the cost per ton of methane abated is about $15 — compared to $50 to $100 for many other climate mitigation strategies. Luna: So it's both a safety and a climate win. And the ROI is there even without a carbon price. Lucas: Exactly. That's why I think this is one of the most underrated IoT applications. It's not flashy — no one builds a consumer product around it — but it saves lives, saves money, and cuts emissions. And it's happening right now, not in some pilot five years out. Luna: Alright, episode 100 in the books. Next week we're looking at something a bit different — I think Lucas has an angle on IoT in agriculture that goes beyond soil moisture sensors. Lucas: That's right. We'll talk about autonomous weeding robots and the computer vision models that let them tell a weed from a crop without spraying herbicide everywhere. Until then.