If you’ve ever driven by a petroleum refinery, you’ve probably seen a flare stack “on fire.” What looks like shooting flames are the waste gases produced during chemical manufacturing burning off. At the stacks’ mouths, the heat can reach into the hundreds of degrees, and the metal is under constant, intense pressure. The slightest crack or the smallest amount of corrosion in the wrong place could cause a fire that would shut down the plant — not to mention endanger the lives of the people who work there. “It’s basically a blowtorch,” says John Spirtos, managing director of New Business Creations at GE Ventures. “You don’t want to get too close to it.”
But the stacks require regular inspections for regulatory and operational purposes. That means inspectors must get close enough to discover what are often tiny traces of wear that can lead to gas leaks and performance issues with related systems. It’s a dangerous, time-consuming and ultimately expensive task. Traditionally, these types of assets are inspected only after the asset and related systems are shut down, allowed to cool and prepared for inspection — which is a slow and costly endeavor that often involves scaffolding, rope inspection teams and visual analysis. The inspection itself can last several days, and the necessary repairs can add on several days more. By the time the process is completed, downtime is often measured in weeks.
But Spirtos’ team is now exploring another option. It wants to use ground- and air-based robots or drones equipped with special sensors to check the stack while it’s still active. These robots can take readings and video footage of a variety of industrial assets. The data is then processed through a series of inspection algorithms and also made available to an inspector, who can choose to be on-site or seated comfortably in a monitoring room.
Because the information is recorded, inspections can be a lot more cooperative and thorough. “We can have multiple people look at the same asset at the same time,” Spirtos says. “You can also interact remotely with the inspection as it takes place. You can freeze the inspection video while it is being taken, move backwards in time, move forward to the current inspection, annotate the images of the inspection and send them to other experts if something is identified as needing further attention.”
The data coming back from the drone enables operators to create a 3D model of the asset and track changes over time through subsequent inspections. For example, they can see if a piece of metal is starting to expand or buckle or if a miniscule crack is growing. Because robotic inspections don’t require a plant to shut down entirely, they can happen more frequently, leading to earlier detection of potential problems. GE estimates drones and other robots can cut the amount of time needed to provide inspection by as much as 50 percent, depending on the assets being inspected.
GE Oil & Gas also can combine the inspection data with input from fixed sensors placed on various assets and handheld sensors carried by technicians to get a much more complete picture of plant health. The data and image analytics platform runs on Predix, GE’s software platform for the Industrial Internet. “We take historic data from different sources, combine it with surface analysis, and we create sort of a mosaic of how the asset actually works,” Spirtos says.
Although the program is still only in a pilot phase, Spirtos’ team has already conducted inspections at five industrial facilities and says the results are very encouraging.