Inspecting industrial equipment such as oil rigs and electrical towers is a messy, expensive, time-consuming and dangerous undertaking. Nvidia Corp. and General Electric Co.’s Avitas Systems are looking to change all that.

The two companies today announced that Avitas, which was formed by GE Ventures about four months ago, will use Nvidia’s graphics chip-powered computers on a variety of tasks involved with industrial inspection. That includes creating optimal flight plans for drones (pictured) and subsea vehicles to do inspections instead of humans, as well as doing data analytics for automated recognition of equipment defects such as fractures or hotspots.

Currently, “they’re literally having people do this manually, via truck or helicopter,” Alex Tepper, managing director at Avitas, said during a conference call with reporters. And sometimes that delays critical inspections, such as on flare stacks for burning off waste gas at refineries, which must be shut down for days to be cool enough for humans to inspect.

Drones can get much closer much more quickly to gear at refineries, power plants and underwater pipelines to detect corrosion, leaks and other defects. They also can gather much more information, including video, photos, infrared images and temperature, which then is combined with operational plant data and information from human inspectors and fed into Predix, GE’s cloud platform for the analysis of data from industrial machines.

Indeed, it’s all that data that is most critical to applying artificial intelligence and machine learning powered by Nvidia’s DGX systems. The graphics processing unit chips used in the systems can process a lot more data in parallel compared with standard chips such as Intel Corp.’s, making them a standard for most companies and researchers doing AI and machine learning.

Avitas employs deep learning, a branch of AI that uses artificial neural networks to enable machines to learn without explicit programming. That allows the company to find defects that aren’t obvious to the human eye (inset).

“We’re using artificial intelligence to do automated defect recognition,” Tepper said. “It allows us to truly understand the state of [an] asset — predictively as well.”

One DGX machine is in central data centers for the heavy data lifting, but the DGX Station announced in May can be used out in the field to process all that data near the source to save time. Avitas currently is using prototypes of the machine, which is soon to be shipped in production using a powerful new Volta chip from Nvidia, said Jim McHugh, Nvidia vice president and general manager of the DGX systems business.

Images: Avitas Systems