Avitas, NVIDIA team up for high-tech UAS inspections

By Patrick C. Miller | September 13, 2017

When it comes to inspecting energy infrastructure and industrial facilities, it’s no longer enough to have an aerial image from a drone.

Avitas Systems—a GE Venture—is partnering with graphics processer manufacturer NVIDIA to use some of the latest advances in artificial intelligence and deep learning to optimize unmanned aircraft and robotic systems for inspections that improve the detection of defects.

Alex Tepper, founder and head of corporate and business development for Avitas Systems, said companies in the oil and gas, electrical transmission and transportation industries spend hundreds of millions of dollars each year to inspect and maintain their infrastructure equipment.

“What was strange to us is that most of the way they’re doing this is by manual means for the detection of problems,” he said.

These inspections methods range from hand-carrying sensors to looking down from a helicopter to scaling a structure to inspect it.

“We realized this was a very big opportunity to automate the way those inspections are done using robotics,” Tepper said. “For us, that includes drones, robotic crawlers and sub-sea vehicles to do the data collection across a variety of sensors.”

Avitas is using NVIDIA’s DGX Systems and supporting software not only to automate the flight paths of UAS doing the inspections, but also to fuse the data gathered from other sensors and industrial assets. Based on analysis of the fused data, algorithms can detect flaws, predict problems and determine which equipment is most at risk.

“We’re using artificial intelligence to do what we call automated defects recognition, which is interpreting the sensor information from the robots to detect defects of various kinds—things like corrosion, hot spots, cold spots and micro fractures,” Tepper explained. “The whole goal is to optimize the way a company allocates its resources for inspection and maintenance.”

For example, Tepper said that rather than following a rigid schedule of inspecting industrial assets every two years, a risk-based assessment can determine which assets need to be inspected more frequently than others.

Avitas uses NVIDIA’s DGX Station under harsh conditions to provide a level of computing power not typically available in the field. Jim McHugh, vice president and general manager of DGX Systems with NVIDIA, said the whisper-quiet, super computer workstation is about the equivalent of a half row of servers in a data center that can sit on a desktop or be deployed in a van or on an oil rig.

“We truly believe that people are using this to do a new form of computing, a new way of developing software,” McHugh said. “What Avitas has shown here is a great example of how that can work, not only at the data center, but also in the field out at the edge.”