Evaluating the Robustness of Autonomous Inspections in the Energy Industry with a Quadruped Robot

Abstract

Abstract This pilot work evaluates a robotic solution for automating inspection activities on a mineral lube oil console used to lubricate gas turbines. A series of autonomous inspection missions have been carried out at the Baker Hughes site in Florence (Italy). A quantitative analysis of robot performances in tasks that are critical for accomplishing routine autonomous missions – such as locomotion, step climbing, visual inspection, recharge and docking - has been conducted and discussed. To assess the robustness of the quadruped robot, pre-programmed autonomous missions have been carried out to perform robotic visual inspection activities on the package of a gas turbine. The walkway was characterized by different ground conditions including climbing up and down steps throughout the path. We evaluated the robot's performance by measuring self-localization precision estimated with QR tags, robustness under different environmental conditions, locomotion quality, and success rate in step climbing behaviors. We carried out 40 autonomous missions, which lasted on average 5’36". Robot climbed steps successfully at the first attempt in 99.16% of cases (down: 99.16%, up: 98.30%). Average localization errors were on average 1.8 cm. Results indicate that this technology is robust enough to perform the autonomous inspection on energy assets with low mitigation action requirements.