[2A4] Autonomous, scalable, mobile robotic-arm NDT: stitching digital twins with visual artefacts
A Poole¹,², M Sutcliffe¹, G Pierce² and A Gachagan²
¹TWI Wales, UK
²University of Strathclyde, UK
Flexible robotic deployment of NDT equipment traditionally requires an initial step to produce an accurate digital twin of the part. The digital twin is then used with robotic path planning software in order to subsequently deploy NDT sensors. For structures significantly larger than the robot’s workspace, external large-volume profiling devices such as photogrammetric equipment are required for full surface planning and part-robot localisation. For these systems, either a feature-rich surface or static work volume is required for accurate reconstructions. Several processes have investigated autonomous robotic scanning procedures in lieu of a digital twin, either using autonomous full surface profiling methodologies or by using low-cost vision sensors. For all these processes, the reconstruction is limited by the working volume of the robotic or sensor platform, or by the surface’s complexity for accurate relative localisation.
This paper introduces a method for simple and scalable reconstructions of parts larger than the working volume of the robotic platform and repeatable localisation for parts that are featureless or with significant spatial symmetries such as cylinders. This method has been found to provide relative localisation accurate to within 2.5 mm and enables scalable on-site deployment of ultrasonic data stitching to noisy, dynamic industrial environments when scanning large, featureless parts where no inertial information is required.
This paper introduces a method for simple and scalable reconstructions of parts larger than the working volume of the robotic platform and repeatable localisation for parts that are featureless or with significant spatial symmetries such as cylinders. This method has been found to provide relative localisation accurate to within 2.5 mm and enables scalable on-site deployment of ultrasonic data stitching to noisy, dynamic industrial environments when scanning large, featureless parts where no inertial information is required.
