NDT – Autonomous robotic system for ultrasonic and eddy current inspection of metal and composite parts of complex shapes

Partners: LA composite s.r.o., ATG s.r.o.
Duration: 2021-2024

The project goal is research and development of technologies of an autonomous robotic system for inspection of composite materials, spot welds, and rivet joints on parts of complex shapes using ultrasonic and eddy current methods. Autonomy will be provided by a developed navigation method, which combines image and ultrasound/eddy current data.

The partners expect to utilize the project outputs to support their production, services and sales. LAC will employ the developed equipment for output inspection of composite parts in its production and as a customer service. ATG will use the technologies as a new sales item for customers requiring autonomous and advanced technologies for output control. CVUT will acquire new knowledge and technologies for future research, development, and teaching.

This project is co-financed from the state budget by the Technology Agency of the Czech Republic within the TREND Programme under grant agreement No FW03010600. (STARFOS TACR)




The IMR group is developing a navigation system for the robotic inspection of complex shapes using a 3D model of the part and a camera system. The inspection points are a list of oriented positions corresponding to the defined part 3D model. The goal of the inspection task is to navigate an inspection probe to each inspection point by the robot to measure the location of the part.

The main problem of the inspection task is that the part model never precisely corresponds to its dimensions in real-world conditions. The origins of these imperfections are production deviations given by production tolerances and the imprecise position of the part relative to the robotic system.

The developed system demonstrates the ability to mitigate the accuracy problem and precisely navigate the inspection probe to the inspection point using feedback from the camera system. Deep convolutional neural nets are utilized to detect and classify inspection points in the camera image. The developed system can be universally applied to any measurement method.

The system's modular architecture allows for easy future functionality and application domain expansion. The navigation system includes methods for optimizing the inspection trajectory, thus reducing the time needed to inspect all target points. Furthermore, the system provides managed storage for part models with defined inspection points and models of the robot environments. Part and environment models can be arbitrarily paired. Therefore once described, the system can reuse the part model on multiple robotic work cells without change. The developed software solution can be deployed on arbitrary hardware consisting of a robotic arm and a color camera.

Video demonstration:

Page last modified on January 31, 2023, at 11:30 AM EST
Powered by PmWiki