Robotics and AI systems for dust detection and decontamination of Tokamak systems

  • Neil Harrison

Student thesis: Master of Philosophy


The project aim was to produce a system to carry out autonomous scanning, detection and removal of beryllium oxide dust using a robotic manipulator. The intended use of this system is for maintenance in nuclear fusion reactors, where plasma-facing components made from beryllium erode to form beryllium oxide (Be-O) dust when struck by the plasma. As Be-O is toxic if inhaled, the components removed from a reactor must be decontaminated before handling can be done safely. Currently this is a manual swabbing process which is time-consuming, susceptible to human error, and inherently dangerous for the operator. This project aimed to improve on this process by using a robotic manipulator in conjunction with a Be-O sensor to scan components and identify the presence of Be-O autonomously. Throughout this project, progress was made towards the overall aim in three main areas; inverse kinematics, path planning, and over-actuated manipulators. A variety of inverse kinematics algorithms were investigated and implemented to learn each algorithm’s strengths and weaknesses. Including obstacle avoidance within the inverse kinematics algorithm was explored, with a conference proceedings written which compares obstacle avoidance algorithms. The path planning process of scanning using a manipulator was reviewed. The need for a path generation algorithm capable of generating a scan path on an arbitrary shape was identified. An algorithm for automatically generating a scan path across the surface of a component was developed. The scan path defined by this algorithm uses the sensor requirements as constraints which are maintained throughout the path. This algorithm provided good coverage on arbitrary complex shapes and successfully identified ‘blind zones’ which were impossible to scan with the modelled sensor. Suggestions are made for taking this algorithm further and incorporating it with a robotic manipulator system. A prototype wrist mechanism was developed for use in conjunction with a robotic manipulator with the aim of increasing the workspace volume and dexterity of the overall manipulator. This wrist mechanism provides two additional degrees of freedom, resulting in an over-actuated manipulator with eight total degrees of freedom. A working prototype was designed and built to be integrated with a Universal Robots UR5, with the mechanism providing full hemispherical movement.
Date of Award31 Dec 2022
Original languageEnglish
Awarding Institution
  • The University of Manchester
SupervisorBarry Lennox (Supervisor), Guido Herrmann (Supervisor) & Joaquin Carrasco Gomez (Supervisor)

Cite this