Presently, remotely deployed automated inspection systems, used to inspect pressure vessels and cooling circuit components, tend to be dedicated to just to one specific task and cannot be used for any other purpose. For example one might have to inspect a nodal weld joint in the side of a coolant feeder pipe. Current practice is to use a sensor manipulator which is constrained to move along some sort of rigid guide rail built round the weld line. Such a guide system cannot be adjusted to fit welds in nodal joints of different angles in pipes of different diameter. Indeed it is quite common for as many as 6 different manipulators to cover all the different kinds of weld geometry occurring in a cooling circuit. Given the variety of types of nuclear reactor in commission throughout the world, the number of single task inspection manipulators (they could also be called fixed or immobile robots) must be very large. Even though inspection costs are very small compared with the overall running costs of a reactor, the fact remains that a proliferation of inflexible, one - off inspection tools is not very cost competitive. Furthermore if an urgent need arises for the inspection of a previously untested weld geometry there could be problems in prototyping a new inspection tool to deadlines. However the main shortcomings of some existing automated inspection tools is the time it takes to position them during which human operators are exposed to radiation.

With this background the aim of the authors' research described here is to create a generic instrument for nuclear power plant inspection, which can inspect many kinds of complex weld contours with few or no modifications and can be rapidly placed in position. This idea is to be realised by a mobile robot vehicle which can climb along and around pipes of different diameters and any orientation, whilst supporting a 7 axis robot arm which can scan an NDT sensor array around the complete circumference of a nodal joint weld. As many pipe structures in nuclear plant are made of stainless steel, vehicle adhesion and stability has to be achieved with pneumatic suction rather than magnetic adhesion, and this greatly complicates the design.

Initial trials on a first prototype vehicle are described here. To minimise the substantial prototyping costs the vehicle was designed round standard off-the shelf components. Movement in any direction over highly curved surfaces, including spheres as well as pipes, is achieved by means of two orthogonal walking mechanisms, each providing forwards/backwards using 8 legs and having 3 key design features. Firstly, thigh hinges which can tilt leg pairs relative to the rigid vehicle payload platform (chassis). Secondly, universal ankle joints which can be made alternately free during a walking step, but otherwise locked rigid for vehicle stability during the data acquisition stages. Thirdly, suction feet which can adapt and adhere to curved surfaces whilst remaining sufficiently rigid for vehicle stability. In all there are 12 axes of linear motion, 4 axes of rotation and 8 universal joints involved in the walking process.

Translational motion, raising and lowering of feet and thigh tilting are achieved entirely by pneumatic power and the pneumatic control circuit involves 52 proximity sensors and a minimum of 8 vacuum sensors depending upon the number of suction cups employed. To achieve sufficient vehicle stability the feet must be rigid and thus shaped to fit a given surface curvature. A range of curvatures is then accommodated by having a range of feet of different contour which can be rapidly exchanged via a single bolt on procedure. Vacuum adhesion is achieved by the use of a thin rubber skirt round the perimeter of the feet or by attaching a number of smaller, standard rubber suction cups to the rigid feet. The latter option reduces the suction area and thus the payload capacity but increases the range of surface curvatures that a foot with a given nominal curvature can tolerate. With this range of foot options the best specifications of the vehicle is that it can carry a payload (on-board plus umbilical control cable) of 38 kg with a safety factor of 3, over surfaces of any diameter down to 1 metre. The envisaged payload is a 7 axis robot arm of 22 kg whose end-effector can scan an ultrasonic probe array with a repeatability of 1 mm with a load capacity (array mass plus contact pressure) of 45N. The vehicle mass, without payload and without optional mechanisms for walking over spheres, is 37 kg.

Ideally, design improvements are now needed to reduce this to 30 kg so that just two operatives can position the vehicle without mechanical assistance and within typical safety regulations on manual handling.

This work forms part of a Brite Euram framework 4 project in the Plant life assessment thematic network.