This thesis discusses the design and development of a pipe crawling robot for the purpose of CANDU fuel channel inspection. The pipe crawling robot shall be capable of deploying the existing CIGAR (Channel Inspection and Gauging Apparatus for Reactors) sensor head. The main focus of this thesis is the design of the localization system for this robot and the many tests that were completed to demonstrate its accuracy.

The proposed localization system consists of three redundant resolver wheels mounted to the robot’s frame and two resolvers that are mounted inside a custom made cable drum. This cable drum shall be referred to in this thesis as the emergency retrieval device. This device serves the dual-purpose of providing absolute position measurements (via the cable that is tethered to the robot) as well as retrieving the robot if it is inoperable. The estimated accuracy of the proposed design is demonstrated with the use of a proof-of-concept prototype and a custom made test bench that uses a vision system to provide a more accurate estimate of the robot’s position. The only major difference between the proof-of-concept prototype and the proposed solution is that the more expensive radiation hardened components were not used in the proof-of-concept prototype design. For example, the proposed solution shall use radiation hardened resolver wheels, whereas the proof-of-concept prototype used encoder wheels. These encoder wheels provide the same specified accuracy as the radiation hardened resolvers for the most realistic results possible.

The rationale behind the design of the proof-of-concept prototype, the proposed final design, the design of the localization system test bench, and the test plan for developing all of the components of the design related to the robot’s localization system are discussed in the thesis. The test plan provides a step by step guide to the configuration and optimization of an Unscented Kalman Filter (UKF). The UKF was selected as the ideal sensor fusion algorithm for use in this application. Benchmarking was completed to compare the accuracy achieved by the UKF algorithm to other data fusion algorithms. When compared to other algorithms, the UKF demonstrated the best accuracy when considering all likely sources of error such as sensor failure and surface unevenness. The test results show that the localization system is able to achieve a worst case positional accuracy of +/- 3.6 mm for the robot crawler over the full 6350 mm distance that the robot travels inside the pressure tube. This is extrapolated from the test results completed over the shorter length test bench with simulated surface unevenness

The key benefits of the pipe crawling robot when compared to the current system include:​ reduced dosage to workers and the reduced outage time. The advantages are due to the fact that the robot can be automated and multiple inspection robots can be deployed simultaneously. The current inspection system is only able to complete one inspection at a time.