Experimental AR Fault Detection Methods for a Hydraulic Robot

Robots in hazardous environments must meet stringent requirements of both durability and reliability to be safe, as robust systems and efficient reliability techniques are critical to safe operation in these environments. Rosie, a large hydraulic robot designed for nuclear reactor decontamination and dismantlement, is a prime example of such a robot. The study discussed here uses a hydraulic testbed which closely models Rosie's wheel actuators to gather data for development and testing of effective data analysis procedures aimed at reducing the dangers associated with dynamic failures this worksystem and others like it. This paper focuses on practical use and theoretical elaboration of the analytical redundancy technique which is used to efficiently detect faults that have been determined to be mission-hazardous by previous FMECA and fault tree analyses of the Rosie system. We believe we have contributed significant improvements to the potential overall reliability of the system. Additionally, we have expanded the applicability of the AR method to nonlinear systems in the course of our work, making this valuable fault detection method more broadly applicable.