Self-evolving, advanced test stand reasoning for closed loop diagnostics

Abstract

Advanced test stand reasoning can yield dramatically increased Line Replaceable Unit (LRU) operational availability through reduction of Mean Time to Repair (MTTR) while offering better utilization of related maintenance and test equipment resources. However, technical challenges related to development of reasoning systems, typically requiring expert LRU domain knowledge, present hurdles that can be prohibitive. Reasoning methods developed using simple pattern recognition and repair instance statistics offer a first order approach that is functional from a proof of concept perspective. Unfortunately, the small numbers of statistics available historically provide reduced reasoner reliability and effectiveness when confronted with dynamic and complex avionic systems. This paper discusses the issues, trade-offs, and potential benefits to be gained through the application of robust, self-evolving, hybrid reasoning techniques. A reasoner that can utilize and leverage the constraints found in typical test stand procedures to provide a best, safe path to diagnosis, while learning and optimizing in-situ, may offer an ideal, scaleable solution for optimizing test stand operations. Related information management and diagnostic model visualization techniques are also presented in the context of diagnostic and avionic system evaluation and improvement.