Using non-monotonic reasoning to manage uncertainty in railway asset diagnostics

When a system instantaneous fails there is no time to perform preventative maintenance. In certain circumstances, where the system subsequently recovers without intervention, the system is left to run undiagnosed. Increasing the certainty of a particular diagnosis leads to a greater likelihood of the corrective action being carried out. Therefore, moving to a situation where available symptom and associated condition data is used to diagnose a failure is desirable. This paper proposes the implementation of a distributed network of semantic nodes that supports the inference of asset status. Querying is then performed to produce concrete facts regarding the status of the asset. The facts are used to support non-monotonic, probabilistic reasoning to increase the certainty that a particular symptom is the cause of the instantaneous failure.