Cubic Paraconsistent Analysers with Evidence Filter and Temporal Analysis

Abstract

Ambiguous or inconsistent information is frequently present in the input of any real-world system that relies on multiple data sources. In the presence of inconsistencies, an expert system based on classical logic is trivialised due to the principle of explosion, that forces any proposition (and its negation) to be inferred. Instead, a family of non-classical logics, called Paraconsistent Logics, formalises the idea that even from inconsistent premises, useful conclusions can be drawn. Within this context, the Paraconsistent Annotated Logics with Annotation of Two Values (PAL2v), which use a pair of favourable and unfavourable evidences as annotations, have been particularly successful in engineering applications. Structures called Para-Analysers can be built with these logics that provide the appropriate means to analyse the truth value of a proposition under conflicting/contradictory evidences. The work reported in this paper extends the traditional Para-Analysers with an evidence filter. This new model can be represented as a cubic structure representing various Para-Analyser lattices, each of which is selected according to the quality of the evidence, that is updated with respect to measurements executed at run time. The resulting three-dimensional paraconsistent analysis structure is encapsulated as an expert-system cell called Cubic Paraconsistent Analyser with Evidence Filter and Temporal Analysis (CPAet). Sets of CPAet nodes are combined constituting distinct network topologies that are used to infer the operating conditions of equipment in data networks of electrical systems control and supervision, resulting in a fault detection and classification system. Tests on simulated systems show that the CPAet networks were effective to deal with inconsistencies without trivialising the inferences, while also providing a more informed (finer grained) decision on the data network equipment state, when compared to the traditional paraconsistent analysers.