Industrial Diagnostics using Algebra of Uncertain Temporal Relations

Industrial diagnostics is an important application area for many AI formalisms. Temporal diagnostics, based on analyzing temporal relations between values of crucial variables, is one possible approach to industrial diagnostics. Often, the information obtained from an industrial object can be uncertain, making the task of diagnostics more complex. In this paper, we propose an approach to temporal industrial diagnostics, which uses algebra of uncertain temporal relations. We estimate temporal relations between the set of symptoms (crucial values of important variables) obtained from an industrial object to build the temporal relational network for this particular situation. After that, we compare the obtained network with known temporal scenarios (patterns) of failures, using the numerical measure of the distance between a network and a scenario. Using this approach we derive the probabilities of possible diagnoses for the particular situation. We also show how the learning for the database of scenarios can be performed, which will make diagnostics for future cases more precise.