Comparison based diagnostics as a probabilistic deduction problem

A novel modeling approach for system-level diagnosis of multiprocessor systems has been introduced in previous publications. In this approach the diagnostic process is formulated as an optimization problem. The possible logical relations are identified between the different pieces of diagnostic information of the system and an optimal consistent combination of the relations is determined during the solution method. A part of the information is that, which can be observed at the outputs of the system. Another part is composed by hypotheses on the states of units. Relations between these information are described by consequence rules having probabilities assigned to them. These probabilities express the uncertainty of test results. The object is to draw back the set of observed information to a subset of hypotheses on unit states with the maximum likelihood, i.e. to determine the states of system units on the basis of the syndrome.