Fault diagnosis using first order logic tools

Abstract

While numerous diagnostic expert systems have been successfully developed in recent years, they are almost uniformly based on heuristic reasoning techniques (i.e. shallow knowledge) in the form of rules. This paper reports on an automated circuit diagnostic tool implementing Reiter's theory of diagnosis based on deep knowledge (i.e. knowledge based on certain design information) and using first order logic as the representation language. In this approach, the automated diagnostician uses a description of the of the system structure and observations describing its performance to determine if any faults are apparent. If there is evidence that the system is faulty, the diagnostician uses the system description and observations to ascertain which component(s), if faulty, would explain the behavior. In particular, Reiter's method finds all combinations of components which explain this behavior.The inference mechanism which is incorporated as part of the diagnostic tool is based on bidirectional constraint propagation. When all components are assumed to be functioning correctly, the reasoning can be done in a "forward" fashion with the outputs of the components determined by inputs. This represents simulation of the operation of the device. However, when one or more components is assumed to be functioning abnormally, the reasoning task becomes more complex and simulation i s no longer sufficient for discovering contradictions. The output of an abnormally functioning component cannot be predicted from its inputs. However, there may be enough constraints on the operation of the other components in the device that the outputs of the malfunctioning component may be predicted from the behavior and interconnections of the other components. This may require that inferences be made about the value of inputs based on output values of a functioning component. Thus, the reasoning proceeds in both a forward (outputs determined from inputs by simulation) and backward (inputs determined from outputs by inferences) fashion.A prototype version of the diagnostic program which finds all diagnoses has been developed and successfully demonstated on several small but nontrivial combinational and sequential circuits.

A New Centralized Detection-Based Process for Evaluating Anomalies and Analyzing the First Causes Using Machine Learning and Web Semantic

“…Therefore, the anomalies detected are so difficult to interpret and analyze in accordance with the laws and rules imposed by the detection models [9]. DCA techniques are based on detection models inspired ontree structures and logics [10].…”