Browse Prior Art Database

Expert Systems Diagnosis Via an Assertion-Justification Backtracking Algorithm for Diagnosing Anomalies in Sensor-Based Systems

IP.com Disclosure Number: IPCOM000040045D
Original Publication Date: 1987-Sep-01
Included in the Prior Art Database: 2005-Feb-01
Document File: 3 page(s) / 62K

Publishing Venue

IBM

Related People

Crehan, DT: AUTHOR [+3]

Abstract

This article describes a device to locate failures in sensor-based systems in which the sensors themselves can fail. It describes an algorithm which propagates unique sensor IDs through a network of assertion nodes and justification arcs to localize the assumed assertion about the health of a component which, if false, would explain the anomalous values reported by the sensors. (Image Omitted) The hardware being diagnosed is represented by a network of nodes representing the components and links representing how they are connected. Each component (node) has a state attribute (ON/OFF) and a health attribute (OK/NO GOOD). Each link has a state attribute. Some of the nodes are sensors that report on the state of the links they are attached to and which themselves have a health attribute and a reported-value attribute.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 51% of the total text.

Page 1 of 3

Expert Systems Diagnosis Via an Assertion-Justification Backtracking Algorithm for Diagnosing Anomalies in Sensor-Based Systems

This article describes a device to locate failures in sensor-based systems in which the sensors themselves can fail. It describes an algorithm which propagates unique sensor IDs through a network of assertion nodes and justification arcs to localize the assumed assertion about the health of a component which, if false, would explain the anomalous values reported by the sensors.

(Image Omitted)

The hardware being diagnosed is represented by a network of nodes representing the components and links representing how they are connected. Each component (node) has a state attribute (ON/OFF) and a health attribute (OK/NO GOOD). Each link has a state attribute. Some of the nodes are sensors that report on the state of the links they are attached to and which themselves have a health attribute and a reported-value attribute. These nodes and links form a model internal to an expert system that allow it to have an expectation of what the sensors should be reporting if operation is normal. Actual sensor values from the real system (e.g., a satellite) are compared to these expected values, and when a discrepancy is found, the following diagnosis process is triggered. When an anomaly occurs, the failure is localized to a single piece of equipment by propagating sensor data through an assertion-justification network which shadows the hardware network of links and nodes. A fragment of such a network is shown in Fig. 1. This diagram is a hardware network consisting of two modules C1, C2, three links L1, L2, L3 and sensors S1, S2, S3 on each link. Shadowing the hardware network is a set of assertions linked by justification arcs (Fig. 2).

(Image Omitted)

Assertions describe the logical belief (true or false) in the value of the quantities:
(1) link state, (2) module health, and (3) sensor reported value. Assertions are either assumed or derived; assumed assertions are those assertions which are accepted as true during diagnosis, while derived assertions are those assertions whose truth is deduced from other (assumed or derived) assertions. Assumed assertions deal with the health of nodes (e.g., module (1 has health "OK") and the state of source links, while derived assertions deal with the position of relays, reported value of sensors, and the state of non-source links (e.g., link L2 is in state "ON"). Justifications describe the dependency of one assertion upon another. For example, the assertion that L2 is in state "ON" is justified by the assertion that has health "OK" and by the assertion that L1 is in state "ON." Also, the assertion that S2 is reporting "ON" is justified by the assertion that S2 has health "OK" and by the assertion that L2 is in state "ON." The truth of the assumed assertions is confirmed or disproved by propagating sensor information through the assertion-justification network. In the case of those se...