Techniques for Sensor-Based Diagnosis.
Abstract
This paper describes a system called PDS, a forward chaining, rule-based architecture designed for the online, realtime diagnosis of machine processes. Two issues arise in the application of expert systems to the analysis of sensor-based data: spurious readings and sensor degradation. PDS implements techniques called retrospective analysis and meta-diagnosis as solutions to these problems. These techniques and our experiences in knowledge acquisition in a large organization, and the implementation of PDS as a portable diagnostic tool are described. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- May 14, 1983
- Accession Number
- ADA133297
Entities
People
- Mark S. Fox
- Pamela Kleinosky
- Simon Lowenfeld
Organizations
- Carnegie Mellon University