USE OF ANALYTIC REDUNDANCY IN FAULT-TOLERANT SENSOR SYSTEMS |
|---|
| J. Milek, F. Kraus |
- Abstract:
- The paper presents a systematic design of fault-tolerant measurement systems for linear time-invariant processes. The discussed methods utilize analytic redundancy of the underlying process. Two multivariable linear model types (PCA for static and ARX for dynamic modeling) are considered. Both model types can be directly identified from the measured process data. Analysis of the primary and structured residuals enables very sensitive detection, isolation, and identification of single and multiple sensor and actuator faults. The fault diagnosis is conducted via analysis of the identified faults and can be followed by appropriate fault accommodation actions. A faulty sensor variable can be reconstructed using faultfree process variables and appropriate models. Faults which degrade sensor performance in a recoverable way (c.f. bias and gain errors) can be compensated using model-based sensor recalibration. Best data source in groups of doubled sensors can be selected through application of the voting scheme (where the third sensor is the soft sensor), thus enabling optimal sensor reconfiguration.
- Keywords:
- fault-tolerant systems, fault accommodation, multivariable models
- Download:
- IMEKO-WC-2000-TC7-P197.pdf
- DOI:
- -
- Event details
- Event name:
- XVI IMEKO World Congress
- Title:
Measurement - Supports Science - Improves Technology - Protects Environment ... and Provides Employment - Now and in the Future
- Place:
- Vienna, AUSTRIA
- Time:
- 25 September 2000 - 28 September 2000