G. Calderón-Espinoza, J. Armengol, and A. Aldea (Spain)
Analytical redundancy, Interval models, Uncertainty, Fault detection and diagnosis
When analytic redundancy is used for fault detection, a mathematical numerical model of the system is required. However, often false alarms occur due to errors in the mod elling process. When knowledge about the system behav ior is imprecise or partially known, interval models may be considered. Interval models make explicit uncertainty in parameters and measurements of sensors. In addition, interval simulation generates numeric envelopes which are formed by all possible behaviors of an uncertain system. This paper describes a system that deals with fault detec tion by using Modal Interval Analysis (MIA). The pro posed system avoids false alarms because when a modal interval simulator is used if a fault is detected, that means that the fault exists. Another advantage of using MIA is the reduction of spurious solutions often presented in semi qualitative methods which may cause missed alarms. Once a fault has been detected, an analysis of tendency is pro posed for fault isolation. This analysis consists of ob serving if the deviation of the variables values are toward smaller values or toward bigger values than those included in the envelopes of normal behavior. The proposed system has been tested with a simple electric circuit and some pre liminary results are analyzed.
Important Links:
Go Back
