FUZZY NEURAL FAULT DETECTION AND ISOLATION

Abstract

The paper focuses on the application of fuzzy neural techniques in fault detection and isolationof single failures in greenhouses. The objective of this paper is to detect and isolate faults ingreenhouses, with emphasis on faults occurred in actuators and sensors. The developed method isbased on a comparison between the measured greenhouse climate and the predictions of areference model. This comparison is performed according to the knowledge-based approach,which ensures robustness of the diagnosis with respect to noise and modeling imperfections.Using the method developed, all the failures investigated are detected and isolated, within a veryshort time. This approach is not limited to greenhouse applications but there is a broader range offuture application, especially in livestock housing, growth chambers and poultry houses.