Fuzzy neural fault detection and isolation

The paper focuses on the application of fuzzy neural techniques in fault detection and isolation of single failures in greenhouses. The objective of this paper is to detect and isolate faults in greenhouses, with emphasis on faults occurred in actuators and sensors. The developed method is based on a comparison between the measured greenhouse climate and the predictions of a reference 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 very short time. This approach is not limited to greenhouse applications but there is a broader range of future application, especially in livestock housing, growth chambers and poultry houses.