Robust model-based failure detection and identification in greenhouses

A model-based method for the detection and identification of single failures in greenhouses is presented. The method relies solely on climatic measurements currently available in commercial greenhouses, and combines hybrid physical/neural-network models with robust failure detection and identification theory for non-linear systems. Both sensor and actuator failures are considered, and the detection of crop water stress is also addressed. The first part of the paper is devoted to a simulation study to estimate the economical cost of the failures. In the second part, a method for robust failure detection and identification is described and tested on experimental data. The application of the method to experimental data shows that, under most circumstances, the failures are correctly detected and identified, leading to a significant reduction of the losses caused by the failures.