Multi-sensor fault detection and diagnosis using combined qualitative and quantitative techniques

This paper describes a fault detection and diagnosis technique for multisensor data fusion systems with an application to process plant monitoring. In particular this work concentrates on three aspects: 1) how the qualitative reasoning techniques are combined with conventional numerical sensor fusion techniques is investigated and a method of developing a combined semi-quantitative model is introduced; 2) how the semi-quantitative information is analysed to perform the system monitoring and fault detection is discussed; and 3) the implementation of the diagnosis using fault model free methodology is introduced. A constraint violation based approach is described. The application of the algorithms developed is demonstrated through a number of examples, based on real data.<<ETX>>