In this paper, the problem of fault diagnosis in drinking water transport networks\ud(DWTNs) is addressed. Two different fault diagnosis approaches are proposed to deal\udwith this problem. The first one is based on a model-based approach exploiting a-priori\udinformation regarding physical/temporal relations existing between the measured variables\udin the monitored system, providing fault detection and isolation capabilities by\udmeans of the residuals generated using these measured variables and their estimations.\udThis a-priori information is provided by the topology and the physical relations between\udthe elements constituting the system, which is used by design in order to derive\udfault diagnosis. Differently, the second approach does not require the physical a-priori\udinformation of the network to operate. It relies on a data-driven solution meant to exploit\udthe spatial and temporal relationships present in the acquired data streams to detect\udand isolate faults. Relationships between data streams are modelled through sequences\udof linear dynamic time-invariant models whose estimated coefficients are used to feed\uda Hidden Markov Model (HMM). When the pattern of estimated coefficients cannot be\udexplained by the trained HMM, a change is detected. Afterwards, a cognitive method\udbased on a functional graph representation of the system isolates the fault. Finally, a\udperformance comparison between these two approaches is carried out using a part of\udthe Barcelona water transport network.
展开▼
