Interactive diagnosis for a grid network of rain gauges using fuzzy reasoning

This paper aims at designing a diagnosis tool to support experts for detecting and localizing faults in a network of rain gauges. This problem is presented in the context of human–machine cooperation. In this problem, it is impossible to model completely the whole expert knowledge about misbehavior. Diagnostic becomes a process where only a part of the expert knowledge is formalized, the remaining is kept implicit and is exploited gradually during the diagnostic process thanks to interactions with experts. At each step, the proposed diagnosis tool supports the expert by presenting selected data to be analyzed, i.e. rainfall hyetographs for a cluster of rain gauges for which the expert has to identify possible discrepancies. A fuzzy logic based diagnostic reasoning is then used because it proves to be more relevant to express the expert conclusions, which may be dubious. The way of handling such diagnosis processes is presented in this paper.