Wired Network Distributed Diagnosis and Sensors Fusion by Multi-carrier Time Domain Reflectometry and Graph Theory

This paper presents a new application of graph theory to the cancellation of location ambiguity of faults in branched networks. The main objective is to solve ambiguities due to multiple paths and to detect and locate accurately electrical faults. To do so, a distributed diagnosis where several reflectometers (sensors) are implemented at different point of the network with MCTDR (Multi-carrier Time Domain Reflectometry) method is used. The MCTDR signal has a good autocorrelation property that allows detecting and locating more accurately the soft-faults in the wired networks. In this paper, a new method for data fusion between the different reflectometers is presented. The fusion is performed by using MCTDR test signal combined with the graph theory to isolate the nodes surrounding the soft-fault first, and then identify the faulty branch. Based on the different graph theory algorithms and its connection matrix, graph theory seems to be a promising solution to facilitate soft-fault location in the complex wired networksaffected by several soft-faults.