DISTRIBUTED SENSOR DIAGNOSIS IN TWISTED PAIR NETWORKS FOR SOFT FAULT IDENTIFICATION USING REFLECTOMETRY AND NEURAL NETWORK

This paper aims at developing an approach allowing to detect, locate, and characterize soft faults (i.e., isolation damage) in branched network composed of shielded twisted pair (STP) cables. To do so, a distributed reflectometry diagnosis where several sensors (reflectometers) are placed at different ends of the network is used to maximize the diagnosis coverage. The soft fault identification is achieved by using Multi-Carrier Time Domain Reflectometry (MCTDR) combined with a Multi-Layer Perceptron Neural Network (MLP-NN). The main novelty here lies in the fact that the MLP-NN method is used for data fusion from several distributed reflectometers, which eliminates ambiguities related to the fault location. The required datasets for training and testing of the NN are generated by simulation. Simulation and experimental results are conducted to verify the effectiveness of the proposed approach for locating and characterizing the soft faults in branched networks.