Soft fault detection in cables using the cluster time-frequency domain reflectometry

Faced with the continual increase in complexity of wiring networks, their reliability, in particular in embedded fields (such as automotive and aerospace industries) becomes a major issue. In wiring diagnostics, reflectometry methods are commonly used. They are quite efficient for detecting important damage (hard faults) such as short-or open-circuits. These faults can have heavy material (e.g.: fires) and economical consequences. Then detecting them before they happen would be very interesting and useful. Unfortunately no current method is efficient enough for addressing soft faults. This article proposes an original approach to overcome this problem: the Cluster Time-Frequency Domain Reflectometry (CTFDR). Based on the fact that a wire is most often in a bundle with several other wires, this method takes advantage of the resultant near end crosstalk signals to get more information about the state of the wires in the bundle. To make the detection of incipient faults easier, a normalized time-frequency cross correlation function has also been used.