Sensors Selection for Distributed Reflectometry-based Soft Fault Detection using Principal Component Analysis

In this paper, a new approach is introduced for the selection of the most relevant sensors to monitor and diagnose soft faults in complex wired networks. Although reflectometry offers good results in point to point topology networks, it introduces ambiguity related to fault location in more complex wired networks. As a solution, distributed reflectometry method is used. However, several challenges are enforced, from the computing complexities and sensor fusion problems, to the energy consumption. In this context, the proposed method combines Time Domain Reflectometry (TDR) with Principal Component Analysis (PCA). It is applied to a Controller Area Network (CAN) bus connected in a network structure in which sensors perform reflectometry measurements consecutively. The simulated TDR responses are then arranged into a database. With this latter, a PCA model is developed and used to detect the existing soft faults. Coupled with statistical analysis based on Hotelling T-2 and Squared Prediction Error (SPE), the most relevant sensors for monitoring and diagnosing soft faults occurred in the network are identified with high accuracy.