Wiring Diagnosis using Time Domain Reflectometry and Random Forest

Abstract

In this paper a new wiring network diagnosis approach dedicated to embedded system based on Time Domain Reflectometry (TDR) response is proposed. The method is based on two complementary steps namely the forward and inverse models. The forward model is used to generate the TDR response using RLCG circuit model and Finite-Difference Time-Domaine (FDTD) method, and to create the datasets. The inverse model allows to detect, localize and characterize the faults from the time response of the faulty network by using Random Forest (RF) technique. Two types of RF models have been used in the diagnosis procedure: RF classifiers and RF regression models. Numerical and experimental investigations have been performed in order to test the performances and the feasibility of the proposed approach.