Structural Health Monitoring (SHM) is the continuous acquisition of data recorded on an industrial installation to monitor the integrity of the installation and to localize potential flaws. This data acquisition could be achieved by a well-known method called Electrical Time Domain Reflectometry (ETDR), which consists in sending a high frequency electro-magnetic pulse within a transmission line to analyze the backscattered signal. This backscattered signal represents the impedance variations along the line. These impedance discontinuities can be induced by a variety of property changes, either within the sensor or in the surrounding environment, like local defects (due to mechanical strain for instance) or property perturbations (such as temperature or hygrometry fluctuations). As a consequence, the use of the ETDR method for industrial applications relies on a relevant signal processing, in close correlation with the knowledge of material properties.

The aim of this study is to investigate the ageing of different sets of materials possibly used for the fabrication of ETDR systems: alumina as a dielectric and various metals (Pt, Pt-based alloy, W, Ni-based alloy,…) as conductor lines. In order to relate the backscattered signal to material properties, a systematic design of experiments has been performed: ETDR signal was then related to the physico-chemical evolutions of the materials. The experimental data obtained were then coupled to an analytical modelling of the transmission line. This model considers intrinsic material properties such as permittivity, resistivity, and permeability, and take into account the geometry variabilities such as coaxial geometry, twin lead geometry, or interspace between the conductors of the sensor.

The objective of this work is twofold. The first is the design of the transmission line, in order to have the most efficient and usable signal for post-processing. The second goal is to have a numerical simulator of backscattered signal in order to build a wide dataset to implement processing by machine learning.
 

By coupling these combined experimental and modelling approaches, the determination of the most efficient materials (dielectric and conductor) and geometries will be discussed, as well as the impact of the environment (such as temperature, hygrometry or strain) on their behaviors.