Physics-Based ML Tools for In-situ Cable Condition Monitoring through FDR

We propose a machine-learned model to use current Nondestructive Examination (NDE) signal processing techniques data for condition Monitoring of Cable Insulations. Karax aims to develop physics-informed machine-learned tools to use NDE data to approximate the remaining useful life and the loss of material properties such as the impedance of cables insulation. Our goal is to use frequency domain reflectometer (FDR) readings along with insulation resistance and nano-indentation for condition monitoring of aged cables, i.e. to approximate state of decay, locate local damages, and predict remaining service life. Through the precise identification of impedance variations, our models intend to establish a direct correlation between the outcomes of FDR tests and the aging state, attributed to the combined effects of thermal and radiation stressors. The project’s goal is to augment the FDR method’s effectiveness in monitoring cable insulation condition, by creating a comprehensive mapping between aging scenarios and the dielectric properties of the material. This approach aims at enhancing maintenance strategies and potentially increasing the operational lifespan of facilities. Phase I objectives encompass (1) developing decay models to accurately forecast impedance modifications resulting from synergistic thermal-radiation (T+R) aging effects, and (2) applying these decay models within Finite Element Method (FEM) simulations to model the electromagnetic field in cable sections experiencing varying aging levels, with subsequent validation of the reflectometry response against actual FDR test outcomes. Collaborative efforts with PNNL are planned for the validation process.