A Novel Methodology to Detect Faults on Anchor Rods Using Reflectometry and Machine Learning

Abstract

This paper presents a new methodology based on a FDR technique and machine learning structure to detect faults on anchor rods used in guyed towers of power transmission lines. Such faults are due to material loss in the rods, which it may break them, and may case the tower falling. The methodology is a non-destructive technique (NDT) that is based on frequency domain reflectometry (FDR). For detection, high frequency pulses are inserted into anchor rod samples and the reflection coefficient is analyzed and compared to that obtained in the simulation results. Samples of anchor rods with and without faults were simulated, fabricated, measured and classified by a machine learning structure to determine the presence or absence of the faults.