Detecting faults on the anchor rods of power transmission line towers is an essential procedure for guaranteeing the safety, operability, and availability of the electric power system. Such a measure aims to minimize the risk of tower collapse and the associated costs due to the lack of energy, injuries to people, and damage to the reputation of the companies involved. In this letter, we report an alternative noninvasive and nondestructive detection technique through the acquisition of electromagnetic scattering parameters. A novel machine learning (ML) approach is proposed in order to achieve the advantages of such a technique and to deal with the respective drawbacks due to the complexity of signals acquired. A strategy to artificially increase the number of relevant samples in the dataset through the simulations of computational models is performed. Different ML classification algorithms are compared in the task of detecting structural faults in anchor rods in order to define the most suitable to this application. Finally, experimental results show that the methodology proposed achieves a high detection accuracy, surpassing other known methods.
Structural faults on anchor rods due to corrosive processes are among the major natural causes worldwide for the collapse of guyed towers. As such towers are widely used on both power transmission lines and antenna systems, their collapse may cause unexpected interruptions of telecommunications and power delivery services. Despite being of utmost importance, underground corrosion monitoring is still a technological challenge, because it is a highly nonlinear problem, influenced by several soil characteristics and its complex iterations with the metallic structure of the rods. Some techniques have already been proposed to locate such faults; however, either they require previous knowledge about the medium or they need the interpretation of the acquired signal by a human specialist. These requirements make it difficult to use such solutions in automatic monitoring as well as to make their integration with Internet of Things (IoT) systems. In this article, a novel technique is proposed to automatically locate structural faults on anchor rods, based on a two-step machine learning (ML) solution. The system is fed by a vector composed of frequency-domain samples of the input impedance electromagnetic (EM) signals measured on the rods through a dedicated designed connector. Experimental results have demonstrated that the proposed combination of EM signal acquisition and ML data processing was able to perform the fault location on anchor rods accurately and with a total independence of any additional human analysis.
Eye bolts are part of a system used to anchor electrical insulators in substations and on transmission lines, connected by concrete structures. Structural failures in these devices may lead to service interruption, financial losses and risks for the general population. In this article, we propose a signal adapter capable of applying and receiving high frequency signals to a portable vector network analyzer for the eye bolt. The technique consists of injecting high frequency signals into the samples in order to detect their integrity and analyzing their return loss (S11), using the principle of reflection and transmission lines. Should the eye bolt present a failure, the ∣S11∣ response is changed, thereby enabling its detection. The development of the adapter, for both intact and defective bolt models, was based on electromagnetic simulation models using the Ansys HFSS program. The measurements were made with models of real bolts presenting approximate wear, in order to validate the simulations.
1 March 2022 in Journal of Microwaves, Optoelectronics and Electromagnetic Applications