Corrosion Grade on Anchor Rods of Guyed Transmission Towers Applying Machine Committee / Grau De Corrosão Em Hastes De Âncora De Torres De Transmissão Guiadas Comitê De Aplicação De Máquinas

Abstract

The use of guyed structures in electric power transmission lines is a growing practice because of their cost efficiency. However, the anchor systems are subject to corrosion, which can lead to their rupture and loss of tower support. Monitoring the evolution of the corrosion of the anchor rods by visual inspection is a destructive and costly method; therefore, there is considerable interest in developing methods and tools that are capable of generating a maintenance diagnosis of the system. This work aimed to propose a classification tool for guyed towers in terms of the corrosion degree by a machine committee with neural networks and applied it to the Paraiso-Açu line located in Rio Grande do Norte in Brazil. Thirty-eight samples were collected and 33 variables related to the soil corrosion along the line were analyzed. The targets for training the networks were obtained from the inspection of anchor rods taken from the field. A simplification of the problem's dimension was proposed by principal component analysis, describing the phenomenon with 6 variables instead of 33, simplifying the practical application by massively reducing the requirements for data sampling in the field. Several network typologies were trained and the best ones in terms of their generalist and specialist capacities were combined in a machine committee for the final proposal of this work.

Comparison Between Python and Matlab Algorithms for the Evaluation of the Corrosion Grade in Guyed Transmission Towers Using Artificial Neural Networks and Machine Committees / Comparação Entre Algoritmos Em Python E Matlab Na Avaliação Do Grau De Corrosão Em Torres De Transmissão Estaiadas Utilização Comitês De Máquina De Redes Neurais Artificiais

“…In this item, the ANN models that formed two distinct machine committees, implement in Python and in MATLAB R2014a, are described along with the techniques and considerations made in each. The database used was the same as the one in [9] with the addition of the soil resistivity measurements made in the field. The number of samples was 269 towers.…”

“…The current work builds upon the results obtained on a previous paper [9], which presented a method consisting of a machine committee based on ANN with the objective of substituting the current use of visual inspection, that is highly costly and requires the transportation of machines and equipment to remote locations. In order to evaluate the problem several characteristics of the focused upon TL were measured both in situ and with laboratory assessments.…”