Estimating fatigue behavior of a family of aluminum overhead conductors using ANNs

Abstract

This study aimed to create an artificial neural network (ANN) architecture capable of estimating the fatigue behavior of aluminum overhead conductors, considering specific weight (W) and bending stiffness (EI) as parameters of influence. ANN training and testing is conducted by using a dataset obtained from fatigue tests carried out in a 50 m resonant bench at the University of Brasilia (UnB). ANNs are used to construct constant life diagrams for this family of conductors, and to compare the results obtained experimentally. Our findings show that for the architectures analyzed, it is possible to accurately estimate the fatigue behavior of this family of aluminum conductors.

Fatigue modeling using neural networks: A comprehensive review

“…Generating data for ANN training, based on prior knowledge of the problem (trivial data such as the average stress and number of cycles when the alternating stress is zero) to train the ANN is an interesting solution. This is done in previous works 74,76–78,91,107,108 …”

“…The number of six stress ratios was presented in Júnior et al, 74 and there were significant improvements (using MNN) where it is possible to obtain generalization in the results and with three stress ratios 75 . A similar approach as Júnior et al's 74 approach is applied to predict that the fatigue lives can be seen in the work conducted by Pestana et al, 76 Kalombo et al, 77 and Cãmara et al 78 Jimenez‐Martinez and Alfaro‐Ponce 79 used FNN to predict the fatigue life of chassis component subjected to sequence loading and different temperatures. The fatigue life is used as one of the inputs.…”