Application of 2-D Convolutional Neural Networks for Damage Detection in Steel Frame Structures

Application of Deep Learning Techniques for the Impedance-based SHM to the Oil & Gas Industry

“…The convolution operation adopted by a convolutional layer is made from a sliding data window, commonly called a kernel, which is constituted by a mesh of synaptic weights and passes through all the inputs of the evaluated layer. This slippage and abstraction procedure is done to highlight the main local features of each dataset (Ghazvineh et al [2021]).…”

“…Among these functions, we can mention the value choice: maximum, average, and Euclidean norm, among others. However, they aim to represent the analysis region in a single numerical value, transmitted to the next layer through a reduced feature map (Chen, Li and Sanchez [2015], Ghazvineh et al [2021]).…”

“…These structures, when in operation, are susceptible to failure due to wear conditions, fatigue, or impacts of their mechanical components (Moura Jr and Steffen Jr [2006], Rezende, Barella and Moura Jr [2020]). Therefore, for such facilities to perform their activities correctly, the failures present in the mechanical system must be identified and repaired promptly, which makes integrity monitoring processes a critical aspect nowadays (Palomino et al [2011], Eleftheroglou et al [2018], Ghazvineh et al [2021]).…”

“…In addition, it is known that incorrect identification of the presence of damage can have serious economic and safety consequences (Ghazvineh et al [2021]). In this sense, the maintenance of systems considered critical becomes even more fundamental, and, consequently, different inspection and damage assessment techniques have been researched in recent years.…”

“…Due to this increase in computational complexity, recent SHM studies have sought to reconcile integrity monitoring methods with innovative and functional tools for pattern abstraction and classification (Verstraete et al [2017], Ghazvineh et al [2021]). The use of pattern abstraction techniques in the monitoring data sets aims to delimit a causal relationship between the different integrity states of the structures under study (Gulgec, Takáč and Pakzad [2017]).…”