PSI - Issue 37

ScienceDirect Available online at www.sciencedirect.com ScienceDirect Structural Integrity Procedia 00 (2019) 000 – 000 Available online at www.sciencedirect.com ScienceDirect Structural Integrity Procedia 00 (2019) 000 – 000 Available online at www.sciencedirect.com

www.elsevier.com/locate/procedia www.elsevier.com/locate/procedia

Procedia Structural Integrity 37 (2022) 187–194

© 2022 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under responsibility of Pedro Miguel Guimaraes Pires Moreira Abstract Recently, due the automation of many industrial processes the intelligent tools for undertaking decisions started to be highly demanded solutions in many industrial branches. This is also the case in non-destructive evaluation, where big amount of data from inspections need to be analyzed, and based on such an analysis a decision must be undertaken. In many cases, the step of analysis is already automated with using a dedicated software, however, the decision-making process still engages a human. In this paper, the authors presented the preliminaries of the damage classification algorithm, which is intended to classify different types of a structural damage in composites based on X-ray computed tomography scans. The proposed approach was based on deep neural networks, which creates a possibility to obtain high values of a classification accuracy. The obtained results within this study clearly show the effectiveness of the proposed approach and create a path for further development of the algorithm. © 2022 The Authors. Published by ELSEVIER B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) 1. Introduction Modern demands on structural safety require periodic inspections of elements and structures being under operation in numerous industrial branches. Such inspections are usually performed with non-destructive testing (NDT) methods. However, such inspections may provide information on damage presence, location, and type only without a possibility of evaluation the structural residual life. The prediction of a structural response, being the last step of structural damage evaluation (Rytter, 1993), is one of the most challenging and still unsolved problems. To predict the structural residual life, its proper classification from the NDT results is a crucial step. Abstract Recently, due the automation of many industrial processes the intelligent tools for undertaking decisions started to be highly d manded solu ions in ny industrial branches. This is also t e cas in non-destructive evaluat on, where big amount f data from inspection need to be anal ze , and based on uc an analysis a d c sion must be undert ken. I many cases, the step of analysis is already automated with using edicated software, however, th decision-making process still engages a human. In this paper, the authors presented the preliminaries of the damage classification algorith , which is intended to cla sify differe t types of a structural damage in com osites b s d n X-ray computed tomography scans. The proposed approach was based o dee neur l netwo ks, which creates a possibility to btain high values of a classification accuracy. The obt ined results within this study cl a ly sho the effe tiveness of the proposed approach and cr ate p th for further development of th algorithm. © 2022 The Authors. Published by ELSEVIER B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under responsibility of Pedro Miguel Guimaraes Pires Moreira Keywords: X-ray computed tomography; Damage classification; Deep neural networks 1. Introduction Modern demands on structural safety require periodic inspections of elements and structures being under operation in num rous industrial bran hes. Such inspections are usually performed with on-de iv test (NDT) m thods. However, such in pections may provide information on d mag presence, location, and type only without a possibility of evaluation the tru tural residual l fe. The pred cti of a structural response, being the last step of str ctural damage evaluation (Rytter, 1993), is one of th most challengi g and still unsolved probl ms. To predict the residual life, its proper classification from the NDT resu ts is a crucial step. ICSI 2021 The 4th International Conference on Structural Integrity Damage classification in composite structures based on X-ray computed tomography scans using features evaluation and deep neural networks Tomasz Rogala a,1 , Piotr Przystałka a , Andrzej Katunin a ICSI 2021 The 4th International Conference on Structural Integrity Damage classification in composite structures based on X-ray computed tomography scans using features evaluation and deep neural networks Tomasz Rogala a,1 , Piotr Przystałka a , Andrzej Katunin a a Department of Fundamentals of Machinery Design, Silesian University of Technology, Konarskiego 18A, 44-100 Gliwice, Poland a Department of Fundamentals of Machinery Design, Silesian University of Technology, Konarskiego 18A, 44-100 Gliwice, Poland Peer-review under responsibility of Pedro Miguel Guimaraes Pires Moreira Keywords: X-ray computed tomography; Damage classification; Deep neural networks

1 * Corresponding author. Tel.: +48-32-237-2741; fax: +48-32-237-1360. E-mail address: tomasz.rogala@polsl.pl 1 * Corresponding author. Tel.: +48-32-237-2741; fax: +48-32-237-1360. E-mail address: tomasz.rogala@polsl.pl

2452-3216 © 2022 The Authors. Published by ELSEVIER B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under responsibility of Pedro Miguel Guimaraes Pires Moreira 2452-3216 © 2022 The Authors. Published by ELSEVIER B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under responsibility of Pedro Miguel Guimaraes Pires Moreira

2452-3216 © 2022 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under responsibility of Pedro Miguel Guimaraes Pires Moreira 10.1016/j.prostr.2022.01.076