elib
DLR-Header
DLR-Logo -> http://www.dlr.de
DLR Portal Home | Imprint | Privacy Policy | Contact | Deutsch
Fontsize: [-] Text [+]

Explainability of deep learning classifier decisions for optical detection of manufacturing defects in the automated fiber placement process

Meister, Sebastian and Wermes, Mahdieu A. M. and Stüve, Jan and Groves, Roger M. (2021) Explainability of deep learning classifier decisions for optical detection of manufacturing defects in the automated fiber placement process. In: Proceedings of SPIE - The International Society for Optical Engineering (11787). SPIE Automated Visual Inspection and Machine Vision IV, 21.-25. Jun. 2021, München. doi: 10.1117/12.2592584. ISSN 0277-786X.

[img] PDF - Only accessible within DLR bis 1 January 2032
7MB

Official URL: https://www.spiedigitallibrary.org/conference-proceedings-of-spie/11787/1178705/Explainability-of-deep-learning-classifier-decisions-for-optical-detection-of/10.1117/12.2592584.full?SSO=1

Abstract

Automated fibre layup techniques are commonly used composite manufacturing processes in the aviation sector and require a manual visual inspection. Neural Network classification of defects has the potential to automate this visual inspection, however, the machine decision-making processes are hard to verify. Thus, we present an approach for visualising Convolutional Neural Network (CNN) based classifications of manufacturing defects and quantifying its robustness suitably. Our investigations have shown that especially Smoothed Integrated Gradients and DeepSHAP are particularly well suited for the visualisation of CNN classifications. The Smoothed Integrated Gradients technique also reveals advantages in robustness when evaluating degraded input images.

Item URL in elib:https://elib.dlr.de/142832/
Document Type:Conference or Workshop Item (Speech)
Title:Explainability of deep learning classifier decisions for optical detection of manufacturing defects in the automated fiber placement process
Authors:
AuthorsInstitution or Email of AuthorsAuthor's ORCID iD
Meister, SebastianSebastian.Meister (at) dlr.dehttps://orcid.org/0000-0002-8193-1143
Wermes, Mahdieu A. M.Mahdieu.Wermes (at) dlr.deUNSPECIFIED
Stüve, JanJan.Stueve (at) dlr.dehttps://orcid.org/0000-0003-1483-2476
Groves, Roger M.R.M.Groves (at) tudelft.nlhttps://orcid.org/0000-0001-9169-9256
Date:June 2021
Journal or Publication Title:Proceedings of SPIE - The International Society for Optical Engineering
Refereed publication:Yes
Open Access:No
Gold Open Access:No
In SCOPUS:Yes
In ISI Web of Science:No
DOI :10.1117/12.2592584
Series Name:SPIE
ISSN:0277-786X
Status:Published
Keywords:Defect classifications, CNN, Inline Inspection, xAI, Computer Vision, Composite Manufacturing, Laser Line Scan Sensor
Event Title:SPIE Automated Visual Inspection and Machine Vision IV
Event Location:München
Event Type:international Conference
Event Dates:21.-25. Jun. 2021
Organizer:SPIE
HGF - Research field:Aeronautics, Space and Transport
HGF - Program:Aeronautics
HGF - Program Themes:Components and Systems
DLR - Research area:Aeronautics
DLR - Program:L CS - Components and Systems
DLR - Research theme (Project):L - Production Technologies
Location: Stade
Institutes and Institutions:Institute of Composite Structures and Adaptive Systems > Composite Process Technology
Deposited By: Meister, Sebastian
Deposited On:28 Jun 2021 07:59
Last Modified:19 Jul 2021 12:48

Repository Staff Only: item control page

Browse
Search
Help & Contact
Information
electronic library is running on EPrints 3.3.12
Copyright © 2008-2017 German Aerospace Center (DLR). All rights reserved.