Mixture-of-Experts for Multi-Domain Defect Identification in Non-Destructive Inspection

Reddy Seethi, Venkata Devesh; Rahman, Ashiqur; Yunker, Austin; Lake, Rami; Kral, Zachary; Kettimuthu, Rajkumar; Alhoori, Hamed

doi:10.1109/icmla61862.2024.00200

Mixture-of-Experts for Multi-Domain Defect Identification in Non-Destructive Inspection

Conference · Tue Dec 17 23:00:00 EST 2024 · 2024 International Conference on Machine Learning and Applications (ICMLA)

DOI:https://doi.org/10.1109/icmla61862.2024.00200· OSTI ID:2563185

Reddy Seethi, Venkata Devesh ^[1]; Rahman, Ashiqur ^[2]; Yunker, Austin ^[3]; Lake, Rami ^[2]; Kral, Zachary ^[4]; Kettimuthu, Rajkumar ^[3]; Alhoori, Hamed ^[2]

Northern Illinois University, DeKalb, IL (United States); Spirit AeroSystems Inc
Northern Illinois University, DeKalb, IL (United States)
Argonne National Laboratory (ANL), Argonne, IL (United States)
Spirit AeroSystems, Inc., Wichita, KS (United States)

Composite materials are widely used in aircraft structures because of their superior mechanical properties. However, their complex failure modes require sophisticated inspection methods to ensure structural integrity. Ultrasonic testing (UT) is a common non-destructive inspection (NDI) technique for aircraft composites that can detect internal and external defects with high resolution and accuracy. Despite their effectiveness, traditional UT methods rely on the manual interpretation of ultrasonic signals, which is time-consuming, labor-intensive, and subjective. Furthermore, processing such large-scale data, particularly across materials of varying thicknesses, significantly increases the computational demands of deep learning model optimization. To overcome these challenges, we propose an efficient sparse mixture-of-experts (MoE) model with a multi-level loss function and introduce four novel training objectives to improve computational efficiency and accuracy in identifying surface defects in composite aircraft materials. Here, we evaluated our approach on material with multiple thicknesses or domains comprising various defects. Our experimental results demonstrate higher accuracy and F1-Score, with only 10% training epochs compared to baseline MoE.

View Conference

Research Organization:: Spirit AeroSystems, Inc., Wichita, KS (United States)

Sponsoring Organization:: USDOE Office of Energy Efficiency and Renewable Energy (EERE); USDOE Office of Science (SC)

DOE Contract Number:: EE0009397; AC02-06CH11357

OSTI ID:: 2563185

Journal Information:: 2024 International Conference on Machine Learning and Applications (ICMLA), Journal Name: 2024 International Conference on Machine Learning and Applications (ICMLA)

Country of Publication:: United States

Language:: English

References (17)

Successes and challenges in non-destructive testing of aircraft composite structures Towsyfyan, Hossein; Biguri, Ander; Boardman, Richard Chinese Journal of Aeronautics, Vol. 33, Issue 3 https://doi.org/10.1016/j.cja.2019.09.017	journal	March 2020
Toward Zero Defect Manufacturing with the support of Artificial Intelligence—Insights from an industrial application Leberruyer, Nicolas; Bruch, Jessica; Ahlskog, Mats Computers in Industry, Vol. 147 https://doi.org/10.1016/j.compind.2023.103877	journal	May 2023
A critical review of recent advances in the aerospace materials Soni, Rahul; Verma, Rajeev; Kumar Garg, Rajiv Materials Today: Proceedings, Vol. 113 https://doi.org/10.1016/j.matpr.2023.08.108	journal	January 2024
Learning defects from aircraft NDT data Prakash, Navya; Nieberl, Dorothea; Mayer, Monika NDT & E International, Vol. 138 https://doi.org/10.1016/j.ndteint.2023.102885	journal	September 2023
A comparison of methods for generating synthetic training data for domain adaption of deep learning models in ultrasonic non-destructive evaluation McKnight, Shaun; Pierce, S. Gareth; Mohseni, Ehsan NDT & E International, Vol. 141 https://doi.org/10.1016/j.ndteint.2023.102978	journal	January 2024
Ultrasonic signal classification and imaging system for composite materials via deep convolutional neural networks Meng, Min; Chua, Yiting Jacqueline; Wouterson, Erwin Neurocomputing, Vol. 257 https://doi.org/10.1016/j.neucom.2016.11.066	journal	September 2017
An Online Framework for Cognitive Load Assessment in Industrial Tasks Lagomarsino, Marta; Lorenzini, Marta; De Momi, Elena Robotics and Computer-Integrated Manufacturing, Vol. 78 https://doi.org/10.1016/j.rcim.2022.102380	journal	December 2022
An improved automated ultrasonic NDE system by wavelet and neuron networks Bettayeb, Fairouz; Rachedi, Tarek; Benbartaoui, Hamid Ultrasonics, Vol. 42, Issue 1-9 https://doi.org/10.1016/j.ultras.2004.01.064	journal	April 2004
Ultrasonic detection of manufacturing defects in multilayer composite structures Kokurov, A. M.; Subbotin, D. E. IOP Conference Series: Materials Science and Engineering, Vol. 1023, Issue 1 https://doi.org/10.1088/1757-899X/1023/1/012013	journal	January 2021
Machine learning protocol from ultrasound data for monitoring, predicting, and supporting the analysis of dam slopes Rocha, Werickson F. C.; Lucena, Antônio U.; Sarmanho, Gabriel F. 2022 21st IEEE International Conference on Machine Learning and Applications (ICMLA) https://doi.org/10.1109/ICMLA55696.2022.00084	conference	December 2022
Multi-Modal Machine Learning for Navigating Noisy Objectives of Automotive Manufacturing Quality Inspection Shirazi, Majid; Safronov, Georgii; Rizk, Amr 2023 International Conference on Machine Learning and Applications (ICMLA) https://doi.org/10.1109/ICMLA58977.2023.00284	conference	December 2023
Designing Effective Sparse Expert Models Zoph, Barret 2022 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW) https://doi.org/10.1109/ipdpsw55747.2022.00171	conference	May 2022
Comparative Study on Deep Learning Methods for Defect Identification and Classification in Composite Aerostructure Material Yunker, Austin; Lake, Rami; Kettimuthu, Rajkumar 2023 50th Annual Review of Progress in Quantitative Nondestructive Evaluation https://doi.org/10.1115/QNDE2023-108602	conference	July 2023
Neural network approach of active ultrasonic signals for structural health monitoring analysis Kral, Zachary; Horn, Walter; Steck, James SPIE Proceedings https://doi.org/10.1117/12.815503	conference	March 2009
Nondestructive testing and evaluation techniques of defects in fiber-reinforced polymer composites: A review Chen, Jian; Yu, Zhenyang; Jin, Haoran Frontiers in Materials, Vol. 9 https://doi.org/10.3389/fmats.2022.986645	journal	October 2022
Computerized Ultrasonic Imaging Inspection: From Shallow to Deep Learning Ye, Jiaxing; Ito, Shunya; Toyama, Nobuyuki Sensors, Vol. 18, Issue 11 https://doi.org/10.3390/s18113820	journal	November 2018
A Systematic Review of Advanced Sensor Technologies for Non-Destructive Testing and Structural Health Monitoring Hassani, Sahar; Dackermann, Ulrike Sensors, Vol. 23, Issue 4 https://doi.org/10.3390/s23042204	journal	February 2023

Similar Records

Comparative Study on Deep Learning Methods for Defect Identification and Classification in Composite Aerostructure Material

Conference · Thu Jul 27 00:00:00 EDT 2023 · 2023 50th Annual Review of Progress in Quantitative Nondestructive Evaluation · OSTI ID:2562929

Evaluating Model Robustness for Defect Identification and Classification in a Composite Aerostructure Material

Journal Article · Fri Jul 26 00:00:00 EDT 2024 · Journal of Nondestructive Evaluation, Diagnostics and Prognostics of Engineering Systems · OSTI ID:2563184

A Quantitative Assessment of Advanced NDI Techniques for Detecting Flaws in Composite Laminate Aircraft Structures. Draft

Technical Report · Fri Aug 01 00:00:00 EDT 2014 · OSTI ID:1762097

Related Subjects

Mixture of experts
composite materials
efficient machine learning
non destructive inspection
ultrasonic signals

Mixture-of-Experts for Multi-Domain Defect Identification in Non-Destructive Inspection

Citation Formats

References (17)

Similar Records

Related Subjects