Hidden corrosion detection in aircraft structures with a lightweight magnetic convolutional neural network.
Saved in:
| Title: | Hidden corrosion detection in aircraft structures with a lightweight magnetic convolutional neural network. |
|---|---|
| Authors: | Vu, Thuy Phuong1 (AUTHOR), Luong, Van Su1,2 (AUTHOR), Le, Minhhuy1,2 (AUTHOR) huy.leminh@phenikaa-uni.edu.vn |
| Source: | Nondestructive Testing & Evaluation. May2025, Vol. 40 Issue 5, p1797-1819. 23p. |
| Subjects: | Convolutional neural networks, Airframes, Electromagnetic testing, Deep learning, Signal processing |
| Abstract: | Detection of hidden corrosion within aircraft structures poses a significant challenge in non-destructive testing (NDT) methodologies, particularly in the electromagnetic testing (ET) method. In the contemporary era of Deep Learning, the imperative for intelligent NDT systems to employ lightweight models becomes apparent, facilitating their operational feasibility without the necessity of extensive computational resources or high-end configurations. In this paper, we propose a Lightweight Magnetic Convolutional Neural Network (LMagNet) model tailored for the efficient processing of electromagnetic signals from corrosion. The proposed LMagNet model achieves better performance compared to other lightweight models such as ShuffleNet, SqueezeNet, and MobileNet structures. The model archives an accuracy of 92% with only 16K parameters when being evaluated on hidden corrosion of aircraft structure, having volumes from 2.8 to 195.4 mm3. The model size is about 224×, 67×, and 96× smaller compared to the ShuffleNet G1, SqueezeNet Complex, and MobileNet V3 models. When deploying on low-resource microcontrollers (i.e. STM32 MCUs), the LMagNet model requires only 90kB of flash and 36kB of RAM, allowing it to run within 40 ms per prediction. In addition, we employed explainable techniques to interpret how the decision-making process of the model is made to achieve reliable results. [ABSTRACT FROM AUTHOR] |
| Copyright of Nondestructive Testing & Evaluation is the property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) | |
| Database: | Engineering Source |
|
Full text is not displayed to guests.
Login for full access.
|
|
| Abstract: | Detection of hidden corrosion within aircraft structures poses a significant challenge in non-destructive testing (NDT) methodologies, particularly in the electromagnetic testing (ET) method. In the contemporary era of Deep Learning, the imperative for intelligent NDT systems to employ lightweight models becomes apparent, facilitating their operational feasibility without the necessity of extensive computational resources or high-end configurations. In this paper, we propose a Lightweight Magnetic Convolutional Neural Network (LMagNet) model tailored for the efficient processing of electromagnetic signals from corrosion. The proposed LMagNet model achieves better performance compared to other lightweight models such as ShuffleNet, SqueezeNet, and MobileNet structures. The model archives an accuracy of 92% with only 16K parameters when being evaluated on hidden corrosion of aircraft structure, having volumes from 2.8 to 195.4 mm3. The model size is about 224×, 67×, and 96× smaller compared to the ShuffleNet G1, SqueezeNet Complex, and MobileNet V3 models. When deploying on low-resource microcontrollers (i.e. STM32 MCUs), the LMagNet model requires only 90kB of flash and 36kB of RAM, allowing it to run within 40 ms per prediction. In addition, we employed explainable techniques to interpret how the decision-making process of the model is made to achieve reliable results. [ABSTRACT FROM AUTHOR] |
|---|---|
| ISSN: | 10589759 |
| DOI: | 10.1080/10589759.2024.2360052 |
