Using deep learning improve the aerial engine nondestructive radiographic tests

Zhi Hao Chen, Jyh Ching Juang

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

This paper aim use of deep convolutional neural networks (CNNs) with generative adversarial networks for aircraft engine X-ray cracks image classification and detection posed. On the basis of the CNNs approach requires large amounts of X-ray defect imagery data. Those data facilitate a cracks image segmentation and tracking on multiple defect of aircraft engine defection by edge detection feature extraction and classification process. The use of the deep CNNs approach deep learning model seeks to augment and improve existing automated nondestructive testing (NDT) diagnosis. Within the context of X-ray screening, limited numbers insufficient types of X-ray aircraft engine defect data samples can thus pose another problem in support vector machine (SVM) model accuracy. To overcome this issue, we employ a deep learning paradigm of generative adversarial network such that a pre-trained deep CNNs. We are primarily trained for aircraft engine defect X-ray image classification eight types where sufficient training data exists. This result are empirically show that deep learning net complex with the pre-tuned model features also more yield superior performance to human crafted features on object identification tasks. Overall the achieve result get more then 90% accuracy based on the DetectNet features model retrained with 8 types of composite material defect classifiers.

Original languageEnglish
Title of host publicationProceedings - 2020 International Symposium on Computer, Consumer and Control, IS3C 2020
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages315-318
Number of pages4
ISBN (Electronic)9781728193625
DOIs
Publication statusPublished - 2020 Nov
Event2020 International Symposium on Computer, Consumer and Control, IS3C 2020 - Taichung, Taiwan
Duration: 2020 Nov 132020 Nov 16

Publication series

NameProceedings - 2020 International Symposium on Computer, Consumer and Control, IS3C 2020

Conference

Conference2020 International Symposium on Computer, Consumer and Control, IS3C 2020
CountryTaiwan
CityTaichung
Period20-11-1320-11-16

All Science Journal Classification (ASJC) codes

  • Electrical and Electronic Engineering
  • Control and Optimization
  • Instrumentation
  • Atomic and Molecular Physics, and Optics
  • Computer Science Applications
  • Energy Engineering and Power Technology
  • Artificial Intelligence

Fingerprint Dive into the research topics of 'Using deep learning improve the aerial engine nondestructive radiographic tests'. Together they form a unique fingerprint.

Cite this