Non-contact detection of railhead defects and their classification by using convolutional neural network

Ghafoor, Imran; Tse, Peter W.; Munir, Nauman; Trappey, Amy J.C.

doi:10.1016/j.ijleo.2022.168607

Cited by 9 publications

(5 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The convolutional layer uses convolution between filters and data to extract meaningful features from the data. In other words, when the input data are filtered, the data dimension is reduced several times, which has the advantage of reducing the amount of computational data but may cause data loss in some cases [30]. Therefore, a pooling layer is used to retain the characteristics.…”

Section: Convolutional Neural Network (Cnn)mentioning

confidence: 99%

“…A [31] 0.035 B [32] 0.025 C [30] 0.04 D [21] 0.045 E [29] 0.046 F [28] 0.057 Table 4 shows the depths at which the maximum stress occurs in data A and B, which are 0.035 and 0.025 mm. When using Equation (5), the higher the frequency of the Rayleigh wave, the more accurate the conversion in the range close to the surface.…”

Section: Xrd Validation Data From Referencesmentioning

confidence: 99%

See 1 more Smart Citation

Development of Maximum Residual Stress Prediction Technique for Shot-Peened Specimen Using Rayleigh Wave Dispersion Data Based on Convolutional Neural Network

Choi,

Lee,

Yeom

et al. 2023

Materials

View full text Add to dashboard Cite

Shot peening is a surface treatment process that improves the fatigue life of a material and suppresses cracks by generating residual stress on the surface. The injected small shots create a compressive residual stress layer on the material’s surface. Maximum compressive residual stress occurs at a certain depth, and tensile residual stress gradually occurs as the depth increases. This process is primarily used for nickel-based superalloy steel materials in certain environments, such as the aerospace industry and nuclear power fields. To prevent such a severe accident due to the high-temperature and high-pressure environment, evaluating the residual stress of shot-peened materials is essential in evaluating the soundness of the material. Representative methods for evaluating residual stress include perforation strain gauge analysis, X-ray diffraction (XRD), and ultrasonic testing. Among them, ultrasonic testing is a representative, non-destructive evaluation method, and residual stress can be estimated using a Rayleigh wave. Therefore, in this study, the maximum compressive residual stress value of the peened Inconel 718 specimen was predicted using a prediction convolutional neural network (CNN) based on the relationship between Rayleigh wave dispersion and stress distribution on the specimen. By analyzing the residual stress distribution in the depth direction generated in the model from various studies in the literature, 173 residual stress distributions were generated using the Gaussian function and factorial design approach. The distribution generated using the relationship was converted into 173 Rayleigh wave dispersion data to be used as a database for the CNN model. The CNN model was learned through this database, and performance was verified using validation data. The adopted Rayleigh wave dispersion and convolutional neural network procedures demonstrate the ability to predict the maximum compressive residual stress in the peened specimen.

show abstract

Section: Convolutional Neural Network (Cnn)mentioning

confidence: 99%

Section: Xrd Validation Data From Referencesmentioning

confidence: 99%

Development of Maximum Residual Stress Prediction Technique for Shot-Peened Specimen Using Rayleigh Wave Dispersion Data Based on Convolutional Neural Network

Choi,

Lee,

Yeom

et al. 2023

Materials

View full text Add to dashboard Cite

show abstract

“…The model has a simple structure and faster processing speed, achieving a recognition accuracy of 92.08%, but the method is mainly effective for the detection of scar defects. Wang et al (2018) ; Ni et al (2021) , and Ghafoor et al (2022) analyzed the image features of rail defects, removed interference noise by image filtering, and then trained the model to improve the detection of surface defects, but the image enhancement algorithm is not universal and the image processing is time-consuming. Han et al (2021) presented a multi-level feature fusion model for rail surface defects detection, which fuses the image features of different receptive field of multiple levels for target detection and enhances the accuracy of detection results and decreases the missing detection rate of small area defects, but the method detects too few types of defects and is not applicable to the detection of multiple complex defects of the rail.…”

Section: Introductionmentioning

confidence: 99%

Research on steel rail surface defects detection based on improved YOLOv4 network

Ren²,

Zhao³

2023

Front. Neurorobot.

View full text Add to dashboard Cite

IntroductionThe surface images of steel rails are extremely difficult to detect and recognize due to the presence of interference such as light changes and texture background clutter during the acquisition process.MethodsTo improve the accuracy of railway defects detection, a deep learning algorithm is proposed to detect the rail defects. Aiming at the problems of inconspicuous rail defects edges, small size and background texture interference, the rail region extraction, improved Retinex image enhancement, background modeling difference, and threshold segmentation are performed sequentially to obtain the segmentation map of defects. For the classification of defects, Res2Net and CBAM attention mechanism are introduced to improve the receptive field and small target position weights. The bottom-up path enhancement structure is removed from the PANet structure to reduce the parameter redundancy and enhance the feature extraction of small targets.ResultsThe results show the average accuracy of rail defects detection reaches 92.68%, the recall rate reaches 92.33%, and the average detection time reaches an average of 0.068 s per image, which can meet the real-time of rail defects detection.DiscussionComparing the improved method with the mainstream target detection algorithms such as Faster RCNN, SSD, YOLOv3 and other algorithms, the improved YOLOv4 has excellent comprehensive performance for rail defects detection, the improved YOLOv4 model obviously better than several others in Pr, Rc, and F1 value, and can be well-applied to rail defect detection projects.

show abstract

“…2 With the development of non-contact detection technology, manual visual inspection is gradually being replaced. 3 The traditional non-contact inspection methods include eddy current detection, 4 ultrasonic guided wave detection, 5 and infrared detection. 6 However, these methods have limitations, such as being costly, complex, and have limited visual feedback during the defect identification process.…”

Section: Introductionmentioning

confidence: 99%

“…The existing methods for detection of end-face defects of metal gaskets are commonly detected through manual visual inspection, which can be influenced by the subjective factors of personnel and lead to inaccurate detection 2 . With the development of non-contact detection technology, manual visual inspection is gradually being replaced 3 . The traditional non-contact inspection methods include eddy current detection, 4 ultrasonic guided wave detection, 5 and infrared detection 6 .…”

Section: Introductionmentioning

confidence: 99%

SR-HorNetYOLO: an efficient algorithm for defect detection of metal gasket end-face

Ding,

Pan,

Zhang

et al. 2023

J. Electron. Imag.

View full text Add to dashboard Cite

.Metal gaskets are an essential component of the manufacturing industry, and detecting end-face defects is critical to ensuring their quality. However, current detection methods struggle to detect such defects due to the gaskets’ uneven end-face structure and small defect size. To address such issues, an approach called SR-HorNetYOLO has been proposed. This approach combines traditional machine vision and deep learning techniques. The visual saliency method is used to simplify the image and remove irrelevant features. YOLOv5 is then optimized in three ways: the GhostNet backbone replaces the CSPDarkNet53 backbone to reduce model weight, the self-attention mechanism and bottleneck transformer block improve the precision of detecting small targets, and the feature pyramid is improved through the HorNet network to enhance the predictive performance of the network. Finally, a classification module and position module are employed to classify and locate three-fusion proportion characteristic images. The experimental results show that the SR-HorNetYOLO method achieves a precision of 93.3% and a recall rate of 94.0%.

show abstract

Non-contact detection of railhead defects and their classification by using convolutional neural network

Cited by 9 publications

References 33 publications

Development of Maximum Residual Stress Prediction Technique for Shot-Peened Specimen Using Rayleigh Wave Dispersion Data Based on Convolutional Neural Network

Development of Maximum Residual Stress Prediction Technique for Shot-Peened Specimen Using Rayleigh Wave Dispersion Data Based on Convolutional Neural Network

Research on steel rail surface defects detection based on improved YOLOv4 network

SR-HorNetYOLO: an efficient algorithm for defect detection of metal gasket end-face

Contact Info

Product

Resources

About