Neural network segmentation methods for fatigue crack images obtained with X-ray tomography

Xiao, Ce; Buffière, Jean‐Yves

doi:10.1016/j.engfracmech.2021.107823

Cited by 13 publications

(4 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…On the other hand, the presence of these streak artifacts can assist in locating some invisible cracks. Therefore, a recently developed deep learning based method (U-net) 32 has been used to automatically distinguish cracsks (as well as invisible cracks) from streak artifacts in SRCT images by learning observer judgment. One result is shown in “ ESM Appendix ”.…”

Section: Discussionmentioning

confidence: 99%

Characterization of internal fatigue cracks in aluminum alloys by simulation of phase contrast tomography

Xiao

Létang

Buffière

2022

Sci Rep

View full text Add to dashboard Cite

Synchrotron Radiation Computed Tomography (SRCT) allows a better detection of fatigue cracks in metals than laboratory CT due to the existence of phase contrast. However the presence in reconstructed images of fringes at the edges of objects generated by Fresnel diffraction makes it difficult to identify and analyze the cracks quantitatively. Simulations of phase contrast synchrotron tomography images containing cracks with different sizes and shapes are obtained by using GATE software. Analyzing the simulation results, firstly, we confirmed that the bright parts with strong contrast in SRCT image are streak artifacts; secondly, we found that the gray scale values within the cracks in SRCT images are related to the crack size; these simulation results are used to analyse SRCT images of internal fatigue cracks in a cast Al alloy, providing a clearer visualisation of damage.

show abstract

Section: Discussionmentioning

confidence: 99%

Characterization of internal fatigue cracks in aluminum alloys by simulation of phase contrast tomography

Xiao

Létang

Buffière

2022

Sci Rep

View full text Add to dashboard Cite

show abstract

“…Xu et al [34] propose a novel tunnel defect inspection method based on the Mask R-CNN. Xiao and Buffiere [35] developed an image segmentation method based on a convolutional neural network for fatigue crack images.…”

Section: B Deep-learning Techniquesmentioning

confidence: 99%

Deep Learning Based Classification of Military Cartridge Cases and Defect Segmentation

et al. 2022

View full text Add to dashboard Cite

The final stage of the production process in the industry is quality control. Quality control answers the question of is there a defect on the surface of the products. Frequently the quality control is performed manually. The disadvantages of manual quality control are high error rate (low accuracy), low product rate (low performance) and high expense rate (high cost). The solution is automatic quality control using machine vision systems. These systems classify the products and segment the defects on their surfaces by processing the images taken by cameras during the production process in real-time. Some products like military cartridge cases have metallic, cylindrical, non-uniform texture and highly reflective surface. So, the quality of images is very important. Another factor that affects the accuracy is the nonuniform texture of the product surface. Distinguishing the product non-uniform texture from defect texture is a challenging problem. In previous works, this problem have been tried to be solved with image processing and deep learning techniques and the accuracy of 97% and 96% have been obtained, appropriately. According to NATO standards, the accuracy of the classification of the military cartridge cases should be above 99%. In this work, the methodology for classification of the military cartridge cases and segmentation of the defects on their surfaces with non-uniform texture is proposed to increase the accuracy. In scope of the proposed methodology the datasets with non-defective, defective and labeled/masked image classes of the cartridge cases were created, the deep learning models to classify the military cartridge cases and segment the defects on their surfaces were proposed, implemented and obtained results were evaluated using the metrics such as Accuracy, Precision, Recall, F1-Score, Jaccard Index (JI) and Mean Intersection over Union (mIoU). Obtained results showed that the proposed methodology increased the accuracy of classification to 100% with the DenseNet169 model and the F1-Score of segmentation to 92.1% with Improved U-Net and ResUnet models.

show abstract

“…That is why in recent years, the capabilities of alternative solutions based on artificial neural networks for segmentation tasks are also being investigated. [100] It should also be mentioned that the direct immersion of microstructural input data for the CA SRX simulations coming from the experimental observation becomes demanding in the 3D space. The serial sectioning approach followed by image analysis and a wider range of 3D reconstruction algorithms seems to be the most accessible technique.…”

Section: Ca Models Of Static Recrystallizationmentioning

confidence: 99%

“…That is why in recent years, the capabilities of alternative solutions based on artificial neural networks for segmentation tasks are also being investigated. [ 100 ]…”

Section: Ca Models Of Static Recrystallizationmentioning

confidence: 99%

Computationally Efficient Cellular Automata‐Based Full‐Field Models of Static Recrystallization: A Perspective Review

Madej

Sitko

2022

steel research int.

View full text Add to dashboard Cite

The evolution of metallic material microstructures under thermomechanical treatment is a critical element that influences the in‐use properties of the final product. In the last four decades, many researchers around the world focused on reliable numerical recreation of phenomena that occur during metal processing. Particular attention is put on the phase transformation, texture evolution, and recrystallization predictions to simulate material behavior at the subsequent processing stages. In recent years, approaches taking into account explicit representation of morphological elements during simulation are becoming more frequently used even at the industrial scale. One of those methods addressed within the current article in application to static recrystallization (SRX) modeling is the cellular automata (CA) approach. A review of the CA‐based full‐field SRX models highlights the progress in capabilities of developed approaches in consideration of various mechanisms controlling microstructure evolution like heterogeneous energy distribution, probabilistic or physics‐based nucleation, grain growth driven by different driving forces, or grain boundary migration is presented within the article. The issue of the computational time associated with 3D CA modeling is particularly addressed. Possible approaches to reducing simulation times based on efficient use of modern computer architecture are finally evaluated as guidelines for further model development.

show abstract

Neural network segmentation methods for fatigue crack images obtained with X-ray tomography

Cited by 13 publications

References 42 publications

Characterization of internal fatigue cracks in aluminum alloys by simulation of phase contrast tomography

Characterization of internal fatigue cracks in aluminum alloys by simulation of phase contrast tomography

Deep Learning Based Classification of Military Cartridge Cases and Defect Segmentation

Computationally Efficient Cellular Automata‐Based Full‐Field Models of Static Recrystallization: A Perspective Review

Contact Info

Product

Resources

About