Phase segmentation of uncured prepreg X-Ray CT micrographs

Galvez-Hernandez, Pedro; Gąska, Karolina; Kratz, James

doi:10.1016/j.compositesa.2021.106527

Cited by 14 publications

(27 citation statements)

References 43 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This points towards further possibilities of the approach to identify, e.g., fiber tows, and layers, and then calculate or analyze orientations and deviations. The approach could also be used on CT-scanned micrographs as in [ 22 ].…”

Section: Discussionmentioning

confidence: 99%

“…Neural networks have also recently been introduced for the segmentation of microstructures. Ge et al [ 21 ] reviewed the applications of deep learning on microscopic image analysis and its possibilities, and Galvez-Hernandez et al [ 22 ] investigated interlaminar voids and dry areas in uncured prepreg based on images from micro-CT scanning. They explored the benefits of using machine learning and found that machine learning consistently exceeded a thresholding approach.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A Statistical Porosity Characterization Approach of Carbon-Fiber-Reinforced Polymer Material Using Optical Microscopy and Neural Network

et al. 2022

View full text Add to dashboard Cite

The intensified pursuit for lightweight solutions in the commercial vehicle industry increases the demand for method development of more advanced lightweight materials such as Carbon-Fiber-Reinforced Composites (CFRP). The behavior of these anisotropic materials is challenging to understand and manufacturing defects could dramatically change the mechanical properties. Voids are one of the most common manufacturing defects; they can affect mechanical properties and work as initiation sites for damage. It is essential to know the micromechanical composition of the material to understand the material behavior. Void characterization is commonly conducted using optical microscopy, which is a reliable technique. In the current study, an approach based on optical microscopy, statistically characterizing a CFRP laminate with regard to porosity, is proposed. A neural network is implemented to efficiently segment micrographs and label the constituents: void, matrix, and fiber. A neural network minimizes the manual labor automating the process and shows great potential to be implemented in repetitive tasks in a design process to save time. The constituent fractions are determined and they show that constituent characterization can be performed with high accuracy for a very low number of training images. The extracted data are statistically analyzed. If significant differences are found, they can reveal and explain differences in the material behavior. The global and local void fraction show significant differences for the material used in this study and are good candidates to explain differences in material behavior.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

A Statistical Porosity Characterization Approach of Carbon-Fiber-Reinforced Polymer Material Using Optical Microscopy and Neural Network

et al. 2022

View full text Add to dashboard Cite

show abstract

“…8 In the composites field, Deep Learning has been successfully applied to a range of applications, such as the analysis of multiclass damage, 9,10 the characterisation of woven composites from low-contrast and low-resolution X-Ray images, 11 and the generation of high-fidelity digital twins. 12 Furthermore, Deep Learning has shown a superior performance to thresholding in the phase segmentation of uncured composite samples in the presence of moderate levels of noise 13 and has enabled an accurate assessment of the porosity content in optical microscopy images of composite samples displaying a significant image quality variability. 14 However, as Deep Learning is increasingly gaining attention, questions arising around which CNN to use to segment the features of interest in composite images, and which combination of user-defined parameters, also known as hyperparameters will provide optimum segmentation results.…”

Section: Introductionmentioning

confidence: 99%

“…14 However, as Deep Learning is increasingly gaining attention, questions arising around which CNN to use to segment the features of interest in composite images, and which combination of user-defined parameters, also known as hyperparameters will provide optimum segmentation results. While U-Net 15 is a common choice to segment composite images, 11,13,14 other networks such as FCDenseNet 16,17 or DeepLabv3+ 12,18 have also been applied in previous studies.…”

Section: Introductionmentioning

confidence: 99%

The effect of convolutional neural network architectures on phase segmentation of composite material X-ray micrographs

Galvez-Hernandez

Kratz

2023

Journal of Composite Materials

Self Cite

View full text Add to dashboard Cite

Porosity severely reduces the mechanical performance of composite laminates and methods for automatic segmentation of void phases are growing. This study investigates porosity in composite materials that take the form of interlaminar voids and dry tow areas. Deep Learning was used for the segmentation of X-ray micrographs via the implementation of eight state-of-the-art Convolutional Neural Network (CNN) architectures trained with data sets containing twenty-five, fifty, and one-hundred images. The combination of hyperparameters providing the highest accuracy for each architecture and training set size was achieved through the optimisation of six relevant hyperparameters, including the cut-off probability applied to output probability maps. Additionally, the properties of the CNN architectures ( e.g., layer typology, connections, density…) were found to play a determining role, not only in the segmentation results but also in the associated computing effort. U-Net and FCDenseNet outperformed the FCN-8s, FCN-16, SegNet, LinkNet, ResNet18 and Xception CNN architectures. However, the CNNs generally outperformed the standard thresholding approaches, especially in sub-volumes containing low porosity (1.07%) where the influence on strength is very sensitive in high-performance composites. In low porosity samples, U-Net and FCDenseNet consistently segmented voids to 85% + accuracy, whereas thresholding was only half as accurate, at around 40%. The results provide a strong motivation to replace thresholding as a segmentation method for composite X-ray micrographs. In terms of efficiency, the reduced complexity of the U-Net network allowed for an average reduction of the training time (−36%) and prediction time (−17%) when compared to FCDenseNet.

show abstract

“…Algorithm-based automatic segmentation, however, requires the user to have programming knowledge and a thorough understanding of mathematical algorithms related to the image processing software being used (Rathnayaka et al, 2011). Deep learning-based segmentation methods may out-perform simple thresholding, but at a high cost for initial training (Galvez-Hernandez et al, 2021).…”

Section: Introductionmentioning

confidence: 99%

A New Pipeline to Automatically Segment and Semi-Automatically Measure Bone Length on 3D Models Obtained by Computed Tomography

Diaz

H’ng

et al. 2021

Front. Cell Dev. Biol.

View full text Add to dashboard Cite

The characterization of developmental phenotypes often relies on the accurate linear measurement of structures that are small and require laborious preparation. This is tedious and prone to errors, especially when repeated for the multiple replicates that are required for statistical analysis, or when multiple distinct structures have to be analyzed. To address this issue, we have developed a pipeline for characterization of long-bone length using X-ray microtomography (XMT) scans. The pipeline involves semi-automated algorithms for automatic thresholding and fast interactive isolation and 3D-model generation of the main limb bones, using either the open-source ImageJ plugin BoneJ or the commercial Mimics Innovation Suite package. The tests showed the appropriate combination of scanning conditions and analysis parameters yields fast and comparable length results, highly correlated with the measurements obtained via ex vivo skeletal preparations. Moreover, since XMT is not destructive, the samples can be used afterward for histology or other applications. Our new pipelines will help developmental biologists and evolutionary researchers to achieve fast, reproducible and non-destructive length measurement of bone samples from multiple animal species.

show abstract

Phase segmentation of uncured prepreg X-Ray CT micrographs

Cited by 14 publications

References 43 publications

A Statistical Porosity Characterization Approach of Carbon-Fiber-Reinforced Polymer Material Using Optical Microscopy and Neural Network

A Statistical Porosity Characterization Approach of Carbon-Fiber-Reinforced Polymer Material Using Optical Microscopy and Neural Network

The effect of convolutional neural network architectures on phase segmentation of composite material X-ray micrographs

A New Pipeline to Automatically Segment and Semi-Automatically Measure Bone Length on 3D Models Obtained by Computed Tomography

Contact Info

Product

Resources

About