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

Abstract

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.

Multiscale Characterisation of Staple Carbon Fibre-Reinforced Polymers

“…In the literature, a vast number of methods exist that have been explored within the field of fibre-reinforced composite materials, such as fibre orientation [30][31][32], fibre length [33,34], interfibre spacing [35], fibre diameter [36], fibre connectivity [37], fibre curvature [37], fibre volume fraction (FVF) [38], and porosity [39,40]. Each microstructural descriptor is measured using individual approaches by destructive methods, e.g., optical microscopy, including bright-field and polarised light microscopy [41,42] and mechanical testing, such as tensile testing or compression testing [11,21,27]; or non-destructive methods, e.g., X-raybased techniques, such as X-ray diffraction [43], X-ray computed tomography [35,37,[44][45][46][47] and ultrasound computed tomography [48]; ultrasonic testing [49][50][51], nuclear magnetic resonance spectroscopy and Raman spectroscopy [52,53]; image analysis using different algorithms, such as Fourier analysis [54], gradient-based methods, and structure tensor methods. Destructive testing provides precise measurements but destroys the sample, while non-destructive testing preserves sample integrity and allows for in situ applications but may offer less detailed information and require specialized equipment.…”