Microstructure reconstruction of porous copper foams based on deep convolutional generative adversarial networks with physical characteristics of materials

Porous copper foams are widely used thermal insulation materials. The CT scan is a widely used method to obtain a two-dimensional digital images of porous copper foams and the foam’s performance can be studied with the images. Due to the high cost of the CT scan, it is not easy to obtain a large number of CT images. In order to cope with this problem, a method based on deep convolutional generative adversarial networks is proposed to generate huge amounts of images have similar characteristics of CT images. The proposed method uses convolutional neural networks to extract microstructure features of copper foams and uses generative adversarial networks to generate images. In order to keep the similarities between reconstructed images, porosity loss is introduced. 8000 two-dimensional tomographic images of copper foams are used to verify the effectiveness of the proposed method. Experimental results show that the proposed method could reconstruct two-dimensional microstructure of copper foams with different sizes quickly and retain inherent characteristics of copper foams. Compared with existing algorithms, reconstructed images reconstructed by the proposed method are similar to the microstructures of copper foams in many metrics.