Packing,compressibility,and crushability of rockfill materials with polydisperse particle size distributions and implications for dam engineering

In rockfill dam engineering,particle breakage of rockfill materials is one of the major factors resulting in dam settlement.In this study,one-dimensional compression tests on a series of coarse granular materials with artificially-graded particle size distributions(PSDs)were carried out.The tests focused on understanding the role of initial PSDs in the dense packing density,compressibility and crushability of coarse granular materials.The effects of fractal dimension(D)and size polydispersity(θ)of PSDs were quantitatively analyzed.Two different loading stages were identified from the logarithms of the stress-strain relationships,with the turning point marked as the yield stress.A similar effect of initial PSDs was observed on the packing density and low-pressure modulus of coarse granular materials.The packing density and low-pressure modulus increased monotonically withθ,and their peak values were attained at a D value of approximately 2.2.However,there was no unique correspondence between the dense packing density and low-pressure modulus.The particle breakage was influenced differently by the initial PSDs,and it decreased with the values of D andθ.The emergence of the unique ultimate state was also identified from both the compression curves and PSDs of the samples after the tests.The potential implications of the test results in the design of both low and high rockfill dams were also demonstrated.

Generative Adversarial Networks with Joint Distribution Moment Matching

Generative adversarial networks(GANs)have shown impressive power in the field of machine learning.Traditional GANs have focused on unsupervised learning tasks.In recent years,conditional GANs that can generate data with labels have been proposed in semi-supervised learning and have achieved better image quality than traditional GANs.Conditional GANs,however,generally only minimize the difference between marginal distributions of real and generated data,neglecting the difference with respect to each class of the data.To address this challenge,we propose the GAN with joint distribution moment matching(JDMM-GAN)for matching the joint distribution based on maximum mean discrepancy,which minimizes the differences of both the marginal and conditional distributions.The learning procedure is iteratively conducted by the stochastic gradient descent and back-propagation.We evaluate JDMM-GAN on several benchmark datasets,including MNIST,CIFAR-10 and the Extended Yale Face.Compared with the state-of-the-art GANs,JDMM-GAN generates more realistic images and achieves the best inception score for CIFAR-10 dataset.