Elastic-Wave Reverse Time Migration Random Boundary-Noise Suppression Based on CycleGAN

In elastic-wave reverse-time migration(ERTM),the reverse-time reconstruction of source wavefield takes advantage of the computing power of GPU,avoids its disadvantages in disk-access efficiency and reading and writing of temporary files,and realizes the synchronous extrapolation of source and receiver wavefields.Among the existing source wavefield reverse-time reconstruction algorithms,the random boundary algorithm has been widely used in three-dimensional(3D)ERTM because it requires the least storage of temporary files and low-frequency disk access during reverse-time migration.However,the existing random boundary algorithm cannot completely destroy the coherence of the artificial boundary reflected wavefield.This random boundary reflected wavefield with a strong coherence would be enhanced in the cross-correlation image processing of reverse-time migration,resulting in noise and fictitious image in the migration results,which will reduce the signal-to-noise ratio and resolution of the migration section near the boundary.To overcome the above issues,we present an ERTM random boundary-noise suppression method based on generative adversarial networks.First,we use the Resnet network to construct the generator of CycleGAN,and the discriminator is constructed by using the PatchGAN network.Then,we use the gradient descent methods to train the network.We fix some parameters,update the other parameters,and iterate,alternate,and continuously optimize the generator and discriminator to achieve the Nash equilibrium state and obtain the best network structure.Finally,we apply this network to the process of reverse-time migration.The snapshot of noisy wavefield is regarded as a 2D matrix data picture,which is used for training,testing,noise suppression,and imaging.This method can identify the reflected signal in the wavefield,suppress the noise generated by the random boundary,and achieve denoising.Numerical examples show that the proposed method can significantly improve the imaging quality of ERTM.

STOCHASTIC ANALYSIS OF GROUNDWATER FLOW SUBJECT TO RANDOM BOUNDARY CONDITIONS

A stochastic model was developed to simulate the flow in heterogeneous media subject to random boundary conditions. Approximate partial differential equations were derived based on the Karhunen-Loeve （KL） expansion and perturbation expansion. The effect of random boundary conditions on the two-dimensional flow was examined. It is shown that the proposed stochastic model is efficient to include the random boundary conditions. The random boundaries lead to the increase of head variance and velocity variance. The influence of the random boundary conditions on head uncertainty is exerted over the whole simulated region, while the randomness of the boundary conditions leads to the increase of the velocity variance in the vicinity of boundaries.

THE NATURAL BOUNDARY OF SOME RANDOM POWER SERIES

Suppose that {X(n)(omega)} are independent random complex variable sequence, E(X(n)) = 0 and [GRAPHICS] (V(X(n) = sigma(n)2). If reversed capital E-epsilon > 0 such that for all P (H) > 1-epsilon, we have [GRAPHICS] Then the circle {\Z\ = rho} is almost surely a natural boundary of the random series [GRAPHICS]

Statistical second-order two-scale analysis and computation for heat conduction problem with radiation boundary condition in porous materials

This paper discusses a statistical second-order two-scale（SSOTS） analysis and computation for a heat conduction problem with a radiation boundary condition in random porous materials.Firstly,the microscopic configuration for the structure with random distribution is briefly characterized.Secondly,the SSOTS formulae for computing the heat transfer problem are derived successively by means of the construction way for each cell.Then,the statistical prediction algorithm based on the proposed two-scale model is described in detail.Finally,some numerical experiments are proposed,which show that the SSOTS method developed in this paper is effective for predicting the heat transfer performance of porous materials and demonstrating its significant applications in actual engineering computation.

Optimizing the wavefield storage strategy in reflection-acoustic logging reverse-time migration

In this paper,wavefield storage optimization strategies are discussed with respect to reverse-time migration(RTM)imaging in reflection-acoustic logging,considering the problem of massive wavefield data storage in RTM itself.In doing so,two optimization methods are proposed and implemented to avoid wavefield storage.Firstly,the RTM based on the excitation-amplitude imaging condition uses the excitation time to judge the imaging time,and accordingly,we only need to store a small part of wavefield,such as the wavefield data of dozens of time points,the instances prove that they can even be imaged by only two time points.The traditional RTM usually needs to store the wavefield data of thousands of time points,compared with which the data storage can be reduced by tens or even thousands of times.Secondly,the RTM based on the random boundary uses the idea that the wavefield scatters rather than reflects in a random medium to reconstruct the wavefield source and thereby directly avoid storing the forward wavefield data.Numerical examples show that compared with other migration algorithms and the traditional RTM,both methods can effectively reduce wavefield data storage as well as improve data-processing efficiency while ensuring imaging accuracy,thereby providing the means for high-efficiency and highprecision imaging of fractures and caves by boreholes.

CYCLIC DEFORMATION OF COARSE GRAINED POLYCRYSTALLINE PURE Ai——Ⅰ.SLIP CHARACTERISTICS

Comparing with ordinary ploycrystalline materials sized to μm grade,the slip morphology of the coarse grained polycrystalline pure Al is characterized by:(1)several slip domains occur in a grain,and in same domain,several slip systems operate at same time or one after another intensely,a beautiful and neat slip pattern is forming on the specimen surface;(2)for high Σ-value coincident and random grain boundaries,the grain boundary affecting zone (GBAZ),bout 50—120μm wide,is favourable site to form intergranular crack at early fa- tigue life easily,and migration or slide of the boundaries were often observed.While low Σ-value near-coincident grain boundaries show a higher degree of slip continuity and strain compatibility than high Σ-value ones.Intergranular crack is not easily nucleated at low Σ-value near-coincident boundaries;and(3)due to suppression of grain boundary slip at triple grain boundary node,the high Σ-value and random grain boundary among the three boundaries of tricrystal crack easily during cyclic deformation.

Retarding the precipitation of η phase in Fe-Ni based alloy through grain boundary engineering

It is important to inhibit the precipitation of η phases in precipitation strengthened Fe-Ni based alloys,as they will deteriorate not only the mechanical property but also the hydrogen resistance.The present investigation shows that grain boundary engineering(GBE) can retard the formation and growth of ηphase in J75 alloy.After GBE treatment with 5% cold rolling followed by annealing at 1000℃ for 1 h,the fraction of special boundaries(SBs) increases from 38.4% in conventional alloy to 77.2% and the fraction of special triple junctions increases from 10% to 74%.During 800℃ aging treatment,quite amount of cellular η phases adjacent to random grain boundary(RGB) will be found in conventional alloy,and only a few small η phases have been observed in GBE treatment alloy subjected to the same aging treatment for long time.The reason for GBE in inhibiting precipitation of η phase can be attributed to not only introducing high fraction of SBs but also breaking the connectivity of RGB networks.As nucleation and growth of η phases on SBs are difficult due to their lower Ti concentration and diffusion rate,and the disruption of RGB networks reduces supply of Ti atoms to the η phases significantly,which impedes their growth at RGB.