Search

Deep neural networks can significantly improve the quality of super-resolution. However, previous work has made insufficient use of low-resolution scale features and channel-wise information, hence hindering the representational ability of CNNs. To address these issues, a multi-scale fractal residual attention network (MFRAN) is proposed. Specifically, MFRAN consists of fractal residual blocks (FRBs), dual-enhanced channel attention (DECA), and dilated residual attention blocks (DRABs). Among them, FRB applies multi-scale extension rule to continuously expand into a fractal structure that detects multi-scale features; DRAB constructs a combined dilated convolution to learn a generalizable and expressive feature space with a larger receptive field; DECA employs one-dimensional convolution to achieve cross-channel information interaction, and enhance the flow of information between groups by channel shuffling. Then, we integrate horizontal feature representations via local residual and feature fusion. Extensive quantitative and qualitative evaluations of benchmark datasets show that our proposed approach outperforms state-of-the-art methods in terms of quantitative metrics and visual results. [Display omitted] • We propose a Multi-scale Fractal Residual Attention Network (MFRAN) for SISR, where Dual-Enhanced Channel Attention can capture inter-channel dependencies more efficiently and lightly while enabling cross-channel information interaction for enhanced channel modeling capabilities, so that the model can reconstruct SR images with richer details at large-scale factors; Fractal Residual Block with multiple local feature extraction paths of different lengths, each path consists of dilated residual attention blocks with different size receptive fields, so that LR features of different scales can be extracted efficiently. [ABSTRACT FROM AUTHOR]

6DoF camera localization is an important component of autonomous driving and navigation. Deep learning has achieved impressive results in localization, but its robustness in dynamic environments has not been adequately addressed. In this paper, we propose a framework based on hybrid attention mechanism which can be generally applied to existing CNN-based pose regressors to improve their robustness in dynamic environments. Specifically, we propose a novel Transformer Bottleneck (TBo) block including convolution, channel attention, and a position-aware self-attention mechanism, which extracts more geometrically robust features by capturing the corresponding long-term dependencies between pixels. Furthermore, we introduce shuffle attention (SA) before the pose regressor, which integrates feature information in both spatial and channel dimensions, forcing the network to learn geometrically robust features, reducing the effects of dynamic objects and illumination conditions to improve camera localization accuracy. We evaluate our method on commonly benchmarked indoor and outdoor datasets and the experimental results show that our proposed method can significantly improve localization performance compared compare favorably to contemporary pose regressors schemes. In addition, extensive ablation evaluations are conducted to prove the effectiveness of our proposed hybrid attention bottleneck block for pose regression networks. • We propose a novel Transformer Bottleneck block with self-attention and channel attention to overcome the limitations of convolution. This coupling allows them to be optimized in a mutually reinforcing manner, significantly improving fine-grained feature extraction for accurate localization. • We propose a novel end-to-end hybrid attention network for single image localization, which improves the accuracy and robustness of camera localization, especially in dynamic scenes. • We conduct extensive experiments in both indoor and outdoor dataset, which shows that our model performs better than the existing competitive methods. [ABSTRACT FROM AUTHOR]

A multi-state system (MSS) employs more than two discrete states to indicate different performance rates. Methods using a universal generating function (UGF) and Monte Carlo (MC) simulation are primary approaches for the reliability analysis of an MSS. However, these approaches incur a large computational overhead because the number of system states increases significantly with the number of components in an MSS. In this paper, stochastic multi-valued (SMV) models are proposed for evaluating the reliability of an MSS with dependent multi-state components (MSCs). The performance rates and their corresponding probabilities of the MSCs are simultaneously encoded in multi-valued non-Bernoulli sequences using permutations of fixed numbers of 1 s and 0 s. The sequences are then processed by logic gates. The effectiveness of the proposed approach is demonstrated via a comparative evaluation of a multi-state system consisting of dependent components with steady and time-varying state probabilities. [ABSTRACT FROM AUTHOR]

The mechanical performance of concrete is primarily dominated by the characteristics and interrelation of its ingredients, especially the content, shape and grading of coarse aggregates. Consequently, constructing a realistic mesostructure of concrete is essential for adequate mesoscale studies on the corresponding mechanical properties. In this study, a novel three-dimensional coarse aggregate generation scheme, namely the random walking algorithm (RWA), is proposed for constructing physically feasible mesostructures of concrete. The proposed approach is able to generate a series of aggregates within an initial placing domain, and subsequently, move them into the target domain by both translation and rotation. Within the proposed analysis framework, the high compactness of mesostructures with an improved aggregate content can be robustly achieved by randomly shifting previously placed aggregates, such that the later generated ones can be ingeniously blended in. Typical samples of random aggregate structure (RAS) are generated under specified grading curves. Parameters relating to aggregate content and efficiency of modelling are critically evaluated. By thoroughly investigating practically motivated examples, it is evidently illustrated that the present method is capable of obtaining a relatively realistic and random distribution of coarse aggregates, and more importantly, the grading of the generated aggregate samples is in compliance with the Fuller's Curve. [ABSTRACT FROM AUTHOR]

The spatial pattern and rate of strain accumulation on a fault during the pre- and inter-seismic phases are very important for interpreting the mechanism of earthquakes and evaluating seismic potentials. Here we use global positioning system (GPS) data and the block-dislocation model to invert for the locking degree and slip rate deficit of the Main Himalayan Thrust (MHT) fault in the southern margin of Tibet before the 2015 M w 7.9 Nepal earthquake. Results show that the locking depth and slip rate deficit increase from the west to the east. Along the western segment of the MHT fault (80°E–84°E), the locking depth is estimated to be 12–17 km with a slip rate deficit of 0–5 mm/a; along the central Nepal segment (84°E–87°E), the locking depth is 16–21 km with a slip rate deficit of 6–10 mm/a, whilst along the eastern segment (87°E–90°E), the locking depth increases to 23–26 km with a slip rate deficit of 8–13 mm/a. The 2015 Nepal earthquake initiated at the boundary between the western and central segments, an area with as where the slip rate deficit varies dramatically from 0 to 9 mm/a within 50 km resulting in high energy gradients. High strain concentration along the central and eastern segments leads to unilateral propagation of the rupture to the east. Given the paucity of large seismic events over the previous decades and the current high slip rate deficit, seismic hazard on the eastern Nepal segment remains high. [ABSTRACT FROM AUTHOR]

Three-dimensional (3-D) fault geometry and surface slip are key factors for understanding fault behaviour and earthquake mechanics. Previous studies of the 2008 Yutian earthquake occurred on the northwestern Tibetan Plateau, and are limited by a lack of detailed measurements of fault structures and near-fault motion due to the inaccessibility and extreme weather of the area and severe decorrelation of InSAR. In this paper, we use a very-high resolution (VHR) DEM of the 2008 rupture zone extracted from the Pléiades satellite stereo imagery to measure the 3-D fault geometry and on-fault slip. The maximum vertical and horizontal coseismic displacement is 3.6 m and 2.8 m respectively. Despite the overall normal faulting, near-fault reverse vertical motion was observed, ranging from 0.5 m to 1.9 m, caused by the transformation of strike-slip motion into localised shortening due to local fault geometric variations. Measurements of the near-surface dip for the three fault segments (55°, 62°, and 58°) show good consistency, different from previous studies (43°, 60°, and 52°). Joint inversion of InSAR and the near-fault measurements of surface slip also show that the fault geometry does not change much along strike. The local ratio of the cumulative to the coseismic vertical offset recorded on an old alluvial fan and an ice tongue ranges from 1.86 to 2.79, suggesting at least 1–2 historical events occurred here before the 2008 earthquake. • 3D fault geometry and slip of Yutian earthquake from very-high resolution DEM • Near-fault reverse vertical motion caused by local fault geometric variations • Measurements and inversion results show good consistency for dips along strike. [ABSTRACT FROM AUTHOR]

Previous studies using paleoseismic trenches and numerical models proposed two different fault rupture models for the Pingdingshan-Annanba segments of the Altyn Tagh Fault, which are both poorly validated because of a lack of measurements of surface rupture displacements. In this paper, we use high-resolution satellite imagery to measure offset geomorphic markers. Taking the fault geometry into consideration, the most likely paleoseismic rupture parameters are determined by comparing the results of the cumulative offset probability densities (COPDs) and the empirical relationships of seismic parameter. The results show that the most recent earthquake (1270A.D.-1775A.D.) along the Suoerkuli segment only ruptured the Suoerkuli Valley, with a surface rupture length of 140 km, an average horizontal displacement of 3.12 m and a moment magnitude (Mw) of 7.4–7.6. The penultimate earthquake (676A.D.-1347A.D.) ruptured both the Annanba segment and part of the Suoerkuli segment with an average horizontal displacement of greater than 5 m, a surface rupture length of 120–260 km and Mw 7.4–7.8. The rupture parameters of the above two earthquakes demonstrate that the Suoerkuli and Annanba segments are individual fault segments which can rupture independently with a surface rupture length more than 100 km. The Pingdingshan double bend to the west of the Suoerkuli segment is likely to be an endpoint that inhibits the rupture propagation towards further west. [ABSTRACT FROM AUTHOR]

Abstract The Mw6.9 Yushu earthquake occurred on 14 April 2010, in Qinghai, China; the largest aftershock, a Mw5.8 event, occurred west of the mainshock on 29 May 2010 (˜40 days later). The aftershock had a different focal mechanism from the mainshock. Furthermore, seismicity after 29 May showed different spatial characteristics in terms of focal depth and distribution direction. To better understand the faulting and the relationship between these two events, we derived the whole displacement field caused by the Yushu mainshock and the Mw5.8 aftershock based on multi-perspective Interferometric Synthetic Aperture Radar (InSAR) data. We then conducted a robust inversion of the slip distribution jointly constrained by InSAR and GPS data. The results indicate that the Mw5.8 aftershock produced a separated deformation field with significant displacement changes of up to ˜4–6 cm, which indicates another intersecting ruptured fault at the west end of the Yushu fault. The slip distribution shows a 75-km NW rupture with a maximum slip of ˜2.1 m at a depth of ˜0–10 km on the main Yushu fault, and a 20 km NE rupture with peak slip of ˜0.4 m at a depth of ˜5–15 km on a vertical hidden fault. Both events showed a dominant left-lateral component. The total rupture length associated with the 2010 Yushu earthquake sequence reached ˜95 km. By calculating Coulomb stress changes, we confirmed that the mainshock triggered the Mw5.8 aftershock. Our results imply that the increased stress at the western end of the Yushu fault caused by the mainshock rupture may have played an important role in transferring the rupture plane from the Yushu fault to the NE hidden fault. [ABSTRACT FROM AUTHOR]

Based on the dense GPS velocity field in the northeastern margin of the Tibetan Plateau from 1999 to 2016, we have produced the deformation and strain characteristics of the Haiyuan fault and the Liupanshan fault. Estimated long-term slip rate along the Haiyuan-Liupanshan fault zones show a gradual decrease from 6.4 ± 1.6 mm/yr at the Tuolaishan fault to 2.9 ± 1.2 mm/yr at the Southern Liupanshan fault. Left-lateral thrusting movement was inverted for the Xiangshan-Tianjingshan fault (XS-TJS), which has an average slip rate of 2.1 ± 3.4 mm/yr during the study period. We also calculated the heterogeneous distribution of interseismic coupling along the fault zones. Our result also shows the locking depth of the Tianzhu seismic gap is ∼22 km. The slip rate deficit, the seismic moment accumulation rate, and the Coulomb stress accumulation rate are high on the fault planes, whereas the second invariant of the strain rate is low at the surface. The Liupanshan fault is locked to a depth of ∼23 km, and the corresponding seismic moment accumulation rate on the fault plane is high, while the strain rate at the surface is low. The accumulated strain along the Tianzhu seismic gap and the Liupanshan fault could be balanced by earthquakes with magnitudes of Mw7.9 and Mw7.4, considering the absence of large earthquakes over the last 1000 years and 1400 years respectively. The Haiyuan segments had ruptured during 1920 Haiyuan earthquake, and the estimated locking depth for period 1999–2016 is 5–10 km. Its seismic moment accumulation rate at depth is low and the strain rate at the surface is high. Our result indicates that 70% of the strike-slip along the Haiyuan segments transforms into thrusting along the Liupanshan fault, while the remaining 30% is related to the orogeny of the Liupanshan. For slip between the Haiyuan fault and the XS-TJS, about 27–34% of the slip is partitioned on the XS-TJS. [ABSTRACT FROM AUTHOR]

This study comprehensively investigated the FRP-concrete debonding mechanism in the pull-off test using a mesoscale cohesive zone modelling approach. Pull-off tests were performed on the FRP-strengthened concrete elements, and corresponding 2-D mesoscale finite element models were established. The numerical stress-separation responses and strain/crack initiation and propagation in the pull-off test were examined. Subsequently, a parametric study numerically investigated the effect of adhesive and FRP properties, concrete heterogeneity, loading fixture stiffness, and sample scale on the debonding mechanism. Finally, the 2-D mesoscale model was extended to the 3-D. The main conclusions are: (1) global separation between FRP and concrete in the pull-off test is minimal; (2) the normal bond strength of the epoxy resin-concrete interface controls the failure mode; however, FRP stiffness does not affect the result; (3) the aggregate content and mortar porosity significantly influence the bond strength, and the effect of aggregate shape and gradation is slight; (4) the loading fixture stiffness should increase with an increase in sample size; (5) the pull-off bond strength aligns with the size effect theory; (6) the bond stiffness and strength of the 3-D model is significantly greater than that of the 2-D model, which is attributed to the out-of-plane concrete constraining. [ABSTRACT FROM AUTHOR]

Coseismic deformation fields caused by the moment magnitude (Mw)7.8 mainshock and Mw7.3 aftershock of the 2015 Gorkha-Nepal earthquake are obtained by analyzing Sentinel-1A/IW ascending and descending interferometry data. Results show that the deformation field associated with the Mw7.8 mainshock roughly resembles a prolate ellipse, extending from the epicenter about 20° east by south. The main region of deformation is about 160 km by 110 km, comprising a large southern area of uplift, and a small northern area of subsidence. Assuming that rupture occurred in a homogeneous elastic half-space, the coseismic fault slip models of the mainshock and aftershock are inverted based on a shallow dip fault constrained by the three data sets, Sentinel-1A/IW descending data, ascending data, and ALOS-2 descending data, separately or in combination. Mainshock slip distributions generated from all three data sets are similar, and inversion constrained by all three in combination reveal a comprehensive fault slip model. Indeed, coseismic slip is mainly distributed within a narrow 40 km zone to the north of the Main Frontal Trust (MFT), and at 6–15 km subsurface depth. In addition, the maximum slip in this event was about 5.1 m, the Mw7.8 mainshock ruptured the deep part of the seismogenic zone, while the region between the southern boundary of the rupture area and the MFT remained locked. Therefore, a considerable earthquake risk remains to the south of Kathmandu. The inverted coseismic slip of the Mw7.3 aftershock was concentrated in a small area, close to, and southeast of the epicenter, with maximum displacement of about 3 m. Finally, because there is no overlap between the two slip areas of the mainshock and aftershock, the gap between them, about 15 km in length, has additional potential to generate future earthquakes. [ABSTRACT FROM AUTHOR]

Abstract: We obtained the coseismic deformation field of the Ms7.1 earthquake in Yushu, Qinghai, China on 14 April 2010 using L band and C band SAR images. The results show that the deformation fields derived from L and C bands separately are consistent in general. They are all characterized by concentric elliptic interference fringes surrounding a NW trending fault with two local intensively deformed regions. Meanwhile they have small differences in area coverage of the deformation field, magnitude of LOS displacements and other details of deformation nearby the fault. The deformation from C band and L band covers 89∗59km square and 77∗43km square, with maximum displacements of about 45cm and 65cm in the radar line of sight, respectively. Using these two kinds of InSAR measurements separately and jointly, we constructed a three-segment fault model and inverted the coseismic fault slip of the Yushu event. The inversion results show that the slip distributions constrained by the two kinds of inSAR data are in agreement on the whole. They all indicate two concentration regions of slip distribution, one is located near the Jiegu town on the southeastern segment of the Yushu fault with a large area and the maximum slip of 2.4m, and the other is at the epicenter on the northwestern segment of the fault with a relatively small area. There are also discrepancies between the two kinds of slip distribution models. In general, the scale, magnitude, and depth of slip distribution constrained by C band ASAR data are larger than that by L band PALSAR data. The slip distribution from a joint inversion of the two kinds of SAR data is considered to be the most robust model compared to the results of that constrained by L band SAR data or C band SAR data alone. [Copyright &y& Elsevier]

Abstract: On 21 March 2008, an Ms7.3 earthquake occurred at Yutian County, Xinjiang Uygur Autonomous Region, which is in the same year as 2008 Mw 7.9 Wenchuan earthquake. These two earthquakes both took place in the Bayar Har block, while Yutian earthquake is located in the west edge and Wenchuan earthquake is in the east. The research on source characteristics of Yutian earthquake can serve to better understand Wenchuan earthquake mechanism. We attempt to reveal the features of the causative fault of Yutian shock and its co-seismic deformation field by a sensitivity-based iterative fitting (SBIF) method. Our work is based on analysis and interpretation to high-resolution satellite (Quickbird) images as well as D-InSAR data from the satellite Envisat ASAR, in conjunction with the analysis of seismicity, focal mechanism solutions and active tectonics in this region. The result shows that the 22km long, nearly NS trending surface rupture zone by this event lies on a range-front alluvial platform in the Qira County. It is characterized by distinct linear traces and a simple structure with 1–3m-wide individual seams and maximum 6.5m width of a collapse fracture. Along the rupture zone are seen many secondary fractures and fault-bounded blocks by collapse, exhibiting remarkable extension. The co-seismic deformation affected a big range 100km×40km. D-InSAR analysis indicates that the interferometric deformation field is dominated by extensional faulting with a small strike-slip component. Along the causative fault, the western wall fell down and the eastern wall, that is the active unit, rose up, both with westerly vergence. The maximum subsidence displacement is ∼2.6m in the LOS, and the maximum uplift is 1.2m. The maximum relative vertical dislocation reaches 4.1m, which is 10km distant from the starting rupture point to south. The 42km-long seismogenic fault in the subsurface extends in NS direction as an arc, and it dipping angle changes from 70° near the surface to 52° at depth ∼10km. The slip on the fault plane is concentrated in the depth range 0–8km, forming a belt of length 30km along strike on the fault plane. There are three areas of concentrating slip, in which the largest slip is 10.5m located at the area 10km distant from the initial point of the rupture. [Copyright &y& Elsevier]