Search

In this study, we investigated a means to improve the robustness of deep network training on visual recognition tasks without sacrificing accuracy. The contribution of this work can reduce the dependence on model decay to gain a strong defense against malicious attacks, especially from adversarial samples. There are two major challenges in this study. First, the model defense capability should be strong and improved over the training stage. The other is that the degrading of the model performance must be minimized to ensure visual recognition performance. To tackle these challenges, we propose active dropblock (ActDB) by incorporating active learning into a dropblock. Dropblock effectively perturbs the feature maps, thus enhancing the invulnerability of gradient-based adversarial attacks. In addition, it selects an optimal perturbation solution to minimize the objective loss function, thereby reducing the model degradation. The proposed organic integration successfully solved the model robustness and accuracy simultaneously. We validated our approach using extensive experiments on various datasets. The results showed significant gains compared to state-of-the-art methods.

For visual tracking methods based on reinforcement learning, action space determines the ability of exploration, which is crucial to model robustness. However, most trackers adopted simple strategies with action space, which will suffer local optima problem. To address this issue, a novel reinforcement learning based tracker called AEVRNet is proposed with non-convex optimization and effective action space exploration. Firstly, inspired by combinatorial upper confidence bound, we design an adaptive exploration strategy leveraging temporal and spatial knowledge to enhance effective action exploration and jump out of local optima. Secondly, we define the tracking problem as a non-convex problem and incorporate non-convex optimization in stochastic variance reduced gradient as backward propagation of our model, which can converge faster with lower loss. Thirdly, different from existing reinforcement learning based trackers using classification method to train model, we define a regression based action-reward loss function, which is more sensitive to aspects of the target states, e.g., the width and height of the target to further improve robustness. Extensive experiments on six benchmark datasets demonstrate that our proposed AEVRNet achieves favorable performance against the state-of-the-art reinforcement learning based methods.

After Vogel proposed a dimensionless inflow performance equation, with the rise of the horizontal well production mode, a large number of inflow performance relationship (IPR) equations have emerged. In the productivity analysis of deviated and horizontal wells, the IPR equation proposed by Cheng is mainly used. However, it is still unclear whether these inflow performance models (such as the Cheng, Klins-Majcher, Bendakhlia-Aziz, and Wiggins-Russell-Jennings types) are suitable for productivity evaluations of horizontal and deviated wells in low-permeability reservoirs. In-depth comparisons and analyses have not been carried out, which hinders improvements in the accuracy of the productivity evaluations of horizontal wells in low-permeability reservoirs. In this study, exploratory work was conducted in two areas. First, the linear flow function relationship used in previous studies was improved. Based on the experimental pressure-volume-temperature results, a power exponential flow function model was established according to different intervals greater or less than the bubble point pressure, which was introduced into the subsequent derivation of the inflow performance equation. Second, given the particularity of low-permeability reservoir percolation, considering that the reservoir is a deformation medium, and because of the existence of a threshold pressure gradient in fluid flow, the relationship between permeability and pressure was changed. The starting pressure gradient was introduced into the subsequent establishment of the inflow performance equation. Based on the above two aspects of this work, the dimensionless IPR of single-phase and oil-gas two-phase horizontal wells in a deformed medium reservoir was established by using the equivalent seepage resistance method and complex potential superposition principle. Furthermore, through regression and error analyses of the standard inflow performance data, the correlation coefficients and error distributions of six types of IPR equations applicable to deviated and horizontal wells at different inclination angles were compared. The results show that the IPR equation established in this study features good stability and accuracy and that it can fully reflect the particularity of low-permeability reservoir seepage. It provides the best choice of the IPR between inclined wells and horizontal wells in low-permeability reservoirs. The other types of IPR equations are the Wiggins-Russell-Jennings, Klins-Majcher, Vogel, Fetkovich, Bendakhlia-Aziz, and Harrison equations, listed here in order from good to poor in accuracy.

A subsecond and novel approach of electroshock treatment (EST) is used in this study to repair defects in directed-energy-deposited Ti-5Al-5Mo-5V-3Cr-1Zr near-β titanium alloy. After EST, the porosity of the specimen decreased significantly from 0.81 to 0.1 pct. Large cracks observed at the bottom of the above mentioned near-β titanium alloy became intermittent small cracks and the number of voids decreased. The defects in the top and middle regions of the specimens are repaired. The potential defect repair is attributable to energy concentration, which promoted the coalescence of defect tips, and thermal stresses, which compressed the defects inward and closed them.

Friction stir lap welding (FSLW) was employed to weld dissimilar joints of zirconium and stainless steel for the first time. The microstructure, evolution mechanism of intermetallic compounds (IMCs) and mechanical properties of joint were investigated. Discontinuous dynamic recrystallization (DDRX) and continuous dynamic recrystallization (CDRX) occurred in zirconium, while DDRX occurred in stainless steel during FSLW. The formation sequence of IMCs at the interface was FeZr3, FeZr2, Fe2Zr, and Fe23Zr6. The crack nucleated at the hook zone on the advancing side (HookAS), and IMCs in the Zr/SS mixed zone accelerated crack propagation. The fracture surface exhibited ductile-brittle mixed fracture characteristic.

The mammalian brain has enormously complex neuronal diversity and a highly modular structure. The propagation of information in the modular brain network can be modeled by a feedforward network (FFN). Although studies in this area have yielded many important results, neuronal diversity has rarely been considered. In the current work, we investigate the complex interactions between the intrinsic properties of neurons and the FFN structure in the propagation of spiking activity. Here, four typical types of cortical neurons reproduced by the Izhikevich neuron model are introduced. A homogeneous FFN composed of a single type of excitatory neuron (regular spiking, mixed model, or tonic bursting) can propagate spiking activity. However, an FFN with fast spiking neurons does not propagate spiking activity. By modifying the network structure and synaptic weights, the spiking propagation of the homogeneous FFNs can vary from synchronous transmission (with a high firing rate) to asynchronous transmission (with a low firing rate). Among the homogeneous FFNs, both the firing rate and the synchrony of the FFN with tonic bursting neurons are the highest, but those of the FFN with regular spiking neurons is lowest, even when implementing the same FFN structure. For the FFN with mixed neuronal types, interestingly, the spiking propagation is very sensitive to the composition of the four types of neurons. By introducing fast spiking neurons into the homogeneous FFN composed of excitatory neurons, spiking propagation can be modified from synchronous to asynchronous. Similarly, changing the proportion of any of the types of neuron affects the spiking propagation, even for very small changes. The underlying mechanism of these observed results has also been discussed.

Siamese network is highly regarded in the visual object tracking filed because of its unique advantages of pairwise input and pairwise training. It can measure the similarity between two image patches, which coincides with the principle of the matching-based tracking algorithm. In this paper, a variant Siamese network based tracker is proposed to introduce attention module into traditional Siamese network, and relocate the object with some auxiliary relocation methods, when the proposed tracker runs under an untrusted state. Firstly, a novel attention shake layer is proposed to replace the max pooling layer in Siamese network. This layer could introduce and train two different kinds of attention modules at the same time, which means the proposed attention shake layer could also help to improve the expression power of Siamese network without increasing the depth of the network. Secondly, an auxiliary relocation branch is proposed to assist in object relocation and tracking. According to the prior assumptions of visual object tracking, some weights are involved in the auxiliary relocation branch, such as structure similarity weight, motion similarity weight, motion smoothness weight and object saliency weight. Thirdly, a novel response map based switch function is proposed to monitor the tracking process and control the effect of auxiliary relocation branch. Furthermore, in order to discuss the effect of pooling layer in Siamese network, 9 pooling and attention architectures are proposed and discussed in this paper. Some empirical results are shown in the experiment part. Comparing with the state-of-the-art trackers, the proposed tracker could achieve comparable performance in multiple benchmarks.

The outbreak of coronavirus named COVID-19, initially identified in Wuhan, China in December 2019, has spread rapidly at the global scale. Most countries have rapidly stopped almost all activities including industry, services and transportation of goods and people, thus decreasing air pollution in an unprecedented way, and providing a unique opportunity to study air pollutants. While satellite data have provided visual evidence for the global reduction in air pollution such as nitrogen dioxide (NO2) worldwide, precise and quantitative information is missing at the local scale. Here we studied changes in particulate matter (PM2.5, PM10), carbon monoxide (CO), NO2, sulfur dioxide (SO2) and ozone (O3) at 10 urban sites in Hangzhou, a city of 7.03 million inhabitants, and at 1 rural site, before city lockdown, January 1–23, during city lockdown, January 24-February 15, and during resumption, February 16–28, in 2020. Results show that city lockdown induced a sharp decrease in PM2.5, PM10, CO, and NO2 concentrations at both urban and rural sites. The NO2 decrease is explained by reduction in traffic emissions in the urban areas, and by lower regional transport in rural areas during lockdown, as expected. SO2 concentrations decreased from 6.3 to 5.3 μg m−3 in the city, but increased surprisingly from 4.7 to 5.8 μg m−3 at the rural site: this increase is attributed both to higher coal consumption for heating and emissions from traditional fireworks of the Spring Eve and Lantern Festivals during lockdown. Unexpectedly, O3 concentrations increased by 145% from 24.6 to 60.6 μg m−3 in the urban area, and from 42.0 to 62.9 μg m−3 in the rural area during the lockdown. This finding is explained by the weakening of chemical titration of O3 by NO due to reductions of NOx fresh emissions during the non-photochemical reaction period from 20:00 PM to 9:00 AM (local time). During the lockdown, compared to the same period in 2019, the daily average concentrations in the city decreased by 42.7% for PM2.5, 47.9% for PM10, 28.6% for SO2, 22.3% for CO and 58.4% for NO2, which is obviously explained by the absence of city activities. Overall, we observed not only the expected reduction in some atmospheric pollutants (PM, SO2, CO, NO2), but also unexpected increases in SO2 in the rural areas and of ozone (O3) in both urban and rural areas, the latter being paradoxically due to the reduction in nitrogen oxide levels. In other words, the city lockdown has improved air quality by reducing PM2.5, PM10, CO, and NO2, but has also decreased air quality by augmenting O3 and SO2. Electronic supplementary material The online version of this article (10.1007/s10311-020-01028-3) contains supplementary material, which is available to authorized users.

Friction stir processing (FSP) is a new severe plastic deformation technique where microstructure refinement, homogenization, and densification occur simultaneously. In this study, the microstructure and superplastic deformation behavior of AZ80 magnesium alloy prepared by FSP were investigated. FSP led to remarkable grain refinement and modification of texture for AZ80 magnesium alloy. A maximum elongation of 606% was obtained at 3 × 10−4 s−1 and 350 ℃. When superplastic deformation was conducted at 3 × 10-3 s−1 and 300 ℃, or at 1 × 10-4 s−1 and 400 ℃, the superplasticity was relatively lower, thus exhibiting a lower strain rate sensitivity. The coarsening of both grains and β-Mg17Al12 precipitates deteriorated superplastic elongation. Applied deformation promoted the premature dissolution of β precipitates at 400 ℃ during superplastic deformation. Moreover, superplastic deformation weakened and spread the basal texture. Although the activation of prismatic and pyramidal slips is beneficial for superplastic deformation, grain boundary sliding is the main superplastic deformation mechanism.

NiTi-Nb porous shape memory scaffold was in-situ sintered by eutectic reaction using NiTi powder, Nb powder and Ti wire at 1180 °C. Ti2Ni and Ni4Ti3 phases are precipitated between the eutectic region and Ti wire, while the R martensite phase and Nb-Rich phases are formed in both eutectic region and the pre-eutectic region. Moreover, the stacking faults and micro-twinning in the eutectic region are beneficial to nucleation of R-phase martensite. All results demonstrate that the eutectic reaction combined with the interaction of Ti wire, NiTi powder and Nb powder contribute to tremendous increase in compressive strength up to 350 MPa.

It is a persistent target to develop noble-metal-free catalysts that can achieve chemoselective hydrogenation of nitro compounds under mild reaction conditions. In this work, we report a new strate...

Electro-shocking treatment (EST) has been investigated as a pathway to optimise the microstructure and mechanical properties of titanium alloys. The thermal conditions introduced by EST resulted in a phase transformation from α to β. The fraction of β phase decreased from 25.27% to 19.47% after EST for 0.02 s, which was possibly caused by the recrystallization of α phase. The application of EST for 0.04 s resulted in an increase in volume fraction of the β phase to 26.95%. The energy introduced by EST resulted in changes to the direction and intensity of texture within the microstructure with the texture intensity of the α phase increasing from 4.94 to 8.52, and that of β both increased from 3.35 to 9.88 after 0.04 s EST. Keywords: Electroshocking treatment (EST), Microstructure, Phase transformation, Phase content, Texture, TC11 alloy

Electronic skin, a class of wearable electronic sensors that mimic the functionalities of human skin, has made remarkable success in applications including health monitoring, human-machine interaction and electronic-biological interfaces. While electronic skin continues to achieve higher sensitivity and faster response, its ultimate performance is fundamentally limited by the nature of low-frequency AC currents. Herein, highly sensitive skin-like wearable optical sensors are demonstrated by embedding glass micro/nanofibers (MNFs) in thin layers of polydimethylsiloxane (PDMS). Enabled by the transition from guided modes into radiation modes of the waveguiding MNFs upon external stimuli, the skin-like optical sensors show ultrahigh sensitivity (1870 kPa-1), low detection limit (7 mPa) and fast response (10 μs) for pressure sensing, significantly exceeding the performance metrics of state-of-the-art electronic skins. Electromagnetic interference (EMI)-free detection of high-frequency vibrations, wrist pulse and human voice are realized. Moreover, a five-sensor optical data glove and a 2×2-MNF tactile sensor are demonstrated. These initial results pave the way toward a new category of optical devices ranging from ultrasensitive wearable sensors to optical skins.

The effects of shot peening (SP) on mechanical behavior and surface morphology of Ti-6Al-4V were investigated in detail. After SP, the uniformity of compressive residual stress (CRS) was analyzed. The CRS decreased as SP intensity increased because of the severe impact of shots. The CRS was enhanced from − 762.7 to − 795.5 MPa with an increase in SP intensity. The fatigue life was increased from 0.46 × 106 cycles to 8.10 × 106 cycles after SP when the loading amplitude was 610 MPa. This increase was mainly attributed to the enhancement of the crack initiation hindering and propagation by the high CRS and refined domain sizes. The results of surface morphology showed that the convexity and concave regions were generated after SP and the improvement in SP intensity aggravated the increase in roughness. Although the roughness increased after SP, it did not hinder the enhancement of fatigue lives, indicating that CRS played a leading role in the improvement in fatigue lives. Thus, the analysis and discussion revealed that SP was beneficial for the improvement of the surface mechanical behavior of titanium alloys.

Detecting nodal activities in dynamic social networks has strategic importance in many applications, such as online marketing campaigns and homeland security surveillance. How peer-to-peer exchanges in social media can facilitate nodal activity detection is not well explored. Existing models assume network nodes to be static in time and do not adequately consider features from social theories. This research developed and validated two theory-based models, Random Interaction Model (RIM) and Preferential Interaction Model (PIM), to characterize temporal nodal activities in social media networks of human agents. The models capture the network characteristics of randomness and preferential interaction due to community size, human bias, declining connection cost, and rising reachability. The models were compared against three benchmark models (abbreviated as EAM, TAM, and DBMM) using a social media community consisting of 790,462 users who posted over 3,286,473 tweets and formed more than 3,055,797 links during 2013–2015. The experimental results show that both RIM and PIM outperformed EAM and TAM significantly in accuracy across different dates and time windows. Both PIM and RIM scored significantly smaller errors than DBMM did. Structural properties of social networks were found to provide a simple and yet accurate approach to predicting model performances. These results indicate the models’ strong capability of accounting for user interactions in real-world social media networks and temporal activity detection. The research should provide new approaches for temporal network activity detection, develop relevant new measures, and report new findings from large social media datasets.

Due to its speed and ease of use, Spark has become a popular tool amongst data scientists to analyze data in various sizes. Counter-intuitively, data processing workloads in industrial companies such as Google, Facebook, and Yahoo are dominated by short-running applications, which is due to the majority of applications being mostly consisted of simple SQL-like queries (Dean, 2004, Zaharia et al, 2008). Unfortunately, the current version of Spark is not optimized for such kinds of workloads. In this paper, we propose a novel framework, called Meteor, which can dramatically improve the performance for short-running applications. We extend Spark with three additional operating modes: one-thread, one-container, and distributed. The one-thread mode executes all tasks on just one thread; the one-container mode runs these tasks in one container by multi-threading; the distributed mode allocates all tasks over the whole cluster. A new framework for submitting applications is also designed, which utilizes a fine-grained Spark performance model to decide which of the three modes is the most efficient to invoke upon a new application submission. From our extensive experiments on Amazon EC2, one-thread mode is the optimal choice when the input size is small, otherwise the distributed mode is better. Overall, Meteor is up to 2 times faster than the original Spark for short applications.

The strengthening effect of laser remelting on Ti-35Nb-2Ta-3Zr alloy was studied in this study. The surface of the material changed from an equiaxed crystal to a needle-like structure after laser remelting. Dislocations and ω phases were observed during the process, and the orientation relationship between ω phase and β matrix was identified as [11 0]β// [ 11 2 - 0 ] ω . The microhardness of the laser-remelted alloy increased significantly.

There is a growing interest in designing models that can deal with images from different visual domains. If there exists a universal structure in different visual domains that can be captured via a common parameterization, then we can use a single model for all domains rather than one model per domain. A model aware of the relationships between different domains can also be trained to work on new domains with less resources. However, to identify the reusable structure in a model is not easy. In this paper, we propose a multi-domain learning architecture based on depthwise separable convolution. The proposed approach is based on the assumption that images from different domains share cross-channel correlations but have domain-specific spatial correlations. The proposed model is compact and has minimal overhead when being applied to new domains. Additionally, we introduce a gating mechanism to promote soft sharing between different domains. We evaluate our approach on Visual Decathlon Challenge, a benchmark for testing the ability of multi-domain models. The experiments show that our approach can achieve the highest score while only requiring 50% of the parameters compared with the state-of-the-art approaches.

In the past two decades, friction stir processing (FSP) technology has received considerable attention. FSP can be used to adjust and control the microstructure of materials, including eliminating defects, destroying dendrites and controlling fractions in the second stage, and is therefore widely used in titanium and its alloys for biomedical, aerospace and automotive applications. This article comprehensively reviews the methods of studying FSP, the microstructure evolution of materials and the summary of material properties. It begins with the introduction of the FSW system, characterization, structure and elemental analysis, and simulation and performance testing methods, and then introduces the microstructure evolution mechanism of various titanium materials and discusses in detail the material properties of titanium alloy, namely hardness and wear resistance, elasticity and plasticity, corrosion resistance and biocompatibility. Finally, this review presents unresolved issues and outstanding challenges in FSP technology and reveals the direction of this emerging field of research.

After high-load spherical indentation treatment, the variations of hardness on the plastic zone of Ti-6Al-4V were investigated via nanoindentation method. The hardness within the center of plastic zone was measured by nanoindenter, and the magnitude decreased gradually along the depth, which were caused by the different extent of plastic deformation under the residual imprint. The microstructure of indentation were observed using scanning electron microscope (SEM) before and after surface etching, and the results showed that the microhardness revealed the average hardness of α and β phases of Ti-6Al-4V. The maximum hardness reached 6.438 GPa in the depth of 132 μm. In addition, the two and three dimensional contour profiles of residual imprint introduced by high-load spherical indentation were measured by the white-light interferometer and the shape of residual imprint was obtained. All results were discussed in detail. Keywords: Nanoindentation, Spherical loading, Hardness, Microstructure, Ti-6Al-4V

The surface microstructure and mechanical properties of Ti-6Al-4V/Ag nanocomposite processed by friction stir processing (FSP) are studied in this work. Scanning electron microscope (SEM) results show the gradient grain size of surface layer. The pile-up and tangle of dislocations and the deformation twins are observed by transmission electron microscopy (TEM). Atom probe tomography (APT) analyses reveal the presence of Ag-enriched nanoparticles in both the interiors and boundaries of α-Ti grains. Severe plastic deformation via FSP is responsible for considerable dislocations pile-up and tangle in the deformation bands. The enhancement in hardness and elastic modulus is attributed to the grain refinement and formation of much more deformation twinning. Moreover, Ag agglomeration could accelerate the recrystallization process and prevent grain coarsening, resulting in the higher hardness and the larger elastic modulus. This study confirms FSP is effective on surface modification of titanium alloys for improving the surface mechanical properties.

As big data processing is being widely adopted by many domains, massive amount of generated data become more reliant on the parallel computing platforms for analysis, wherein Spark is one of the most widely used frameworks. Spark’s abnormal tasks may cause significant performance degradation, and it is extremely challenging to detect and diagnose the root causes. To that end, we propose an innovative tool, named LADRA, for log-based abnormal tasks detection and root-cause analysis using Spark logs. In LADRA, a log parser first converts raw log files into structured data and extracts features. Then, a detection method is proposed to detect where and when abnormal tasks happen. In order to analyze root causes we further extract pre-defined factors based on these features. Finally, we leverage General Regression Neural Network (GRNN) to identify root causes for abnormal tasks. The likelihood of reported root causes are presented to users according to the weighted factors by GRNN. LADRA is an off-line tool that can accurately analyze abnormality without extra monitoring overhead. Four potential root causes, i.e., CPU, memory, network, and disk I/O, are considered. We have tested LADRA atop of three Spark benchmarks by injecting aforementioned root causes. Experimental results show that our proposed approach is more accurate in the root cause analysis than other existing methods.

It is of great allure but still challenging to construct carbon-based metal free electrocatalysts to replace Pt-based ones for efficient ORR in both acidic and alkaline media. In this work, protein-hydrogel networks are demonstrated as a precursor for preparing N, S-codoped, hierarchically porous carbon (NSPC) materials with tunable pore structure. Of note, NSPC-0.2–900 possesses a high pore volume of 0.8552 cm³ g–1 and shows excellent ORR activity in both alkaline (Eonset, ∼0.93 vs. RHE) and acidic (Eoneset, ∼0.89 V vs RHE) media. As a cathode catalyst for zinc-air batteries, NSPC exhibits higher power density and better long-term durability than Pt/C does. The work presents a great potential of using protein as a precursor for preparing heteroatom doped, hierarchically porous carbon materials for energy and environmental applications.

Additive manufacturing (AM) has a substantial capability to produce superior and divergent properties of titanium alloys for biomedical implants, unlike the existing conventional technologies. This work investigated the mechanical properties and microstructure evolution of a β-type Ti-35Nb-2Ta-3Zr alloy prepared by selective laser sintering (SLS) process. The superelastic properties of the resultant specimen were characterized by cyclic loading-unloading tensile testing to evaluate the effect of SLS-process on the β-type Ti alloy specimen. The zigzag and V-shaped formation of {112} β twins, coexisting with stress-induced ω-formation, were observed by the transmission electron microscopy (TEM). The formation of Type I twin martensite along with β-structure is attributed to superelastic recovery and elastic recovery of SLS-produced specimen. High resolution TEM (HRTEM) observation was used to investigate the transition between β and ω phases. Thin layers of ω-formation in weak interfacial stress regions along with the longitudinal twin boundaries were also analyzed. The orientation relationship between ω-structure and parent β-phase involves an overlapping of ω-phase, observed along with longitudinal β-matrix and β-twins. Moreover, dislocation tangles and dislocation pile-ups form along with twin martensite, stress-induced ω-phase, and β-phase.

Hollow porous carbon (HPC) composed of inner voids and outer shells, is receiving increasing attention in constructing supported catalysts for diverse applications. It is of great allure but still ...

The Ga’erqiong-Galale ore district in the Bangong-Nujiang metallogenic belt hosts two large skarn Cu-Au deposits, associated with voluminous Late Cretaceous volcanism and plutonism. In this study, we present new zircon U-Pb age and Hf isotopic data, whole-rock geochemical and Sr-Nd-Pb isotopic data to constrain the geodynamic mechanisms of Late Cretaceous magmatism and metallogenesis within the Ga’erqiong-Galale ore district. The intrusive rocks in the district include diorite, quartz diorite, granodiorite, granite, and dioritic porphyrite. Zircon U-Pb analyses suggest those rocks mainly emplaced at early Late Cretaceous (84–95 Ma), which coincide with the ore-forming age of Ga’erqiong and Galale deposits (87 and 88 Ma, respectively). Geochemically, the granodiorite and dioritic porphyrite are metaluminous with depletion of Nb, Ta, and Ti and characterized by high MgO (2.68–4.16 wt%), Mg # (55–87), Sr (414–616 ppm), Sr/Y (30–56), Cr (22.5–105.0 ppm), and Ni (26.6–59.7 ppm), low Y (10.6–14.0 ppm), and heavy rare earth elements (HREE), showing geochemical signatures of high Mg# adakitic rocks. Four dioritic porphyrite and three granodiorite samples exhibit homologous Sr-Nd-Pb isotope compositions and have positive whole rock eNd(t) values of +1.1 to +3.0, (87Sr/86Sr)i values of 0.7045–0.7048, (206Pb/204Pb)t ratios of 18.42–18.73, (207Pb/204Pb)t ratios of 15.58–15.63, and (208Pb/204Pb)t ratios of 38.54–38.88. Besides, 116 analyses from the Ga’erqiong-Galale adakitic rocks give positive zircon eHf(t) values (+4.3 to +12.1). The geochemical and isotopic data indicate that the early Late Cretaceous high Mg# adakitic rocks in the Ga’erqiong-Galale ore district originated from the delaminated juvenile lower mafic crust that has been modified by the previously subduction. Partial melting of previously subduction modified, delaminated juvenile lower crust not only formed the hydrous adakitic melts in highly oxidation state, but also provided the ore metals (e.g., Cu, Au, Mo) for the subsequent mineralization. Such melts reacted with surrounding mantle peridotite to elevate the contents of metallic elements again. Then the melts experienced varying degrees of fractional crystallization process in the shallow magma chamber to form the dioritic and granodioritic magmas. Those magmas emplaced into the limestone of Duo’ai and Jiega Formation to form the Ga’erqiong and Galale skarn Cu-Au deposits, respectively. Field observation, together with geochronological and geochemical data, indicate that the early Late Cretaceous mineralization events in the northern and central Lhasa subterranes took place in a transition tectonic setting from collisional orogeny to extension and the input of subduction modified juvenile crust play an important role in the Late Cretaceous ore-forming events.

Titanium and its alloys are among the widely used materials in the biomedical field, but they have poor wear resistance and antibacterial properties. In the present study, anodization, photo-reduction, and spin-coating technologies were integrated to prepare a hybrid modified coating for bio-inert titanium implants, having excellent comprehensive performance. The surface roughness of Ti–35Nb–2Ta–3Zr was specifically optimized by surface modification leading to improved wear resistance. Ag ions are still detectable after 28 days of submersion in saline. The antibacterial rate of the composite coating group reaches 100% by plate counting due to the antibacterial mechanism of direct and indirect contact. Both bacteria morphology and fluorescence staining experiments confirm these results. Besides, no cytotoxicity was detected in our fabricated implants during the CCK-8 assay. Accordingly, fabrication of hybrid modified coatings on Ti–35Nb–2Ta–3Zr is an effective strategy for infection and cytotoxicity prevention. These hybrid modified coatings can be regarded as promising multifunctional biomaterials.

Design an implant similar to the human bone is one of the critical problems in bone tissue engineering. Metal porous scaffolds have good prospects in bone tissue replacement due to their matching elastic modulus, better strength, and biocompatibility. However, traditional processing methods are challenging to fabricate scaffolds with a porous structure, limiting the development of porous scaffolds. With the advancement of additive manufacturing (AM) and computer-aided technologies, the development of porous metal scaffolds also ushers in unprecedented opportunities. In recent years, many new metal materials and innovative design methods are used to fabricate porous scaffolds with excellent mechanical properties and biocompatibility. This article reviews the research progress of porous metal scaffolds, and introduces the AM technologies used in porous metal scaffolds. Then the applications of different metal materials in bone scaffolds are summarized, and the advantages and limitations of various scaffold design methods are discussed. Finally, we look forward to the development prospects of AM in porous metal scaffolds.

The main goal of this research is to scrutinize the effect of texture and grain size on the biological response of hierarchical structured pure titanium (Ti), examining the interrelation between grain refinement mechanisms with texture variation. The hierarchical structure was produced using two methods of severe plastic deformation (SPD). The Ti specimens were first processed up to six passes by equal channel angular pressing (ECAP) and subsequently treated at the top surface using surface mechanical attrition treatment (SMAT). Microstructure examination by Electron backscatter diffraction (EBSD) indicates that the SMAT-treated surface was categorized into three distinct microstructural regions based on the type of grain refinement process involved during SPD: twin induced dynamic recrystallization (TDRX) and geometric dynamic recrystallization (GDRX) in the topmost surface, and continuous (CDRX) and discontinuous dynamic recrystallization (DDRX) in the lower regions of the sample. The biological experiments showed meaningful improvement in the cellular response of SMATed and ECAPed samples. It was demonstrated that grain refinement could have the capability of improving the biological response of Ti surface. In this regard, SMATed + 2ECAPed sample showed the best result although it has not the smallest grain size and the highest texture intensity. It was observed that texture and grain orientation of planes have an important impact on the biological response of pure Ti and dominance of prismatic (1010) texture can improve the cell viability, adhesion and its differentiation. Therefore, microstructure and texture tailoring through combined SPD methods could be a promising strategy for the improvement of the next generation of medical implants.

Magnesium (Mg) and its alloys have attached more and more attention because of their potential as a new type of biodegradable metal materials. In this work, AZ31/ZrO2 nanocomposites with good uniformity were prepared successfully by friction stir processing (FSP). The scanning electron microscope (SEM) and transmission electron microscope (TEM) were used to characterize the microstructure of the composites. The mechanical properties, electrochemical corrosion properties and biological properties were evaluated. In addition, the effect of reinforced particles (ZrO2) on the microstructure and properties of the composite was studied comparing with FSP AZ31 Mg alloy. The results show that compared with the base metal (BM), the AZ31/ZrO2 composite material achieves homogenization, densification, and grain refinement after FSP. The combination of dynamic recrystallization and ZrO2 particles leads to grain refinement of Mg alloy, and the average grain size of AZ31/ZrO2 composites is 3.2 μm. After FSP, the c-axis of grain is deflected under the compression stress of shoulder and the shear stress of pin. The ultimate tensile strength (UTS) and yield strength (YS) of BM were 283 and 137 MPa, respectively, the UTS and YS of AZ31/ZrO2 composites were 427 and 217 MPa, respectively. The grain refinement and Orowan strengthening are the major strengthening mechanisms. Moreover, the corrosion resistance in simulated body fluid of Mg alloy is improved by grain refinement and the barrier effect of ZrO2.

Transfer learning has become the de facto practice to reuse a deep neural network (DNN) that is pre-trained with abundant training data in a source task to improve the model training on target tasks with smaller-scale training data. In this paper, we first investigate the correlation between the DNN’s pre-training performance in the source task and their transfer results in the downstream tasks. We find that high performance of a pre-trained model does not necessarily imply high transferability. We then propose a metric, named Frechet Pre-train Distance, to estimate the transferability of a deep neural network. By applying the proposed Frechet Pre-train Distance, we are able to identify the optimal pre-trained checkpoint, and then achieve high transferability on downstream tasks. Finally, we investigate several factors impacting DNN’s transferability including normalization, different networks and learning rates. The results consistently support our conclusions.

Functional gradient materials (FGMs), as a modern group of materials, can provide multiple functions and are able to well mimic the hierarchical and gradient structure of natural systems. Because biomedical implants usually substitute the bone tissues and bone is an organic, natural FGM material, it seems quite reasonable to use the FGM concept in these applications. These FGMs have numerous advantages, including the ability to tailor the desired mechanical and biological response by producing various gradations, such as composition, porosity, and size; mitigating some limitations, such as stress-shielding effects; improving osseointegration; and enhancing electrochemical behavior and wear resistance. Although these are beneficial aspects, there is still a notable lack of comprehensive guidelines and standards. This paper aims to comprehensively review the current scenery of FGM metallic materials in the biomedical field, specifically its dental and orthopedic applications. It also introduces various processing methods, especially additive manufacturing methods that have a substantial impact on FGM production, mentioning its prospects and how FGMs can change the direction of both industry and biomedicine. Any improvement in FGM knowledge and technology can lead to big steps toward its industrialization and most notably for much better implant designs with more biocompatibility and similarity to natural tissues that enhance the quality of life for human beings.

Ultraviolet is a vital spectrum in wireless optical communication. There are a lot of studies have been done in wireless ultraviolet communication. However, Ultraviolet communication networking mainly focuses on self-organizing networks and multi-hop networks. We proposed a new multi-access scheme of star topology network for confidential communications in semi-open spaces on the surface of ships. We verified the feasibility of the scheme through theoretical analysis and system simulation. It is proved that the scheme is suitable for ultraviolet communication and has better performance in a multi-user network. This scheme could be a sub-scheme for the next generation communication network combined with ultraviolet light communication.

With the continuous progress and development in the biomedicine field, metallic biomedical materials have attracted the considerable attention of researchers, but the related procedures need to be further developed. Since the traditional metal implant materials are not highly compatible with the human body, the modern materials with excellent mechanical properties and proper biocompatibility should be developed urgently in order to solve any adverse reactions caused by the long-term implantations. The advent of the high-entropy alloy (HEA) as an innovative and advanced idea emerged to develop the medical implant materials through the specific HEA designs. The properties of these HEA materials can be predicted and regulated. In this paper, the progression and application of titanium-based HEAs, as well as their preparation and biological evaluation methods, are comprehensively reviewed. Additionally, the prospects for the development and use of these alloys in implant applications are put forward.

The CoCrNi and precipitate-hardened (CoCrNi)94Ti3Al3 medium entropy alloys (MEAs) have attracted much attention, due to their exceptional mechanical properties, whereas the hot deformation characteristics have not been revealed. In the present study, we investigated the dynamic recrystallization behavior and microstructure evolutions of the two MEAs hot-compressed at single-phase temperatures. The constitutive equation was obtained, and microstructures were observed. Discontinuous dynamic recrystallization acted as a key mechanism of grain refinement at a relatively higher temperature and lower strain rate, which leads to the formation of a homogeneous grain structure. The addition of Ti and Al promoted dynamic recrystallization due to the solid solution hardening effect. The results provide valuable guidelines for microstructure refinement via thermomechanical processing.

Deep Neural Network (DNN) trained by the gradient descent method is known to be vulnerable to maliciously perturbed adversarial input, aka. adversarial attack. As one of the countermeasures against adversarial attacks, increasing the model capacity for DNN robustness enhancement was discussed and reported as an effective approach by many recent works. In this work, we show that shrinking the model size through proper weight pruning can even be helpful to improve the DNN robustness under adversarial attack. For obtaining a simultaneously robust and compact DNN model, we propose a multi-objective training method called Robust Sparse Regularization (RSR), through the fusion of various regularization techniques, including channel-wise noise injection, lasso weight penalty, and adversarial training. We conduct extensive experiments to show the effectiveness of RSR against popular white-box (i.e., PGD and FGSM) and black-box attacks. Thanks to RSR, 85 % weight connections of ResNet-18 can be pruned while still achieving 0.68 % and 8.72 % improvement in clean- and perturbed-data accuracy respectively on CIFAR-10 dataset, in comparison to its PGD adversarial training baseline.

Rapidly generated data and the amount magnitude of data analytical jobs pose great pressure to the underlying computing facilities. A distributed multi-cluster computing environment such as a hybrid cloud consequently raises its necessity due to its advantages in adapting geographically distributed and potentially cloud-based computing resources. Different clusters forming such an environment could be heterogeneous and may be resource-elastic as well. From analytical perspective, in accordance with increasing needs on streaming applications and timely analytical demands, many data analytical jobs nowadays are time-critical in terms of their temporal urgency. And the overall workload of the computing environment can be hybrid to contain both time-critical and general applications. These all call for an efficient resource management approach capable to apprehend both computing environment and application features. However, the added up complexity and high dynamics of the system greatly hinder the performance of traditional rule-based approaches. In this work, we propose to utilize deep reinforcement learning for developing elasticity-compatible resource management for a heterogeneous distributed computing environment, aiming for less occurrences of missing temporal deadline while maintaining low average execution time ratio. Along with reinforcement learning we design a deep model employing Long Short-Term Memory (LSTM) structure and partial model sharing for multi-target learning mechanism. The experimental results show that the proposed approach could greatly outperform baselines and serve as a robust resource management for variant workloads.

At present, by establishing complex mathematical model to simulate the ironmaking process in Blast Furnace (BF), the real-time and applicability are not strong in actual production. To solve this problem, an image reconstruction algorithm for BF Burden surface based on Compressive Sensing is proposed. Sparse matrix is constructed by Discrete Cosine Transform (DCT) and random Gaussian matrix is constructed to obtain part of the image information of BF Burden surface directly. Then the original image information is reconstructed by orthogonal matching pursuit algorithm. The final simulation results show that the image reconstruction algorithm proposed in this study works well.

We study how to train a student deep neural network for visual recognition by distilling knowledge from a blackbox teacher model in a data-efficient manner. Progress on this problem can significantly reduce the dependence on large-scale datasets for learning high-performing visual recognition models. There are two major challenges. One is that the number of queries into the teacher model should be minimized to save computational and/or financial costs. The other is that the number of images used for the knowledge distillation should be small; otherwise, it violates our expectation of reducing the dependence on large-scale datasets. To tackle these challenges, we propose an approach that blends mixup and active learning. The former effectively augments the few unlabeled images by a big pool of synthetic images sampled from the convex hull of the original images, and the latter actively chooses from the pool hard examples for the student neural network and query their labels from the teacher model. We validate our approach with extensive experiments., Accepted to CVPR 2020

The propose of this review was to summarize the advances in multi-scale surface technology of titanium implants to accelerate the osseointegration process. The several multi-scaled methods used for improving wettability, roughness, and bioactivity of implant surfaces are reviewed. In addition, macro-scale methods (e.g., 3D printing (3DP) and laser surface texturing (LST)), micro-scale (e.g., grit-blasting, acid-etching, and Sand-blasted, Large-grit, and Acid-etching (SLA)) and nano-scale methods (e.g., plasma-spraying and anodization) are also discussed, and these surfaces are known to have favorable properties in clinical applications. Functionalized coatings with organic and non-organic loadings suggest good prospects for the future of modern biotechnology. Nevertheless, because of high cost and low clinical validation, these partial coatings have not been commercially available so far. A large number of in vitro and in vivo investigations are necessary in order to obtain in-depth exploration about the efficiency of functional implant surfaces. The prospective titanium implants should possess the optimum chemistry, bionic characteristics, and standardized modern topographies to achieve rapid osseointegration.

Object frequency in the real world often follows a power law, leading to a mismatch between datasets with long-tailed class distributions seen by a machine learning model and our expectation of the model to perform well on all classes. We analyze this mismatch from a domain adaptation point of view. First of all, we connect existing class-balanced methods for long-tailed classification to target shift, a well-studied scenario in domain adaptation. The connection reveals that these methods implicitly assume that the training data and test data share the same class-conditioned distribution, which does not hold in general and especially for the tail classes. While a head class could contain abundant and diverse training examples that well represent the expected data at inference time, the tail classes are often short of representative training data. To this end, we propose to augment the classic class-balanced learning by explicitly estimating the differences between the class-conditioned distributions with a meta-learning approach. We validate our approach with six benchmark datasets and three loss functions., Comment: Accepted for publication at CVPR2020

Speaker diarization, which is to find the speech segments of specific speakers, has been widely used in human-centered applications such as video conferences or human-computer interaction systems. In this paper, we propose a self-supervised audio-video synchronization learning method to address the problem of speaker diarization without massive labeling effort. We improve the previous approaches by introducing two new loss functions: the dynamic triplet loss and the multinomial loss. We test them on a real-world human-computer interaction system and the results show our best model yields a remarkable gain of +8% F 1 -scores as well as diarization error rate reduction. Finally, we introduce a new large scale audio-video corpus designed to fill the vacancy of audio-video dataset in Chinese.

Dilated convolution kernels are constrained by their shared dilation, keeping them from being aware of diverse spatial contents at different locations. We address such limitations by formulating the dilation as trainable weights with respect to individual positions. We propose Adaptive Dilation Convolutional Neural Networks (ADCNN), a light-weighted extension that allows convolutional kernels to adjust their dilation value based on different contents at the pixel level. Unlike previous content-adaptive models, ADCNN dynamically infers pixel-wise dilation via modeling feed-forward inter-patterns, which provides a new perspective for developing adaptive network structures other than sampling kernel spaces. Our evaluation results indicate ADCNNs can be easily integrated into various backbone networks and consistently outperform their regular counterparts on various visual tasks.

Decreasing elastic modulus and simultaneously improving biological compatibility has been a major challenge in developing biomedical alloy. In this study, Ti6Al4V/Zn composite with surface nano crystallization is successfully fabricated by friction stir processing (FSP) with simultaneous addition of beneficial Zn element. Nanoscale microstructures in surface region include nanoscaled Zn rich particles (including TiZn2), α nanograins, nanotwins and nano α + β lamellae. Dislocations slip and grain boundary sliding contribute much to the transformation of α-Ti nanograin, while dynamic overlapping of two partial dislocations is the main deformation mechanism for nanotwins. Interestingly, these nanoscale microstructures reduce the elastic modulus of this region.

A simulation model based on the finite-element method (FEM) and transfer matrix method (TMM) is developed for analyzing mechanical long-period fiber gratings (MLPFGs). The anisotropic characteristics of the cross section of a fiber are obtained using FEM and the non-rectangular distribution of refractive index of the stressed area is accounted for to analyze band-rejection filtering characteristics using TMM, both of which are neglected in the conventional coupled-mode method (CMM). The overall attenuation in the entire spectral range is successfully simulated by the proposed method. The deviations of the transmittance at resonance wavelength between theoretical values and the actual conditions are 11.39 dB and 1.18 dB for CMM and FEM-TMM, respectively, for a high-birefringence case under a high applied force such as F = 60 N.

Friction stir processing (FSP) was successfully employed to produce gradient multi-phase structured Ti–6Al–4V/Ag nanocomposite surface layer. Microstructural evolution of amorphous/nanocrystalline/nano-twin in the surface layer was characterized by TEM in detail. The formation of amorphous phase in the upper surface of stir zone (SZ) was attributed to mechanically driven amorphization caused by dislocation gliding plastic deformation and elements mixing process. The dislocation activities played a dominant role in causing the distortion of lattices and nanocrystallization. The plastic deformation via FSP was responsible for considerable dislocations pile-up and tangle in the potential deformation bands.

Corrosion fatigue is one of the major forms for nickel‑aluminum bronze alloy used in marine propeller. It is greatly important to understand the corrosion fatigue behavior of nickel‑aluminum bronze alloy in seawater for the purpose of improving the service life of marine equipment. In this study, corrosion fatigue crack growth rate of nickel‑aluminum bronze alloy in 3.5% NaCl solution was tested under various stress intensity factors, stress ratios and loading frequencies to reveal its corrosion fatigue mechanism. The results show that corrosion environment significantly accelerates the crack growth and decreases the corrosion fatigue threshold, which is attributed to the corrosion reactions between the metals of crack tip and corrosive medium. Among the various factors, loading frequency has more predominant influence on the corrosion fatigue crack growth rate of nickel‑aluminum bronze alloy, which can be rationally explained by the hydrogen embrittlement mechanism. Additionally, in the process of corrosion fatigue crack growth, the acidification and the dealumination would occur at the crack tip, resulting in the corrosion dissolution of κ phases and the deterioration in mechanical properties of crack tip.

In this study, coarse-grained Nickel-Aluminum bronze (NAB) alloy is subjected to shot-peening (SP) treatment, and the grain refinement mechanisms of various phases in shot-peened NAB alloy are investigated separately. The results show that severe deformation layers are produced on the surface. Meanwhile, refined grains with the average size below 100 nm are induced through SP treatment indicated by the results of X-ray diffraction (XRD) patterns. Various phases in NAB alloy including α, β' and three intermetallic phases (κii, κiii and κiv) display different grain refinement mechanisms. The grain refinement in the α phase involves dislocation segmentation in the precipitation-free zone and the twinning intersection in the precipitation zone. The formation of nanotwins by emitting partial dislocations at the grain boundary is found to be responsible for the grain refinement in the β' phase. As for the κ phases, the strip-shaped κiii phases are fragmented after severe deformation, and the κiv and κii phases exhibit the twinning and dislocation deformation modes, respectively. Also, the microhardness is significantly enhanced after SP treatment, which is mainly attributed to the work hardening effect and grain refinement.

This work addresses the susceptibility to stress corrosion cracking in 3.5% NaCl solution (simulating seawater) of three differently obtained microstructures of a nickel-aluminum bronze alloy using slow strain rate tensile tests. The results showed that the stress corrosion cracking susceptibility of annealed and normalized alloys increased with the decrease of the strain rate as the metal became more severely corroded and dissolved at the crack tips. The continuous or semi-continuous α + κⅢ eutectoid structure is sensitive to stress corrosion cracking, resulting in the highest susceptibility of annealed alloy. The thin and overlapped Widmanstatten α grain hinders extension of the cracks and thus decreases the stress corrosion cracking susceptibility of normalized alloy to some degree. The quenched/aged alloy with homogeneous microstructure without α + κⅢ eutectoid structure and the β’ phase exhibits the lowest susceptibility. The stress corrosion cracking process of nickel-aluminum bronze alloys was revealed and is properly explained by the oxide film rupture and anodic dissolution mechanism as well as the hydrogen induced cracking mechanism.

The Lietinggang deposit is a representative skarn iron polymetallic deposit in the Nyainqentanglha skarn polymetallic zone (NSP) in Lhasa Terrane, Tibet. The mineralization is hosted in skarn of the Early and Middle Triassic Chaqupu Formation. Plutons in the deposit include gabbro, diorites (divided into the D1 and D2 diorites), granodiorite, syenogranite and porphyritic granite, and the D2 diorite and syenogranite are the mineralization associated plutons. In this study, we present LA-ICP-MS zircon U-Pb dating, zircon Hf isotopic and whole rock geochemical data of the diorites and porphyritic granite and S-Pb isotopic compositions of the ore sulfides in the Lietinggang deposit. Zircon U-Pb dating results for these intrusions yield Paleocene ages of 62.2–59.0 Ma. All the intermediate and felsic intrusions are I-type, and characterized by arc magma geochemical features, enriched in LREE and LILE, well developed negative Eu anomalies and depletion in HFSE, with most having positive eHf(t) values. These geochemical characteristics are similar with the coeval intrusions of the Gangdese batholith and Linzizong Group volcanic rocks, indicating they share a similar magma source. This magma source is proposed to be partial melting of the young metasomatized asthenospheric mantle wedge-derived basaltic rocks. During the magma evolution the fractionated crystallization has also play an important role. δ34S values for sulfides have features of magma source. The 206Pb/204Pb (18.043–18.711), 207Pb/204Pb (15.520–15.776), and 208Pb/204Pb (38.288–39.215) values of sulfides show that ore-forming metals were derived from mixing sources of mantle with upper crust materials. Geochemical characteristics of the intermediate and felsic intrusions, combined with the dating results, indicate that the Lietinggang magmatism and mineralization were formed within the India-Asia continental-collision tectonic setting.