A series of amorphous Mo-S-Se (a-Mo-S-Se) thin films with different S/Se atomic ratios were deposited on carbon paper (CP) via ratio frequency (RF)-magnetron co-sputtering. The electrocatalytic hydrogen evolution reaction (HER) properties of a-Mo-S-Se thin films were investigated. The overpotential and Tafel slope during the HER process initially decreased and then increased with decreasing S/Se ratio. The a-Mo-S-Se thin film with S/Se = 1:0.7 exhibited the lowest overpotential at 10 mA.cm- 2 (eta 10) of 230 mV and a Tafel slope of 90 mV dec- 1, demonstrating superior HER activity and kinetics. This was attributed to the optimization of Delta GH by adjusting S/ Se atomic ratio and a large number of active sites originating from the unsaturated S2-and Se2- , terminal S22and Se22- ligands on the film surface. In addition, the a-Mo-S-Se thin films show excellent electrocatalytic stability, which is ascribed to the good adhesion between the thin films and CP substrates obtained via the magnetron sputtering method. This work proposes a synergistic strategy combining element tuning, amorphous strategy, and thin film morphology achieved by a simple magnetron co-sputtering method to improve HER activity and stability of molybdenum sulfide.

Most recent research on pavement damage segmentation focuses on fully supervised learning to achieve solid performance. The major limitation of this strategy in practice is the heavy reliance on a large number of pixel-level labels, which are labour-intensive and time-consuming to acquire, especially for curvilinear cracks with fine details and inlaid patches with weak edges. This paper proposes a novel methodology effectively exploit unlabelled data supported by limited label knowledge, thus reducing the heavy annotation burden. The proposed methodology tackles three critical challenges in the pavement damage segmentation based on a semi-supervised paradigm. The first challenge is the noisy pseudo-labels triggered by severely imbalanced prior distributions, which can result in the underutilization of unlabelled samples. To tackle challenge, we propose a labelled prior-support uniform rectification module to explicitly improve the prediction quality of pseudo-labels by rectifying their corresponding semantics conditioned on the class-balanced labelled tokens. For the second challenge of the inter-class indistinction, an intrinsic cross-dimensional constraint mechanism is devised to enforce difficult representations to be aligned with the category distribution the weak vertical constraint within intra-set hierarchical features and the strong parallel cross-set constraint, enabling the model to identify weak texture defects. Finally, an adaptive dual regularization module proposed to handle the challenge of irregular and variable defect morphologies, achieved by regularizing the deep descriptor based on the intrinsic and external correlations across various granularity attributes. Extensive experiments on three datasets with the aforementioned challenges demonstrate the superiority of the proposed methodology over other advanced approaches. Our code and datasets will be available https://github.com/Yayan1128/LLPDS.git.

Graphene with high-density defects was generated by one-step gaseous detonation method, designated as DG. The DG exhibits a highly hydrogen evolution reaction (HER) performance in terms of its lower onset potential of 223 mV and smaller Tafel slope of 120 mV·dec−1, which are superior to those of commercial counterpart (CG). Furthermore, the value of DG acted as an excellent support of active materials was verified by successfully hybridizing the MoS2 with DG and the improvement of DG on the HER performance of MoS2. This work provides a simple, economical, energy-saving and high yield to prepared DG with highly HER electrocatalytic activity and advantages as an excellent support material, making it of great value in practical applications.

The phase separation of proteins is considered to be promoted by driver regions. However, emerging evidence indicates that disordered linkers connecting driver regions in multidomain proteins may regulate phase separation. In this study, we use a protein-DNA system to investigate how disordered linkers affect protein phase separation behavior. We report that the residue-residue interactions within disordered linkers modulate the overall conformational ensemble of the protein, thereby influencing the threshold concentration for phase separation and physical properties of the resulting condensates. We demonstrate that disordered linkers not only serve a connecting role but also play a previously underexplored regulatory role in protein phase separation through their sequence diversity. This addresses a gap in the understanding of linker-mediated regulation of phase separation and advances the “stickers-and-spacers” theory. Furthermore, we show that a molecular grammar governs condensate properties, providing valuable insights into the relationship between disordered linker sequences and condensate functions.

Semi-supervised 3D medical image segmentation aims to achieve accurate segmentation using few labelled data and numerous unlabelled data. The main challenge in the design of semi-supervised learning methods consists in the effective use of the unlabelled data for training. A promising solution consists of ensuring consistent predictions across different views of the data, where the efficacy of this strategy depends on the accuracy of the pseudo-labels generated by the model for this consistency learning strategy. In this paper, we introduce a new methodology to produce high-quality pseudo-labels for a consistency learning strategy to address semi-supervised 3D medical image segmentation. The methodology has three important contributions. The first contribution is the Cooperative Rectification Learning Network (CRLN) that learns multiple prototypes per class to be used as external knowledge priors to adaptively rectify pseudo-labels at the voxel level. The second contribution consists of the Dynamic Interaction Module (DIM) to facilitate pairwise and cross-class interactions between prototypes and multi-resolution image features, enabling the production of accurate voxel-level clues for pseudo-label rectification. The third contribution is the Cooperative Positive Supervision (CPS), which optimises uncertain representations to align with unassertive representations of their class distributions, improving the model’s accuracy in classifying uncertain regions. Extensive experiments on three public 3D medical segmentation datasets demonstrate the effectiveness and superiority of our semi-supervised learning method.

Nondestructive testing of welding images is still a significant challenge due to the imaging characteristics of radiographic images and the extremely random distribution of welding defect types. The practical application of supervised methods for welding nondestructive testing encounters significant challenges due to the limited availability of densely annotated samples and the absence of prior information regarding unknown defects. Hence, this article first proposes a rapid screening method using only defect-free image training to screen suspected defect images and normal images of welds in real time. Specifically, we first combine the correspondence mechanism and the representation mechanism, aiming to: 1) alleviate the smoothing reconstruction behavior caused by small defects and weak texture defects in weld; and 2) mitigate the offset learning behavior resulting from the differences between natural images and industrial weld images in the normalizing flow method. We propose a memory-aware transformer-based encoder, thus improving representations of complex defect-free images. Moreover, a dual-decoder strategy is introduced, which remaps the latent dependencies generated by the encoder through a semantic correspondence mechanism and reconstruction-guided normalizing flow, enabling effective learning of knowledge from weld images. We apply this framework to an industrial case of weld images, the experimental results demonstrate that our method outperforms other existing approaches.

Developing hydrogen energy is an important direction in the future. Industrialized scale electrolysis water for hydrogen production requires the use of low-cost hydrogen evolution electrocatalyst materials to reduce its overpotential. Graphene has shown broad application prospects in hydrogen evolution electrocatalyst materials due to its large specific surface area, excellent conductivity, good stability, adjustable electronic structure, and modification of structure and surface state. This article provides a detailed analysis of the mechanism of graphene application in hydrogen evolution electrocatalysis. Based on different mechanisms, graphene-based hydrogen evolution electrocatalyst materials were classified and their latest research progress was reviewed. Finally, future development direction of graphene-based hydrogen evolution electrocatalytic materials was prospected.

Tiny cracks are often overlooked in the inspection process, causing huge economic losses and dangerous accidents. Therefore, tiny cracks should be detected in a timely and accurate manner to eliminate the disease at the initial stage. Inspired by the fact that humans are more likely to capture conspicuous information when observing objects, we propose a novel three-stage Extraction-Amplification-Fusion network (EAFNet). Specifically, in the extraction stage, we utilize an effective backbone network for feature extraction of tiny cracks. In the amplification stage, we design a Tiny Feature Amplification (TFA) module to amplify the extracted features. In the fusion stage, we propose a Two-Branch Fusion (TBF) module to fully fuse the feature maps at different resolutions. To make tiny crack information more ‘conspicuous’, we propose an activation function TinyReLU to enhance the contrast of the tiny cracks with the background. In addition, we construct a Tiny Crack (T-CRACK) dataset with six different backgrounds and a Cross-scale Crack (C-CRACK) dataset. On both datasets, EAFNet achieves an advantage over the existing 8 advanced networks. The two datasets are available at: https://github.com/EAFNet/EAFNet.

Deep neural network has demonstrated high-level accuracy in rail surface defect segmentation. However, deploying these deep models in actual inspection situations results in generalizability deficits and accuracy degradation. This phenomenon is mainly caused by the appearance difference between training and test images. To alleviate this issue, we propose a feature-based domain disentanglement and randomization (FDDR) framework to improve the generalization of deep models in unseen datasets. Specifically, two encoders are introduced to decompose the defect image into domain-invariant structural features and domain-specific style features. Only the domain invariant features are used to identify the defects. Additionally, we design a shuffle whitening module to remove the style information from the domain-invariant features. Meanwhile, the extracted style features are used to train a style variational autoencoder to randomly generate novel defect styles. Then, the randomly generated style features are combined with the domain-invariant features to obtain new defect images, thus expanding the training sample. We validate the proposed FDDR framework in six defect segmentation datasets. Extensive experimental results show that FDDR demonstrates robust defect segmentation performance in unseen scenarios and outperforms other state-of-the-art domain generalization methods. The source code will be released at https://github.com/Rail-det/FDDR.

Rail surface defect detection is an essential part of railroad maintenance to prevent safety accidents. Recently, deep learning-based defect detection algorithms have shown impressive detection performance. However, the performance improvement from deep neural networks is based on a strong assumption that the training images have the same data distribution as the test images. This requirement is unrealistic in practical railroad surface defect detection. The appearance of rails varies significantly in different railroad sections due to changes in working conditions and maintenance status. In this paper, we propose a novel uncertainty-inspired unsupervised domain adaptation framework (UIUDA) to improve the generalization of deep learning models on data with distribution differences. Specifically, domain adversarial learning is used to align the feature distributions of training and test data. To mitigate the negative transfer caused by global alignment, we design a locally aligned domain adaptation strategy. In addition, self-training is introduced to refine the decision boundaries of the model in the test data. In this process, we propose an uncertainty-inspired evaluation method to remove noisy pseudo labels. We conducted domain adaptation experiments with RSDD and NEU-RSDD as the source domain and FRSD dataset as the target domain. The results show that UIUDA can effectively improve the generalization ability of the model, and the mIoU metric of defect segmentation is 80.17%.

Lead and its compounds can have cumulative harmful effects on the nervous, cardiovascular, and other systems, and especially affect the brain development of children. We collected 4918 samples from 15 food categories in 11 districts of Guangzhou, China, from 2017 to 2022, to investigate the extent of lead contamination in commercial foods and assess the health risk from dietary lead intake of the residents. Lead was measured in the samples using inductively coupled plasma mass spectrometry. Dietary exposure to lead was calculated based on the food consumption survey of Guangzhou residents in 2011, and the health risk of the population was evaluated using the margin of exposure (MOE) method. Lead was detected in 76.5% of the overall samples, with an average lead content of 29.4 mu g kg-1. The highest lead level was found in bivalves. The mean daily dietary lead intakes were as follows: 0.44, 0.34, 0.25, and 0.28 mu g kg-1 body weight (bw) day-1 for groups aged 3-6, 7-17, 18-59, and >= 60 years, respectively. Rice and rice products, leafy vegetables, and wheat flour and wheat products were identified as the primary sources of dietary lead exposure, accounting for 73.1%. The MOE values demonstrated the following tendency: younger age groups had lower MOEs, and 95% confidence ranges for the groups aged 3-6 and 7-17 began at 0.6 and 0.7, respectively, indicating the potential health risk of children, while those for other age groups were all above 1.0. Continued efforts are needed to reduce dietary lead exposure in Guangzhou.

In this paper, we report a special [Formula: see text] nanoflower synthesized by hydrothermal approach, in which a heterostructure petal of [Formula: see text] [Formula: see text] is constructed by growing the [Formula: see text] active layers onto the few-layered MXene [Formula: see text] substrate. Compared with the pristine [Formula: see text] nanoflowers, the heterostructure [Formula: see text] nanoflowers exhibit a lower onset overpotential of 207 mV and a lower Tafel slope of 28 mV dec[Formula: see text]. This can be explained by the heterostructure [Formula: see text] nanoflowers providing a great number of exposed surfaces of [Formula: see text], and increasing more active sites, and strengthening the interlayer coupling between [Formula: see text] and MXene [Formula: see text] to accelerate the charge transfer. All the merits are beneficial to the performance enhancement for hydrogen evolution catalytic.

Vision-based surface defect detection (SDD) for no-service aero-engine blades provides a fast and effective way to monitor product quality. Most existing detection algorithms for aero-engine blades are 1) based on CNN, including artificially designed non-maximum suppression (NMS) operations, and 2) focus on improving the detection accuracy rather than improving the inference speed and even ignoring the latter. To solve the above problems, we introduce a novel object detection paradigm, DEtection TRansformer (DETR), to design a novel network (SDD-DETR) with high accuracy for the SDD of aero-engine blades. To our knowledge, the paper is the first to introduce the DETR detector to SDD of aero-engine blades. While providing high accuracy, the inference speed of DETR remained slow due to self-attention operation and feed-forward network (FFN). Therefore, two lightweight modules have been designed for SDD of aero-engine blades: a progressive feature input multi-scale deformable attention module (PFI-MSDA) and a lightweight FFN (LW-FFN). PFI-MSDA hierarchically reduces the number of tokens input to the self-attention module, thereby reducing the time complexity of the self-attention layer. LW-FFN shrinks the complexity of multilayer perceptron. In addition, no parameter sharing of the detection head is utilized to compensate for the accuracy drop caused by the lightweight. Experiments verify that our method has the same AP and F1-score as DINO (a DETR-based detector), but our approach is lighter. Compared with DINO, the FLOPs are reduced by 113.4 ${G}$ , the inference speed is increased by 42.4%, and the runtime memory usage is reduced by 5.9 ${G}$ , which allows our method to be trained on low-end GPUs with more batch size, further improving the training efficiency. The code is available at https://github.com/VDT-2048/SDD-DETR. Note to Practitioners —The motivation for this paper is to design a high-precision and high-inference speed visual detection method for the SDD of aero-engine blades. Most high-precision vision methods are based on the transformer framework. However, its high complexity and poor compatibility in deployment environments lead to slower detection speeds. Although the application object in this paper is aero-engine blades, it is also applicable in other fields of industry, such as rail detection, plate and strip steel detection, etc. However, the method proposed cannot be supported by deployment environments such as RKNN because of the deformable attention operator, so it takes a certain amount of time to be deployed and put into practical use. Currently, the frameworks of visual and language large models are based on transformers, consistent with the framework of our method, which makes extending our approach to large visual and multi-modal models more accessible.

Accurate classification of surface defects is one of the most important factors in achieving quality inspection for strip steel. Most existing methods are based on fully-supervised learning, which requires a large number of labeled training data samples. In the manufacturing process, collecting defective samples is time-consuming and laborious. So it is very difficult to train a fully supervised model based on few labeled samples. In this paper, we propose a feature-aware network (FaNet) for a few shot defect classification, which can effectively distinguish new classes with a small number of labeled samples. In our proposed FaNet, we use ResNet12 as our baseline. The feature-attention convolution module (FAC) is applied to extract the comprehensive feature information from the base classes, as well as to fuse semantic information by capturing the long-range feature relationships between the upper and lower layers. Meanwhile, during the test phase, an online feature-enhance integration module (FEI) is adopted to average the noise from the support set and query set defect images, further enhancing image features among the different tasks. In addition, we construct a large-scale strip steel surface defects few shot classification dataset (FSC-20) with 20 different types. Experimental results show that the proposed method achieves the best performance compared to state-of-the-art methods for the 5-way 1-shot and 5-way 5-shot tasks. The dataset and code are available at: https://github.com/VDT-2048/FSC-20.

The giant panda is a unique vulnerable mammal in western China, and its main cause of death is digestive system diseases regardless of whether these animals are in the wild or in captivity. The relationship between the intestinal flora and the host exerts a significant impact on the nutrition and health of the giant pandas.

The solid-liquid composite lubrication system is a key technology for achieving high-fuel-economy and high-durability engine systems. Diamond-like carbon (DLC) films have high hardness, a low friction coefficient, and good biocompatibility, thereby being widely applied in engine system, involving valve lifters, piston rings, etc. It is well known that the performance of oil lubricants is highly correlated with the chemical composition of the grinding surface materials. Previous studies have shown that Fe2+ from ferrous materials directly takes part in the tribochemical reactions of oil lubricants. DLC films exhibit chemical inertness, in contrast to traditional ferrous materials. However, the current formulas for engine oil lubricants are universally developed for ferrous materials, and a special lubricant formula for DLC films needs urgent improvement. The extreme-pressure antiwear agent zinc dialkyldithiophosphate (ZDDP), the detergent calcium persulfonate (OBCaSu), and the dispersant polyisobutylene succinimide (PIBSI) are the most widely used additives in formulated lubricants. However, there is still limited research on the tribological properties of DLC films lubricated with the above additives. In this work, amorphous carbon (a-C) films were prepared via nonequilibrium magnetron sputtering, and the tribological properties of the a-C films under boundary lubrication conditions between ZDDP and OBCaSu (ZDDP+OBCaSu) and between ZDDP and PIBSI (ZDDP+PIBSI) were evaluated using a CSM tribometer. The tribochemical reactions were analyzed using Raman spectroscopy, scanning electron microscopy, and energy dispersive spectroscopy (EDS), combined with the full formula (FF) oil and GCr15 steel, to explore the tribological mechanism of a-C films. The morphologies of worn surfaces were determined with a three-dimensional surface profilometer, and it can be seen that the main wear mechanism of a-C films against steel balls is abrasive wear. Under various lubrication conditions, the results indicate that the worn surfaces of the a-C films undergo graphitization compared to the unworn surfaces, which is beneficial for achieving a low friction coefficient. Phosphate tribofilms are formed on the worn surfaces of GCr15 steel and a-C films under ZDDP, ZDDP+OBCaSu, and ZDDP+PIBSI lubrication conditions. Under a ZDDP+OBCaSu lubrication condition, the composite tribofilms of Ca-3(PO4)(2) and Zn-3(PO4)(2 )on the worn surfaces of GCr15 steel and a-C films can improve their wear resistance and lubrication performance, which was confirmed by the low friction coefficients and wear rates. Moreover, similar friction coefficients and wear rates on the surfaces of GCr15 steel and a-C films are obtained. Therefore, the tribochemical reaction of the ZDDP+OBCaSu lubrication has less dependence on the surface materials. Under ZDDP+PIBSI lubrication, the tribological properties of GCr15 steel and a-C films decrease, resulting in high friction coefficients and high wear rates. According to the EDS energy spectrum, it can be derived that the strong dispersion of PIBSI is not conducive to the formation of phosphate tribofilms. Under FF lubrication conditions, the tribological properties of GCr15 steel and a-C films decreased compared to those under ZDDP, ZDDP+OBCaSu, and ZDDP+PIBSI lubrication conditions, which may be influenced by the lubricant concentrations and other lubricants. In this work, the tribological behavior and tribochemical reaction mechanism of traditional oil lubricants on the surfaces of GCr15 steel and a-C films are studied. Overall, the results show that the changes in the friction coefficients and wear rates of GCr15 steel and a-C films with various lubricants are similar. However, the a-C films are not sensitive to the lubricants. That is, compared to GCr15 steel, the tribological performance of the a-C films under different lubricants fluctuates by only a minor degree. Therefore, there is a need to develop more suitable oil lubricant formulas for a-C films. This work serves as a guidance for future design and development of lubricants and a-C films for energy savings and fuel efficiency.

Deep neural networks have greatly improved the performance of rail surface defect segmentation when the test samples have the same distribution as the training samples. However, in practical inspection scenarios, the rail surface exhibits variations in appearance due to different service times and natural conditions. Conventional deep learning models show limited generalization in scenes with distribution differences. To address this problem, we propose a novel one-shot unsupervised domain adaptation framework. Specifically, we introduce a shape-consistent style transfer module that performs pixel-level distribution alignment between the training and test images. Based on the one-shot test image, the training image is reconstructed to have the same appearance as the test image. Meanwhile, we employ a multitask learning strategy to prevent content distortion of the reconstructed images. To improve the robustness of the model to distribution differences, we design an edge-aware defect segmentation model and train the model using the reconstructed training images. The experimental results show that our method effectively improves the robustness of the model to distribution differences and achieves satisfying results in the task of rail surface defect segmentation.

Objective To investigate the contamination level of aflatoxin B 1 in loose-packed peanut oil extracted by a native method in individual workshops in Guangzhou from 2016 to 2021,and to evaluate the corresponding health risk and liver cancer risk of loose-packed peanut oil consumers.Methods In 2016-2021,a total of 203 loose-packed peanut oil extracted by a native method in individual workshops were collected in Guangzhou. Aflatoxin B 1 concentrations were examined by HPLC-fluorometry and HPLC-post-chromatographic derivatization photometry,and evaluated according to the National Food Safety Standard-Limit of Mycotoxins in Food（GB 2761）. Based on the probability assessment theory,methods of margin of exposure（MOE）and mathematical model were used to evaluate the health risk and liver cancer risk of loose-packed peanut oil consumers.Results The detection rate and over-standard rate of aflatoxin B 1 in 203 loose-packed peanut oil were 76.85%（156/203）and 9.85%（20/203）,respectively. The mean and maximum detection values in 2019-2021 were lower than those in 2016-2018. The results of the health risk assessment showed that the MOE values of peanut oil consumers of all ages were bellowed 10 000,and the risk of liver cancer caused by aflatoxin B 1 in high consumption groups was greater than 1 cancer/（1 million people/year）.Conclusions The contamination level of aflatoxin B 1 in loosepacked peanut oil extracted by a native method in individual workshops in Guangzhou has been improved. While the health risk of peanut oil consumers exposed to aflatoxin B 1 in peanut oil still deserves attention. It is suggested to further strengthen the supervision of individual workshops on peanut oil.