In computer vision, fine-grained classification has become an important issue in recognizing objects with slight visual differences. Usually, it is challenging to generate good performance when solving fine-grained classification problems using traditional convolutional neural networks. To improve the accuracy and training time of convolutional neural networks in solving fine-grained classification problems, this paper proposes a tree-structured framework by eliminating the effect of differences between clusters. The contributions of the proposed method include the following three aspects: (1) a self-supervised method that automatically creates a classification tree, eliminating the need for manual labeling; (2) a machine-learning matcher which determines the cluster to which an item belongs, minimizing the impact of inter-cluster variations on classification; and (3) a pruning criterion which filters the tree-structured classifier, retaining only the models with superior classification performance. The experimental evaluation of the proposed tree-structured framework demonstrates its effectiveness in reducing training time and improving the accuracy of fine-grained classification across various datasets in comparison with conventional convolutional neural network models. Specifically, for the CUB 200 2011, FGVC aircraft, and Stanford car datasets, the proposed method achieves a reduction in training time of 32.91%, 35.87%, and 14.48%, and improves the accuracy of fine-grained classification by 1.17%, 2.01%, and 0.59%, respectively. Full article

The precision and consistency of the microdroplet jetting procedure are crucial for the casting sand mold’s performance during binder injection. The generation and jetting of microdroplets in piezoelectric printheads were examined in this study in relation to changes in specific jetting process parameters. Using finite element analysis and a simplified physical model of a microdroplet jetting device, an electromechanically coupled model of a microdroplet jetting device was created in order to study the characteristics of microdroplet jetting. A volume-of-fluid model was also created in order to study the microdroplet jetting process and perform repeatability tests. The effects of altering nozzle radius, actuation pulse width, intake velocity, and fluid viscosity on microdroplet jetting properties were then investigated using the models. We were able to control the development of satellite droplets thanks to the knowledge we gained about how each process parameter affected droplet status. This study demonstrates how the radius of the nozzle and the pulse width of the piezoelectric actuation signal have a significant impact on the jetting properties of piezoelectric printheads and the production of microdroplets. The quantitative correlations between process factors and jetting characteristics can be used to optimize microdroplet production and reduce droplet size. Finally, this study will help create control systems for microdroplet jetting operations and enhance the precision of 3D printed casting sand molds. Full article

As precious cultural heritage sites, the state of preservation of cave temples is closely related to the geological and climatic conditions in which they are located. This paper constructed an analytical method of sized slate stability based on the equivalent support stiffness method. The stability analysis of the roof slab of Yuanjue Cave was carried out by establishing a three-dimensional numerical calculation model. Through comparative analysis of the results of stress and displacement fields under different conditions, the stress and deformation characteristics of the roof slab of Yuanjue Cave were revealed, as well as the study of the main factors affecting the stability of the roof slab of Yuanjue Cave and the key slate to be monitored. The main research results are as follows. The stress deformation of the roof plate of Yuanjue cave is mainly divided into the initial uniform change stage, the medium-term stable change stage or the medium-term accelerated change stage, and the later rapid change stage. With the increase in the number of overhanging and broken slates and the increase in the damage factor of cracked slates, the vertical stress extremum of the stones increases continuously, and the equivalent support stiffness decreases, which aggravates the uneven stress deformation of the roof of the Yuanjue Cave. This study provides a reliable reference basis for the stability analysis and evaluation of the roof slab of a large number of cave temples existing in the Sichuan and Chongqing areas in China. Full article

In order to make up for the shortcomings of existing theory calculation methods for circular diaphragm walls, an alternative calculation method was developed with a clear concept based on the circular cylindrical shell theory and superposition principle in elasticity and named the circular cylindrical shell theory calculation method (CCSTCM). We took the north anchorage circular foundation pit of G3 Tongling Yangtze River Highway and Railway Bridge as an example and calculated and analyzed it by the proposed CCSTCM, finite element numerical simulation method (FENSM) and site monitoring. As a result, we obtained the radial displacement, circumferential stress and vertical bending moment of the circular diaphragm wall, and then summarized their regularities. By researching the results of the CCSTCM (TCR), the results of the FENSM (NSR) and the results of the site monitoring (SMR), the following conclusions could be drawn: the numerical calculation model established was reasonable; the variation trends of the data curves of the TCR were highly similar to those of the NSR and SMR; and the TCR were slightly larger than the NSR but slightly smaller than the SMR, and closer to the SMR in general. Finally, the proposed CCSTCM was proven to be correct and applicable and could be used in similar circular diaphragm wall projects. Full article

This article reports on the comparative study of the hot deformation behaviour of virgin (steel A) and rejuvenated heat treatment creep-exhausted (steel B) P91 steels. Hot uniaxial compression tests were conducted on the two steels at a deformation temperature range of 900–1050 °C and a strain rate range of 0.01–10 s⁻¹ to a total strain of 0.6 using Gleeble^® 3500 equipment. The results showed that the flow stress largely depends on the deformation conditions. The flow stress for the two steels increased with an increase in strain rate at a given deformation temperature and vice versa. The flow stress–strain curves exhibited dynamic recovery as the softening mechanism. The material constants determined using Arrhenius constitutive equations were: the stress exponent, which was 5.76 for steel A and 6.67 for steel B; and the apparent activation energy, which was: 473.1 kJ mol⁻¹ for steel A and 564.5 kJmol⁻¹ for steel B. From these results, steel A exhibited better workability than steel B. Statistical parameters analyses showed that the flow stress for the two steels had a good correlation between the experimental and predicted data. Pearson’s correlation coefficient (R) was 0.97 for steel A and 0.98 for steel B. The average absolute relative error (AARE) values were 7.62% for steel A and 6.54% for steel B. This study shows that the Arrhenius equations can effectively describe the flow stress behaviour of P91 steel, and this method is applicable for industrial metalworking process. Full article

Within the presented research study we want to estimate the State of Health (SOH) of a fleet of electric vehicles solely using field data. This information may not only help operators during mission planning, but it can reveal causes of accelerated aging. For this purpose, we use a customized neural network that is able to process the data of all fleet vehicles simultaneously. Thus, information between batteries of the different vehicles is transferred and the extrapolation properties are enhanced. We firstly show results with data gathered from a fleet of 25 electric buses. A prediction accuracy of below 5 mV could be obtained for most validation sections. Furthermore, a proof-of-concept experiment illustrates the advantages of the fleet learning approach. Full article

The goal of relational triple extraction is to extract knowledge-rich relational triples from unstructured text. Although the previous methods obtain considerable performance, there are still some problems, such as error propagation, the overlapping triple problem, and suboptimal subject–object alignment. To address the shortcomings above, in this paper, we decompose this task into three subtasks from a fresh perspective: entity extraction, subject–object alignment and relation judgement, as well as propose a novel bi-directional translating decoding model (BTDM). Specifically, a bidirectional translating decoding structure is designed to perform entity extraction and subject–object alignment, which decodes entity pairs from both forward and backward extraction. The bidirectional structure effectively mitigates the error propagation problem and aligns the subject–object pairs. The translating decoding approach handles the overlapping triple problem. Finally, a (entity pair, relation) bipartite graph is designed to achieve practical relationship judgement. Experiments show that our model outperforms previous methods and achieves state-of-the-art performance on NYT and WebNLG. We achieved F1-scores of 92.7% and 93.8% on the two datasets. Meanwhile, in various complementary experiments on complex scenarios, our model demonstrates consistent performance gain in various complex scenarios. Full article

A corrosion inhibitor was widely used to improve corrosion resistance of steel bar in reinforcement concrete structure. A kind of multi-component corrosion inhibitor, which is composed of organic and inorganic substances, was developed in this research. This corrosion inhibitor was comparatively studied with various other inhibitors by using open circuit potential (OCP), electrochemical impedance spectroscopy (EIS), and cyclic voltammetry (CV) methods. The results show that the OCP values and charge transfer resistance (calculated by EIS curves) of the multi-component corrosion inhibitor remain, respectively, as high as −0.45 V and 932.19 kΩ·cm⁻² after 60 days immersion, which are significantly better than other groups. Wide passivation interval and various peaks in cyclic voltammograms (CV) were applied to analyze the mechanism of adsorption (organic substance) and oxidation–reduction reactions (inorganic substance). The functional groups -OH in triethanolamine (TEA) and tri-isopropanolamine (TIPA) bond to the steel bar surface quickly, behaving as an adsorbent of organic substance in early age. An additional protective precipitate related to the reactions of Fe³⁺ was formed by inorganic substances (Fe₂(MoO₄)₃ and FePO₄), which is consistent with the EIS results and equivalent electrochemical circuits. As an eco-friendly substitute, multi-component corrosion inhibitors possess similar or even better protecting effects on steel bars in comparison to calcium nitrite. In addition, the concept of a “multi-functional protective film” was proposed, providing a new insight to achieve modified anti-corrosion capacity of inhibitors. Full article

For many years, convolutional neural networks dominated the field of computer vision, not least in the medical field, where problems such as image segmentation were addressed by such networks as the U-Net. The arrival of self-attention-based networks to the field of computer vision through ViTs seems to have changed the trend of using standard convolutions. Throughout this work, we apply different architectures such as U-Net, ViTs and ConvMixer, to compare their performance on a medical semantic segmentation problem. All the models have been trained from scratch on the DRIVE dataset and evaluated on their private counterparts to assess which of the models performed better in the segmentation problem. Our major contribution is showing that the best-performing model (ConvMixer) is the one that shares the approach from the ViT (processing images as patches) while maintaining the foundational blocks (convolutions) from the U-Net. This mixture does not only produce better results ($DICE=0.83$) than both ViTs ($0.80$/0.077 for UNETR/SWIN-Unet) and the U-Net ($0.82$) on their own but reduces considerably the number of parameters (2.97M against 104M/27M and 31M, respectively), showing that there is no need to systematically use large models for solving image problems where smaller architectures with the optimal pieces can get better results. Full article

This paper introduces a novel method to predict when a Google translation is better than other machine translations (MT) in Dutch. Instead of considering fidelity, this approach considers fluency and readability indicators for when Google ranked best. This research explores an alternative approach in the field of quality estimation. The paper contributes by publishing a dataset with sentences from English to Dutch, with human-made classifications on a best-worst scale. Logistic regression shows a correlation between T-Scan output, such as readability measurements like lemma frequencies, and when Google translation was better than Azure and IBM. The last part of the results section shows the prediction possibilities. First by logistic regression and second by a generated automated machine learning model. Respectively, they have an accuracy of 0.59 and 0.61. Full article

A simple supervised learning model can predict a class from trained data based on the previous learning process. Trust in such a model can be gained through evaluation measures that ensure fewer misclassification errors in prediction results for different classes. This can be applied to supervised learning using a well-trained dataset that covers different data points and has no imbalance issues. This task is challenging when it integrates a semi-supervised learning approach with a dynamic data stream, such as social network data. In this paper, we propose a stream-based evolving bot detection (SEBD) framework for Twitter that uses a deep graph neural network. Our SEBD framework was designed based on multi-view graph attention networks using fellowship links and profile features. It integrates Apache Kafka to enable the Twitter API stream and predict the account type after processing. We used a probably approximately correct (PAC) learning framework to evaluate SEBD’s results. Our objective was to maintain the accuracy and confidence levels of our framework to enable successful learning with low misclassification errors. We assessed our framework results via cross-domain evaluation using test holdout, machine learning classifiers, benchmark data, and a baseline tool. The overall results show that SEBD is able to successfully identify bot accounts in a stream-based manner. Using holdout and cross-validation with a random forest classifier, SEBD achieved an accuracy score of 0.97 and an AUC score of 0.98. Our results indicate that bot accounts participate highly in hashtags on Twitter. Full article

The surrounding rock structure of the crossing-seam roadway is poor and is susceptible to anchorage failure phenomena, such as top plate sinking and convergence deformation under high ground stress. These issues can cause significant deformation of the surrounding rock over time, resulting in challenging engineering problems. To address this issue, we studied the failure modes and destabilization mechanisms of the surrounding rock in different crossing-seam roadways by field tests and numerical simulations. The results show that since the rock strata in these roadways are extremely unstable and highly susceptible to high horizontal stress, the weak surrounding rock presents the mode of full-section plastic failure. The roof is damaged more seriously than the floor and both walls. In this case, the basic anchorage layer in the original scheme is not thick and rigid enough to support these roadways. Thus, the surrounding rock deforms severely and persistently, which is one of the engineering failure characteristics. To solve this problem, a new scheme of “prompt thick-layer end anchorage + full-length lag grouting anchorage + secondary continuous reinforcement” was proposed based on the continuous roof control theory. According to the industrial test, this scheme can successfully control the long-term large deformation of the weak surrounding rock in crossing-seam roadways. Notably, the deformation of the top plate decreased by 56.65% and the deformation of the two walls decreased by 66.35%. Its design concept will provide important references for controlling the surrounding rock in similar soft rock roadways. Full article

In this paper, we describe the upgrade of a small electrostatic dust accelerator located at the University of Stuttgart. The newly developed dust source, focusing lens, differential detector and linac stage were successfully installed and tested in the beam line. The input voltage range of the dust source was extended from 0–20 kV to 0–30 kV. A newly developed dust detector with two differential charge sensitive amplifiers is employed to monitor particles with speeds from several m/s to several km/s and with surface charges above 0.028 fC. The post-stage linac provides an additional acceleration ability with a total voltage of up to 120 kV. The entire system of this dust accelerator works without protection gas and without a complex high voltage terminal. The volumes to be pumped down are small and can be quickly evacuated. The new system was used to accelerate micron- and submicron-sized metal particles or coated mineral materials. Improvements in the acceleration system allow for a wider variety of dust materials and new applications. Full article

The supply of energy with the correct parameters to electrical appliances is possible with the use of energy converters. When a direct current is required, rectifier bridges are needed. These can be made using rectifier diodes. The problem of excessive junction temperatures in power diodes, which are used to build rectifier bridges and power converters, was recognized. For this reason, research work was carried out to create a model of a rectifier diode placed on a heat sink and to analyze the heat dissipation from the junction of this diode under forced convection conditions. The results obtained from the simulation work were compared with the results of thermographic temperature measurements. The boundary conditions chosen for the simulation work are presented. A method is also presented that determined the convection coefficient under forced convection conditions. The difference between the simulation results and the results of the thermographic measurements was found to be 0.1 °C, depending on the power dissipated at the junction and the air velocity around the diode. Full article

For the trajectory optimization of the time–jerk of robotic arms with a chicken swarm optimization algorithm, using five-order B-spline interpolation can ensure smooth and continuous acceleration, but, due to the performance problems of the algorithm, the low solution accuracy and the slow convergence speed, the ideal trajectory curve cannot be obtained. To address these problems, an improved chicken swarm algorithm based on a parallel strategy and dynamic constraints (PDCSO) is proposed, where the rooster update method is employed with a parallel strategy using X-best guidance and a Levy flight step. Dynamic constraints for the rooster are given, followed by the hens, and the optimal rooster position that improved the convergence accuracy while preventing the local optimum was determined. Simulation experiments using 18 classical test functions showed that the PDCSO algorithm outperformed other comparative algorithms in terms of convergence speed, solution accuracy and solution stability. Simulation validation in ADAMS and real machine tests proved that PDCSO can effectively reduce the running time and motion shock for robotic arms and improve the execution efficiency of such arms. Full article