Dynamic Loss Research Articles

Disentangling Dynamic from Thermodynamic Summer Ice Area Loss from Observations (1979–2021): A Potential Mechanism for a “First-Time” Ice-Free Arctic

Abstract In recent decades, the Arctic minimum sea ice extent has transitioned from a predominantly thick multiyear ice cover to a thinner seasonal ice cover. We partition the total (observed) Arctic summer area loss into thermodynamic and dynamic (convergence, ridging, and export) sea ice area loss during the satellite era from 1979 to 2021 using a Lagrangian sea ice tracking model driven by satellite-derived sea ice velocities. Results show that the thermodynamic signal dominates the total summer ice area loss and the dynamic signal remains small (∼20%) even in 2007 when dynamic loss was largest. Sea ice loss by compaction (within pack ice convergence) dominates the dynamic area loss, even in years when the export is largest. Results from a simple (Ekman) free-drift sea ice model, supported by results from the Lagrangian model, suggest that nonlinear effects between dynamic and thermodynamic area loss can be important for large negative anomalies in sea ice extent, in accord with previous modeling studies. A detailed analysis of two all-time record minimum years (2007 and 2012)—one with a semipermanent high in the southern Beaufort Sea and the other with a short-lived but extreme storm in the Pacific sector of the Arctic in late summer—shows that compaction by Ekman convergence together with large thermodynamic melt in the marginal ice zone dominated the sea ice area loss in 2007 whereas, in 2012, it was dominated by Ekman divergence amplified by sea–ice albedo feedback—together with an early melt onset. We argue that Ekman divergence from more intense summer storms when the sun is high above the horizon is a more likely mechanism for a “first-time” ice-free Arctic.

Effect of silica fume on frost resistance and recyclability potential of recycled aggregate concrete under freeze–thaw environment

In this study, the effects of silica fume (SF) dosages (0 %, 3 %, 5 %, 8 %, 11 %, 13 %, and 16 %) on the mechanical properties (compressive strength and splitting tensile strength) and frost resistance (relative dynamic elasticity modulus and mass loss rate) of recycled aggregate concrete (RAC) prepared with 100 % recycled concrete aggregate (RCA) under freeze–thaw (F-T) environment were investigated. To assess the recyclability potential of RAC, the apparent density, water absorption, crushing index, and attached mortar content of second-generation RCA after F-T cycles were also studied. The phase composition and microstructure near the interfacial transition zone (ITZ) of RAC were characterized by X-ray diffraction, thermogravimetric, scanning electron microscopy, and microhardness. The results indicate that the incorporation of SF effectively improves the mechanical properties and frost resistance of RAC, with an optimal SF dosage around 13 %. After 400 F-T cycles, the compressive strength of RAC with 13 % SF dosage can still reach 40.1 MPa, which is classified as medium-strength concrete. All second-generation RCAs achieves Class III, but RCA obtained from RAC containing 13 % SF has higher potential to upgrade to Class II and reuse in 40 MPa structural concrete. During the F-T process, the ITZ width of RAC with 13 % SF dosage only increased by 3 μm, and the average hardness of ITZ decreased by 6.7 %. The microscopic analysis showed that an appropriate dosage of SF is conducive to denser ITZ with higher hardness and narrower crack width, thereby hindering crack propagation within RAC during F-T cycles.

Rolling bearing intelligent fault diagnosis towards variable speed and imbalanced samples using multiscale dynamic supervised contrast learning

Deep learning-based fault diagnosis methods have already attained remarkable achievements in this field. However, rolling bearing frequently operates under variable speed conditions, and the number of healthy samples collected is often significantly larger than that of failure samples. In this paper, a multiscale dynamic supervised contrast learning (MDSupCon) framework is proposed. First, a multiscale adaptive feature extraction network is designed as the backbone, which utilizes multiple convolutional kernels to enhance feature extraction capabilities under variable speed conditions, and the branch attention mechanism is incorporated to adaptively adjust the weights of various scale branches. Second, the joint channel-space attention mechanism is constructed to enhance the importance of critical features while reducing redundant information, thereby improving fault identification accuracy and interpretability. Third, the dynamic supervised contrast loss function is designed to assign dynamic compensation factors to samples of various categories according to the training results, which reduces the impact of easily classified samples and enhances the contribution of hard-to-classify samples in imbalanced scenarios. Additionally, a dynamic cross-entropy loss is designed to train the backbone and the classifiers. The MDSupCon has achieved superior results of 89.49% and 92.15% on two bearing datasets with an imbalance ratio of 20:1 and variable speeds.

Energy-saving starting method of electric motor based on the battery-accumulator hybrid drive

The series electro-hydraulic hybrid powertrain has advantages in improving the dynamic characteristics and increasing the cruising range of battery rail vehicles. In order to reduce the large peak starting current of electric motor, an energy-saving starting method is proposed, which is using the hydraulic pump/motor to reversely drive the electric motor to restart at a speed, based on the energy reverse transfer characteristics between electric motor and hydraulic pump/motor. Firstly, the principle of restarting the electric motor at a speed is proposed based on the designed EH3. Then, PID is adopted to control the starting of electric motor which is driven by hydraulic powertrain. And a speed sensorless control algorithm is designed to control the restarting at a speed. Finally, aiming at the problem that the traditional simulation models of electro-hydraulic hybrid powertrain components are too simple and ideal. The estimation model of battery and SOC, dynamic efficiency model of hydraulic motor and dynamic iron loss model of PMSM are improved and optimized to improve the accuracy of powertrain simulation. The results both of co-simulation and experiments shows that the starting current can be reduced by 69.36 % at most and 64.91 % of battery consumption can be saved by restarting the electric motor at a speed.

EFP-Net: A Novel Building Change Detection Method Based on Efficient Feature Fusion and Foreground Perception

Over the past decade, deep learning techniques have significantly advanced the field of building change detection in remote sensing imagery. However, existing deep learning-based approaches often encounter limitations in complex remote sensing scenarios, resulting in false detections and detail loss. This paper introduces EFP-Net, a novel building change detection approach that resolves the mentioned issues by utilizing effective feature fusion and foreground perception. EFP-Net comprises three main modules, the feature extraction module (FEM), the spatial–temporal correlation module (STCM), and the residual guidance module (RGM), which jointly enhance the fusion of bi-temporal features and hierarchical features. Specifically, the STCM utilizes the temporal change duality prior and multi-scale perception to augment the 3D convolution modeling capability for bi-temporal feature variations. Additionally, the RGM employs the higher-layer prediction map to guide shallow layer features, reducing the introduction of noise during the hierarchical feature fusion process. Furthermore, a dynamic Focal loss with foreground awareness is developed to mitigate the class imbalance problem. Extensive experiments on the widely adopted WHU-BCD, LEVIR-CD, and CDD datasets demonstrate that the proposed EFP-Net is capable of significantly improving accuracy in building change detection.

Dynamical Interactions and Mass Loss within the Uranian System

The origin and evolution of planetary rings and moons remains an active area of study, particularly as they relate to the impact history and volatile inventory of the outer solar system. The Uranian system contains a complex system of rings that are coplanar with the highly inclined planetary equator relative to the orbital plane. Uranus also harbors five primary regular moons that play an important role in the distribution of material that surrounds the planet. Here we present the results of a dynamical simulation suite for the Uranian system, intended to explore the interaction between the five primary regular moons and particles within the system. We identify regions of extreme mass loss within 40 planetary radii of Uranus, including eccentricity excitation of particle orbits at resonance locations that can promote moonlet formation within the rings. We calculate a total dynamical particle mass-loss rate of 35% within 0.5 × 106 yr and 40% mass loss within 107 yr. We discuss the implications for postimpact material, including dynamical truncation of stable ring locations and/or locations of moon formation promoted by dynamical excitation of ring material.

Dynamic Loss Minimization Control of a 4-Pole/2-Pole Electronic Pole-Changing Induction Motor Using a Look-Up Table

Dynamic Loss Minimization Control of a 4-Pole/2-Pole Electronic Pole-Changing Induction Motor Using a Look-Up Table

Enhancing structural illumination microscopy with hybrid CNN-transformer and dynamic frequency loss

Deep learning techniques have augmented structured illumination microscopy (SIM) imaging by employing convolutional neural networks (CNNs) for the super-resolution reconstruction of SIM images. However, due to the intrinsic shortcomings of CNNs, i.e. limited receptive field and inadaptability to input content, their reconstruction faithfulness and imaging resolution are limited. Recently, another class of neural backbone, Transformer, has shown complementary advantages to CNNs. To this end, this work aims to borrow the recent Transformers to enhance CNN-based SIM reconstruction, thereby fully unleashing the power of neural networks for SIM imaging. Concretely, we first experimentally explore the pros and cons of CNNs and Transformers for SIM reconstruction. Then, we propose a Hybrid CNN-Transformer (HCT) model, to inherit merits of the both. To further encourage the recovery of high frequencies down-modulated by patterned illumination, we also devise a dynamic frequency loss based on the phase spectrum information, helping models focus on hard frequencies adaptively. Experiments on diverse subcellular structures demonstrate that the proposed method can evidently boost SIM reconstruction performance, even under low-light conditions or with fewer raw frames.

Pore morphology based on graphene oxide modified steel fibre concrete for freeze–thaw resistance

The effect on frost resistance of the pore morphology and pore space distribution of graphene oxide-steel fiber reinforced concrete (GO-SFRC) has been researched systematically by X-ray computed tomography (X-CT) technology. As graphene oxide is added to steel fiber concrete, the pore morphology and spatial distribution will be altered. Combined with the strength loss and dynamic elastic modulus loss of GO-SFRC specimens after freeze–thaw cycles, the effect that graphene oxide shows significant effect on the frost resistance of steel fiber concrete through analyzed. The results demonstrate that GO-SFRC concrete has significantly better frost resistance than ordinary SFRC concrete based on the analysis of compressive strength loss and dynamic elastic modulus loss of the specimens. Experiments show that the porosity and pore distribution was different for the five GO-SFRCs and the pore sphericity was closely related to the content and dispersion of graphene oxide. Overall, the GO coordination regulates the relationship between porosity, sphericity, and pore size. In addition, the compressive strength of graphene oxide-steel fiber-reinforced concrete was also analyzed in terms of pore parameters, such as pore sphericity, flatness, and porosity.

A vulnerability detection algorithm based on residual graph attention networks for source code imbalance (RGAN)

It is essential to detect potential vulnerabilities in software to ensure its safety. As software systems become more complex, traditional static vulnerability detection methods perform poorly. Currently, deep learning-based vulnerability detection models only extract source code vulnerability features using sequences or graphs. Sequential neural networks ignore structural information in the code, such as control flow diagrams and data flow diagrams. Additionally, graph neural networks cannot accurately extract features due to the lack of effective methods for extracting nodes’ features and aggregating global information. To address the above issue, we propose a vulnerability detection algorithm based on residual graph attention networks for source code imbalance (RGAN). Firstly, a local feature extraction module (PE-BL-A module) is designed. Using the sequence neural network, the module extracts various useful features, including node features in a control flow diagram based on local semantic features. Secondly, we present the Residual Graph Attention Network module (RGAT). To learn and update node features along the control flow direction, the module uses a graph attention network with residual connections. In this module, a mean biaffine attention pooling mechanism is proposed that can extract total graph vulnerability features more effectively. Thirdly, a dynamic cross-entropy loss function is designed. Using this function, it can handle sample imbalances during training. Finally, experiments conducted on several benchmark datasets demonstrate that the proposed model achieves state-of-the-art results.

Experimental study with DIC technique on the bond failure of structural concrete under repeated reversed loading with different amplitude

The bond between deformed steel rebar and concrete plays a critical role against seismic action, thus it is important to capture the dynamic bond loss and establish the bond-slip model with cyclic loading. To explain the macroscale performance such as bond-slip relationship and establish the bond strength model, it is meaningful to capture the mesoscale and local behaviors between ribs and concrete. In this paper, pull-out specimens with rebar embedded in concrete and small windows excavated in the cover concrete are adopted with DIC technique to monitor the strain variation and crack development along the interfaces under static and reversed loading of different amplitude. From the test, it is found that DIC is an effective approach to obtain the localized characteristics including the normal and shear deformations. Through parametric study, the effects of concrete cover depth and the existence of stirrups on bond deterioration in terms of both global and local behaviors are studied. Interestingly, the global slip and local crack-width show synchronically increase under the pulling-out action, which demonstrates that the macroscale behaviors could be clarified from mesoscale measurement with the aid of DIC technique.

SHGCN: Socially Enhanced Heterogeneous Graph Convolutional Network for Multi-behavior Prediction

In recent years, multi-behavior information has been utilized to address data sparsity and cold-start issues. The general multi-behavior models capture multiple behaviors of users to make the representation of relevant features more fine-grained and informative. However, most current multi-behavior recommendation methods neglect the exploration of social relations between users. Actually, users’ potential social connections are critical to assist them in filtering multifarious messages, which may be one key for models to tap deeper into users’ interests. Additionally, existing models usually focus on the positive behaviors (e.g., click , follow , and purchase ) of users and tend to ignore the value of negative behaviors (e.g., unfollow and badpost ). In this work, we present a Multi-Behavior Graph (MBG) construction method based on user behaviors and social relationships and then introduce a novel socially enhanced and behavior-aware graph neural network for behavior prediction. Specifically, we propose a Socially Enhanced Heterogeneous Graph Convolutional Network (SHGCN) model, which utilizes behavior heterogeneous graph convolution module and social graph convolution module to effectively incorporate behavior features and social information to achieve precise multi-behavior prediction. In addition, the aggregation pooling mechanism is suggested to integrate the outputs of different graph convolution layers, and a dynamic adaptive loss (DAL) method is presented to explore the weight of each behavior. The experimental results on the datasets of the e-commerce platforms (i.e., Epinions and Ciao) indicate the promising performance of SHGCN. Compared with the most powerful baseline, SHGCN achieves 3.3% and 1.4% uplift in terms of AUC on the Epinions and Ciao datasets. Further experiments, including model efficiency analysis, DAL mechanism, and ablation experiments, confirm the validity of the multi-behavior information and social enhancement.

Global Dynamics and Photon Loss in the Kompaneets Equation

Global Dynamics and Photon Loss in the Kompaneets Equation

Deep learning-based autonomous real-time digital meter reading recognition method for natural scenes

Natural scenes with variable illumination, variable target scale and angular tilt pose significant challenges to the autonomous recognition of digital meter readings. Based on this, this paper proposes a deep learning-based autonomous real-time digital meter reading recognition method for natural scenes. First, the YOLO-style corner point detection method (YOLO-CPDM) for the reading area is proposed by reconstructing the detection heads and incorporating the corner detection loss function. Its localization accuracy is further refined by embedding attention mechanism module, implementing dynamic loss function and enhancing training data diversity through offline augmentation techniques like image rotation and flipping. Then, the detected corner points are used to geometrically correct the distorted reading area by perspective transformation to mitigate the interference caused by the shooting angle. Next, the YOLO-style end-to-end reading recognition method (YOLO-EERRM) is proposed to accurately extract the characters in the reading area. Finally, the validity of the YOLO-CPDM and YOLO-EERRM was verified on a produced dataset named SYSU-DM and 2 public datasets. Compared with the State of the arts (SOTA) keypoint detection model, the mean Average Precision @ 50:95 scores of the YOLO-CPDM improved by 2.8, 4.1, and 1.1 points, respectively, while the inference latency was only 5.3 ms, and YOLO-EERRM achieved 100 % accuracy and 3.1 ms inference latency on the SYSU-DM dataset. Statistically, the complete digital meter reading recognition method has 99.6 % accuracy and 8.6 ms inference latency, indicating that the system has high recognition accuracy and practicality.

The performance analysis of screw pump stator elastomers: Polyamide 6/hydrogenated nitrile blends: Mechanical, oil resistance and tribological properties

In this paper, the mechanical properties, thermal properties, oil resistance and tribology performance of hydrogenated nitrile/polyamide 6 blends were studied systematically. The inter-molecular interaction, phase separation structure, crystallinity, dynamic loss, friction morphology and other aspects of the blends were investigated. The results showed that PA6 phase had significant reinforcing effect in HNBR matrix. Due to the strong interaction between PA6 molecular chain and HNBR molecular chain, the glass transition temperature (Tg) of the blends was increased and the dynamic loss factor (tanδ) was decreased. Meanwhile, the tribology performance of the blends was significantly improved. It was due to the formation of a microball layer with PA6 particle phase as component, which effectively isolated the contact between the friction pairs.

Dynamic hysteretic properties and specific loss power of magnetic nanoparticles in a viscous medium at different thermal baths

A system of magnetite single-domain magnetic nanoparticles in an aqueous colloidal suspension at different temperatures is simulated. In this framework, we study the response of the magnetization of the system to the presence of a time-dependent magnetic field at certain frequency. To do that, a Hamiltonian that includes the Zeeman interaction and the uniaxial magneto-crystalline anisotropy energy is considered. The dynamics of the system is driven according to the solution of the stochastic Landau–Lifshitz–Gilbert differential equation, in combination with a torque term acting on the nanoparticle in order to consider the Brown rotation influenced by the viscosity of the solvent. Thus, both Néel and Brown rotation mechanisms are contemplated. Our results allow us to conclude that Brownian relaxation mediates the alignment of the anisotropy axes with the external field. Thus, depending on the solid or liquid state of the solvent, it is possible to determine the conditions under which a reinforcement of the magnetic anisotropy can take place in order to increase the remanence and squareness of the hysteresis loops.

An inverse Jiles-Atherton model of nanocrystalline magnetic core for nanoseconds square pulsed magnetization.

The magnetic core is a key component of a linear transformer driver (LTD), and the accuracy of the core model affects the calculation of the LTD power flow and the prediction of the output waveform. In this paper, a magnetization model based on the inverse Jiles-Atherton (inverse J-A) model is developed and a particle swarm algorithm is used to identify the parameters and to obtain the variation of the parameters with the excitation characteristic. A nanoseconds square wave LTD magnetic core test platform was built to obtain the magnetization characteristics of nanocrystalline magnetic cores under different excitation characteristic parameters. Under square wave pulses, due to the presence of harmonic components, core loss is more complex. In view of the fitting deviation caused by the traditional J-A model not considering harmonic factors and anisotropy, a dynamic loss correction factor is proposed. Through a comparison of experimental and simulation results, this model can well reflect the magnetization process and has high accuracy in fitting dynamic hysteresis loops and predicting losses, which is important for guiding the design of a square pulse LTD.

Исследование динамической устойчивости модели крыла в потоке воздуха

Investigating the aircraft structure elements for compliance with dynamic stability is one of the most important ways to study their dynamic properties. The paper considers possibility of the wing model dynamic loss of stability (flutter type) in the form of a flat rod of variable geometry close to the real wing geometry in the airflow. Internal force factors manifested under the frontal aerodynamic force influence, as well as increment in the lift and aerodynamic moment arising during possible small model oscillations are taken into account from the oncoming airflow side. Due to these external forces increment, the so-called energy pumping from the airflow side occurs, and the problem becomes non-conservative. Behavior of the first three complex eigenvalues of small oscillations was numerically studied, and the critical flow velocities, at which the model could lose dynamic stability, were determined for various parameters of the external aerodynamic forces.

Enhanced Elderberry Snack Bars: A Sensory, Nutritional, and Rheological Evaluation.

Interest in functional foods is continuously increasing, having the potential to be an ally in reducing cardiometabolic risk factors. This study focuses on developing and evaluating oat- and millet-based snack bars enriched with freeze-dried elderberry powder (FDEBP), aiming to combine great taste with enhanced nutritional value, antioxidant properties, and prebiotic potential. The research encompassed a sensory evaluation, nutritional assessment, and rheological analysis of the snack bars. A hedonic test was conducted to gauge consumer preferences and overall liking, providing insights into taste, texture, and acceptance. Sensory evaluation revealed positive feedback from participants, and acceptance rating scores ranged from 7 to 8.04, the best score recorded by one of the enhanced bars with 1% FDEBP. The rheological analysis determined the bars' dynamic storage modulus (G') and loss modulus (G″), assessing the material's elasticity and mechanical properties. Results showed that the incorporation of 0.5% and 1% FDEBP in the oat and millet snack bars significantly impacted their rheological properties, enhancing structural strength. Nutritional analysis demonstrated that the snack bars provided a complete mix of macronutrients required in a daily diet. The study sheds light on the potential of functional snack bars enriched with FDEBP, offering a delectable way to access essential nutrients and bioactive compounds in a minimally processed form, without the addition of sweeteners or additives, friendly to the gut microbiota.

Accurate detection of microalgae in ship ballast water: An innovative computer vision strategy

Microalgae play a crucial role in marine ecological preservation, especially in the management of ship ballast water, where accurate microalgae detection is essential for preventing biological invasions. However, when extracting and observing microalgae from ship ballast water samples, the intricate water quality background and cell adhesion phenomena present substantial interferences for precise microalgae detection, posing significant challenges for existing microalgae detection methods. To address these issues, this paper proposes an innovative microalgae detection method based on efficient omnidimensional dynamic convolution (EOD) and adhesion loss. The novelty of our study lies in the unique loss strategy for detecting adhesive microalgae cells for the first time and its combination with EOD to address the challenges of microalgae detection in complex environments. First, a large number of microalgae images are generated using an enhanced StyleGAN3, supplementing the dataset in complex scenarios to improve the model's robustness. Second, the proposed EOD is applied to mitigate the interference from the complex background and enhance the capability of microalgae feature extraction. Finally, our newly designed adhesion loss strategy imposes penalties on adhered cells based on intersection over union (IOU) and Gaussian distance, effectively tackling the cell adhesion issue. The experimental results indicate that this method achieves a mAP@[0.5:0.95] of 82.1% for microalgae detection in complex environments, meeting real-time microalgae detection requirements. This study provides an innovative perspective for marine ecological preservation, while its highly accurate and real-time microalgae detection method also pioneers new possibilities in microalgae detection research.