Model Parameters Research Articles

Phenotypic homogenization and potential fitness constraints following non-native introgression in an endemic sportfish.

Introgressive hybridization may lead to contrasting evolutionary outcomes that are difficult to predict, since they depend on the fitness effects of endogenous genomic interactions and environmental factors. Conservation of endemic biodiversity may be more effective with require direct measurement of introgressed ancestry and fitness in wild populations, especially for keystone taxa at risk of hybridization following species introductions. We assessed the relationship of non-native ancestry with growth and body condition in the basin-restricted Neosho Bass (Micropterus velox; NB), focusing on two streams in the NB native range that are admixed extensively with non-native Smallmouth Bass (M. dolomieu; SMB). We quantified genetic composition of 116 fish from Big Sugar Creek (N=46) and Elk River (N=70) at 14 microsatellite loci. Using back-calculated total length-at-age estimated from sagittal otoliths, we assessed whether genetic ancestry explained variation in von Bertalanffy growth model parameters, accounting for sex and stream effects. We then assessed the relationship of ancestry and body condition. We found no differences in growth parameters by sex, stream, or ancestry, suggesting phenotypic homogenization which could be mediated by selection on body size. We found a negative correlation between SMB ancestry and condition, including lower condition in Big Sugar Creek, possibly reflecting a trade-off between maximum length and condition with respect to overall fitness. We show that ongoing non-native introgression, which may be augmented by anthropogenic SMB introductions, may attenuate evolutionary differentiation between species and directly influence fitness, possibly having critical implications for long-term persistence and management of adaptive potential in a popular and ecologically important endemic sportfish.

CGLight: An effective indoor illumination estimation method based on improved convmixer and GauGAN

Illumination consistency is a key factor for seamlessly integrating virtual objects with real scenes in augmented reality (AR) systems. High dynamic range (HDR) panoramic images are widely used to estimate scene lighting accurately. However, generating environment maps requires complex deep network architectures, which cannot operate on devices with limited memory space. To address this issue, this paper proposes CGLight, an effective illumination estimation method that predicts HDR panoramic environment maps from a single limited field-of-view (LFOV) image. We first design a CMAtten encoder to extract features from input images, which learns the spherical harmonic (SH) lighting representation with fewer model parameters. Guided by the lighting parameters, we train a generative adversarial network (GAN) to generate HDR environment maps. In addition, to enrich lighting details and reduce training time, we specifically introduce the color consistency loss and independent discriminator, considering the impact of color properties on the lighting estimation task while improving computational efficiency. Furthermore, the effectiveness of CGLight is verified by relighting virtual objects using the predicted environment maps, and the root mean square error and angular error are 0.0494 and 4.0607 in the gray diffuse sphere, respectively. Extensive experiments and analyses demonstrate that CGLight achieves a balance between indoor illumination estimation accuracy and resource efficiency, attaining higher accuracy with nearly 4 times fewer model parameters than the ViT-B16 model.

Stability discussion and application study of pseudo-corner models

Accurate plastic flow modelling under complex working conditions is crucial for metal deformation simulations. Recently, some advanced pseudo-corner models have been developed to describe corner effects and analyze strain localization problems. The present work consists of three parts. The first part discusses the intrinsic stability of the pseudo-corner model class, which forms the premise of application analysis. The second part applies the pseudo-corner models and the associated flow rule (AFR) to buckling onset estimation, plastic post-buckling analysis and shear band analysis. The experimental conditions are strictly reproduced and the optimal model parameters are determined. The results reveal that the pseudo-corner models and AFR are indistinguishable in the buckling onset estimation. AFR overestimates the post-buckling strength of circular tubes under axial compression, and cannot reproduce the shear band development during sheet bending; while the pseudo-corner models have better prediction performance in both scenarios. The results also suggest that the parameter values of pseudo-corner models are apparently inconsistent in the above two types of problems. Then in the third part, two representative influencing factors including strain gradient plasticity and initial imperfections are discussed, and this inconsistency is finally attributed to the shortwave surface defect which however is usually neglected by previous studies.

A General DNA-Like Hybrid Symbiosis Framework: An EEG Cognitive Recognition Method.

In electroencephalogram (EEG) cognitive recognition research, the combined use of artificial neural networks (ANNs) and spiking neural networks (SNNs) plays an important role to realize different categories of recognition tasks. However, most of the existing studies focus on the unidirectional interaction between an ANN and a SNN, which may be overly dependent on the performance of ANNs or SNNs. Inspired by the symbiosis phenomenon in nature, in this study, we propose a general DNA-like Hybrid Symbiosis (DNA-HS) framework, which enables mutual learning between the ANN and the SNN generated by this ANN through parametric genetic algorithm and bidirectional interaction mechanism to enhance the optimization ability of the model parameters, resulting in a significant improvement of the performance of the DNA-HS framework in all aspects. By comparing with seven typical EEG cognitive recognition models, the performance of the seven hybrid network frameworks constructed using this method on different EEG-based cognitive recognition tasks are all improved to different degrees, verifying the effectiveness of the proposed method. This unified hybrid network framework similar to the DNA structure is expected to open up a new approach and form a new research paradigm for EEG-based cognitive recognition task.

An Optimal Two-Dimensional Maintenance Policy for Self-Service Systems with Multi-Task Demands and Subject to Competing Sudden and Deterioration-Induced Failures

Self-service systems, such as electric vehicle charging piles (EVCPs), are typically deployed without on-site personnel. While frequent maintenance ensures high service revenue, it also leads to significant maintenance setup costs. Therefore, balancing service revenue and maintenance costs is essential for profit maximization. In this paper, we develop a maintenance policy optimization framework to maximize the profit rate of a fleet of self-service systems. First, we propose a maintenance policy that ensures sufficient functional systems while preventing high corrective maintenance costs. Next, we model the fleet's state transition process and its profit rate by characterizing two unique failure-induced demand-and-system interactions: demand switching and stepwise demand arrival rates, where the demands involve multiple tasks and systems are subject to multiple failure modes with non-constant occurrence rates. We develop a Tabu-search algorithm with random exploration to optimize the maintenance policy. Building on this, we investigate the impacts of key model parameters through a case study of thirteen EVCPs in Hong Kong and draw implications for profit maximization.

Modeling the Impact of Short-Term Displacement of Domestic Animals on the Transmission Dynamics of Brucellosis

Brucellosis, a prevalent zoonotic disease caused by bacteria from the genus Brucella, presents a substantial public health and economic challenges worldwide. In this paper, we formulate a two-patch deterministic model to investigate the impact of short-term displacement of domestic animals on the transmission dynamics of brucellosis within and between patches with heterogeneous risk of infection. The model analysis is performed, and the basic reproduction number (R0) is computed and used in the stability analysis of equilibria. A global sensitivity analysis is performed using Latin Hypercube Sampling (LHS) approach to check the impact of model parameters over a period of time. Numerical simulation results indicate that the increase in time spent by domestic animals in high-risk areas, such as communal grazing areas or wildlife interfaces, significantly increases the spread of infections. This underscores the importance of considering short-term animal movements to communal grazing areas and water points in brucellosis management strategies. It equips policymakers and stakeholders with actionable insights to combat the burden of the disease.

COSMO-RS prediction, experimental investigation, and mechanism analysis: A new approach to separating the n-hexane–tert-butanol azeotropic system via liquid-liquid extraction with ionic liquids

n-Hexane and tert-Butanol (TBA) are frequently utilized to refine the conditions for Grignard, Wittig, and Friedel-Crafts alkylation reactions. During the synthesis of loratadine, the formation of azeotropic mixtures n-hexane and TBA, which are challenging to separate through conventional distillation, is inevitable. This study utilizes the COSMO-RS model to identify suitable ionic liquids (ILs) for the separation of the n-hexane–TBA azeotropic system. Based on solvent capacity and selectivity, 1-ethyl-3-methylimidazolium trifluoroacetate ([EMIM][TFA]), 1-butyl-3-methylimidazolium trifluoroacetate ([BMIM][TFA]), and 1-hexyl-3-methylimidazolium trifluoroacetate ([HMIM][TFA]) were selected as extractants. The liquid–liquid equilibrium (LLE) data for the n-hexane–TBA – ILs ternary system were measured at 303.15 K and atmospheric pressure. Distribution coefficients and selectivity were calculated to evaluate the performance of the extractants, with the NRTL model used to correlate the experimental LLE data. The consistency of the NRTL model parameters was corroborated through topological analysis associated with the Gibbs tangent principle. Quantum chemical calculations, including interaction energy, ESP analysis, IGMH analysis, and QTAIM topological analysis, were performed to explore the separation mechanism at the molecular level. The results indicated that the interaction energies between the ILs and TBA were higher than those between the ILs and n-hexane, indicating a stronger attraction of ILs to TBA. Consequently, the ILs effectively separated TBA from the n-hexane–TBA azeotropic system, with extraction capacities ranked as [EMIM][TFA] > [BMIM][TFA] > [HMIM][TFA]. The quantum chemical calculations successfully explained the experimental results, aligning with COSMO-RS model predictions and confirming their reliability. IGMH and QTAIM topological analyses systematically explored the types and strengths of interactions, revealing that hydrogen bonds are predominant between the ILs and TBA, with additional contributions from van der Waals forces. Furthermore, the hydrogen bond strengths between TBA and the anions and cations of the ILs are classified as strong and moderate, respectively. This work provides valuable insights into the separation of azeotropic systems using ILs, elucidating the underlying mechanisms behind this process.

Robust Driving Control Design for Precise Positional Motions of Permanent Magnet Synchronous Motor Driven Rotary Machines with Position-Dependent Periodic Disturbances

Position-dependent periodic disturbances often limit the accuracy and smoothness of the positional motion of permanent magnet synchronous motor (PMSM)-driven rotary machines. Because the period of these disturbances varies with the motion velocity of the rotary machine, spatial domain control methods such as spatial iterative learning control (SILC) and spatial repetitive control (SRC) have been proposed and applied to improve rotary machine motion control designs. However, problems with learning period convergence and rotary machine dynamics significantly affect transient motion, further constraining the overall motion performance. To address these challenges, this study developed a robust driving control (RDC) that integrates a robust control design with position-dependent periodic disturbance feedforward compensation, rotary machine dynamics compensation, and proportional–proportional integral feedback control. A position-dependent periodic disturbance model was developed using multiple position–sinusoidal signals and identified using a spatial fast Fourier transform. RDC compensates for disturbances and dynamics and considers the effects of model parameter uncertainty and modeling error on the stability of the control system. Several motion control experiments were conducted on a PMSM test bench to compare the RDC, SILC, and SRC. The experimental results demonstrated that although both SILC and SRC can effectively suppress position-dependent periodic disturbances, SILC provides slower position error convergence owing to the learning process, and SILC and SRC result in significant position errors because of the influence of the PMSM-driven rotary machine dynamics. RDC not only suppresses position-dependent periodic disturbances, but also significantly reduces position errors with a reduction rate of 90%. Therefore, the RDC developed in this study effectively suppressed position-dependent periodic disturbances and significantly improved both the transient-state and steady-state position-tracking performances of the PMSM-driven rotary machine.

Modeling Nitrogen Recovery and Water Transport in Gas-Permeable Membranes

This study presents a new modeling approach for nitrogen recovery for gas-permeable membrane (GPM) contactors, including both ammonia and water transport dynamics. A distinct feature of the model is its capacity to model water transport across the membrane, which has been overlooked in most publications. Osmotic pressure differences are used to predict the behavior of ammonia and water transport in the GPM. Experiments carried out to develop, test and calibrate the model examined the dynamics of ammonia and water transport through the GPM at various nitrogen concentrations. Specifically, the GPM contactor was tested for nitrogen recovery from high-strength synthetic wastewaters (2.4-10.6 g N/L) at 35°C and at pH 9. The initial volume of the trapping solution (diluted H2SO4) was 10 times lower than that of the synthetic wastewater, aiming to concentrate the recovered nitrogen. The estimated ammonia transport constant (Km) ranged between (1.2 – 2.1)·10-6 m/s and water transport constant Kw between (2.8 – 8.2)·10-10 m/(s bar). Numerical determination of the model parameters revealed high R² values, demonstrating strong agreement with experimental data.

Broad Distributed Game Learning for intelligent classification in rolling bearing fault diagnosis

As a new Single Layer Feedforward Network (SLFN) architecture, Broad Learning System (BLS) has been widely used in the field of fault diagnosis because of its fast-training speed and high generalization capability. However, when features in different classes of signals are similar or weak, BLS generates a large number of redundant features that may be difficult to classify accurately. In view of this, a new Broad Distributed Game Learning (BDGL) method is proposed in this paper, which maps data into the game space by constructing two non-parallel game hyperplanes to achieve game and segmentation of different similar features, thereby making the data linearly differentiable in the game space. Meanwhile, a linear distribution constraint term is designed to reduce noise fitting and weak feature learning in training data learning by limiting the complexity of model parameters, thereby making the solution of the objective function simpler and faster. By comparing the Precision, Recall, F-score, Kappa and Accuracy of BDGL and the comparison methods on the two types of rolling bearing experimental data, the results show that BDGL has a high classification accuracy. In addition, the experimental results on small and noisy samples once again demonstrate the effectiveness of BDGL, which provides an efficient solution for rolling bearing fault diagnosis.

Decoding Pakistan's rainfall: optimizing predictions from ARIMA to SARIMA with seasonal adjustments

Abstract. This study aims to improve the accuracy of rainfall forecasting in Pakistan by comparatively analyzing the performance of two models, ARIMA and SARIMA, to optimize the forecasting methodology. The study points out that although the ARIMA model performs well in time series analysis, it has shortcomings in handling data with significant seasonal variations. Therefore, the SARIMA model was introduced and it performed better in forecasting seasonal variations. Future research should consider combining the SARIMA model with models that can explain global climate phenomena such as El Nio and La Nia to enhance the accuracy of forecasts. In addition, ways to automate and improve the selection of model parameters should be explored to make the SARIMA model more efficient and accurate. The introduction of the SARIMA model has significantly improved prediction accuracy and contributed to more efficient planning and management of water resources. Areas where improvements can be made include reserving water resources in advance during the dry season or allocating water resources appropriately during the rainy season to support irrigation agriculture, urban water supply, flood control measures, etc. These enhanced forecasting methods help Pakistan cope with climate change challenges.

Research on Data Feature Selection of Sharing Platform Based on Particle Swarm Optimization Algorithm

Abstract. This paper is based on the optimization of machine learning models such as SVM (Support Vector Machine) and LightGBM through the application of PSO (Particle Swarm Optimization) and combines shared platform recovery data to provide training material for machine models. The goal is to accelerate the parameter configuration of machine learning models and enhance the rigor of feature selection through the application of algorithmic models. The experimental results show that PSO performs well in optimizing parameters for machine learning models like SVM and LightGBM, achieving high levels in evaluation metrics such as accuracy, recall, precision, and F1 score. Based on this, a framework is proposed for embedding the PSO algorithm into a shared platform architecture, including layers for data collection and preprocessing, algorithm integration and optimization, decision support and service, and feedback and optimization. This framework allows for precise predictions of user behavior, market demand, and other factors, achieving automated scheduling of machine model parameters.

Constraining anisotropic universe under f(R,T) theory of gravity

We try to find the possibility of a Bianchi V universe in the modified gravitational field theory of f(R,T). We have considered a Lagrangian model in connection between the trace of the energy-momentum tensor T and the Ricci scalar R. In order to solve the field equations a power law for the scaling factor was also considered. To make a comparison of the model parameters with the observational data we put constraint on the model under the datasets of the Hubble parameter, Baryon Acoustic Oscillations, Pantheon, joint datasets of Hubble parameter + Pantheon and collective datasets of the Hubble parameter + Baryon Acoustic Oscillations + Pantheon. The outcomes for the Hubble parameter in the present epoch are reasonably acceptable, especially our estimation of this H0 is remarkably consistent with various recent Planck Collaboration studies that utilize the Λ-CDM model.

Sieve estimation of the accelerated mean model based on panel count data

Panel count data are gathered when subjects are examined at discrete times during a study, and only the number of recurrent events occurring before each examination time is recorded. We consider a semiparametric accelerated mean model for panel count data in which the effect of the covariates is to transform the time scale of the baseline mean function. Semiparametric inference for the model is inherently challenging because the finite-dimensional regression parameters appear in the argument of the (infinite-dimensional) functional parameter, i.e., the baseline mean function, leading to the phenomenon of bundled parameters. We propose sieve pseudolikelihood and likelihood methods to construct the random criterion function for estimating the model parameters. An inexact block coordinate ascent algorithm is used to obtain these estimators. We establish the consistency and rate of convergence of the proposed estimators, as well as the asymptotic normality of the estimators of the regression parameters. Novel consistent estimators of the asymptotic covariances of the estimated regression parameters are derived by leveraging the counting process associated with the examination times. Comprehensive simulation studies demonstrate that the optimization algorithm is much less sensitive to the initial values than the Newton–Raphson method. The proposed estimators perform well for practical sample sizes, and are more efficient than existing methods. An example based on real data shows that due to this efficiency gain, the proposed method is better able to detect the significance of practically meaningful covariates than an existing method.

Mechanical properties and constitutive model of deep marble under alternating creep-fatigue action

Given the time dependence and susceptibility to disturbance of deep rocks, this paper analyzes mechanical properties, acoustic emission (AE) signals, and pore radius distribution characteristics of deep marble under alternating creep-fatigue (ACF) action at different confining pressures by experiment and theoretical analysis. The results show that: (1) the alternation of creep and fatigue accelerates the rock damage, and the number of alternations increases with the increasing confining pressure. (2) Strain increments exhibit a pattern of first decreasing and then increasing, and they undergo a transformation from “V”-type to “U”-type with the increase of confining pressure. (3) The AE ringing count is mainly generated in the fatigue segment. The radius of large pores inside the rock is constrained by the confining pressure. (4) The ACF constitutive model is established, where model parameters show a nonlinear decreasing trend of the power exponential function with the strain increment.

Statistical Response of ENSO Complexity to Initial Condition and Model Parameter Perturbations

Abstract Studying the response of a climate system to perturbations has practical significance. Standard methods in computing the trajectory-wise deviation caused by perturbations may suffer from the chaotic nature that makes the model error dominate the true response after a short lead time. Statistical response, which computes the return described by the statistics, provides a systematic way of reaching robust outcomes with an appropriate quantification of the uncertainty and extreme events. In this paper, information theory is applied to compute the statistical response and find the most sensitive perturbation direction of different El Niño–Southern Oscillation (ENSO) events to initial value and model parameter perturbations. Depending on the initial phase and the time horizon, different state variables contribute to the most sensitive perturbation direction. While initial perturbations in sea surface temperature (SST) and thermocline depth usually lead to the most significant response of SST at short and long ranges, respectively, initial adjustment of the zonal advection can be crucial to trigger strong statistical responses at medium range around 5–7 months, especially at the transient phases between El Niño and La Niña. It is also shown that the response in the variance triggered by external random forcing perturbations, such as the wind bursts, often dominates the mean response, making the resulting most sensitive direction very different from the trajectory-wise methods. Finally, despite the strong non-Gaussian climatology distributions, using Gaussian approximations in the information theory is efficient and accurate for computing the statistical response, allowing the method to be applied to sophisticated operational systems. Significance Statement The purpose of this work is to better understand how El Niño–Southern Oscillation (ENSO) responds to changes in its initial state and internal dynamics or external forcings. A statistical quantification of this response allows for the comprehension of the triggering conditions and the effect of climate change on the occurrence frequency and strength of each type of ENSO event. Such a study also allows to detect the most sensitive perturbation directions, which has practical significance in guiding anthropogenic activities. The approach used to study the response in this work is through the framework of information theory, which allows for an unbiased and robust assessment of the statistical response that is not affected by the turbulent dynamics of the system.

Fractional-Order Mathematical Modeling and Stability Analysis of Tumor-Immune System Relation Including Chemotherapy Drug Effect

Cancer diseases rank second in global mortality rates. Among the prevalent approaches for treating tumors, chemotherapy stands out as a prominent method capable of reducing tumor size and managing the advancement of cancer ailments. To gain deeper insights into the intricacies of chemotherapy mechanisms, we devised a Fractional-order mathematical model, depicting tumor growth in the presence of chemotherapy. This comprehensive model encompasses both the immune system's response and the effects of drug therapy. To demonstrate that the system is biologically meaningful, we examined the existence and uniqueness of solutions and proved the positivity and boundedness of solutions. We characterize the dynamics properties of this differential equation model by finding the equilibrium points and exploring the stability conditions in a range of model parameters. In addition, we have performed numerical simulations considering different parameter values. Furthermore, to demonstrate the memory effect of the fractional derivative, we have simulated the dynamic behavior of the system for different orders of fractional derivatives. So, we have concluded that the memory effect occurs as the α decreases from 1 and the chemotherapy drug is quite effective on the populations. We hope that this work will contribute to helping medical scientists take the necessary measures during the screening process and treatment.

Simplified dynamic modeling and Fuzzy gain scheduled PID control of woodward-governor controlled grid-interactive twin-shaft gas turbine plants

Biomass-integrated gas turbines for electricity generation are competitive with coal and nuclear energy production. Gas turbine-based power generation system looks to be an excellent solution amid the emerging environmental conditions and the increased energy crisis. This paper presents a simplified dynamic analysis model for biomass-based, Woodward governor-controlled, twin-shaft heavy-duty gas turbine (HDGT) plants, alongside an intelligent controller for stability enhancement. Grid-connected HDGTs can become unstable under load disturbances, potentially causing system shutdowns. A simplified model for twin-shaft gas turbines, rated from 18.2 MW to 106.7 MW, has been identified based on control characteristics during startup. The speed controller is dominant, while the acceleration controller is only active at startup, and the temperature controller's effect is minimal during normal operation. This model suits all dynamic studies of twin-shaft gas turbines, regardless of varying power ratings and rotor time constants. For improving the grid stability, a Takagi-Sugeno-Kang (TSK) fuzzy gain scheduling PID controller has been proposed in this work and its behavior is validated with 5001M, 7001Ea, and 9001Ea models. Step response of fuzzy PID controller under load disturbances and set-point variations are compared with fixed gain PID controllers tuned by Ziegler-Nichols (ZN) and performance index-based tuning using the typical gas turbine model. Further the controller behavior is validated with field test-based model parameters as derived from the real-world combustion turbines used in Alaskan Railbelt system. Extensive research simulation and validation results reveal that the fuzzy self-tuning PID controller showed superior adaptability for grid-interactive twin-shaft HDGTs under load variations and set-point variations. Time domain specifications and the performance indices confirms that the proposed fuzzy tuned PID controller ensure reliable and stable operation for the energy management in grid-operative mode. The proposed simplified model and intelligent control logic enable various dynamic studies in grid-connected environments for both simple cycle and combined cycle operations.

Fatigue Life Prediction of 2024-T3 Al Alloy by Integrating Particle Swarm Optimization-Extreme Gradient Boosting and Physical Model.

The multi-parameter characteristics of the physical model pose a challenge to the fatigue life prediction of 2024-T3 aluminum (Al) alloy. In response to this issue, a parameter-solving method that integrates particle swarm optimization (PSO) with extreme gradient boosting (XGBoost) is proposed in this study. The fatigue performance and failure mechanism of the 2024-T3 Al alloy are analyzed. Furthermore, the fatigue life prediction physical model of the 2024-T3 Al alloy is established by using the energy method of fracture mechanics. The physical model incorporates critical physical parameters. Meanwhile, the PSO algorithm optimizes the hyperparameters of the XGBoost model based on fatigue data of the 2024-T3 Al alloy. Eventually, the optimized XGBoost model is used to solve the parameters of the physical model. Furthermore, the analytical equation of the fatigue life prediction model is obtained. This paper provides a new method for solving the parameters of the fatigue life prediction model, which reduces the error and cost of obtaining the model parameters in the experiment and shortens the time required.

A lightweight and explainable model for driver abnormal behavior recognition

With the advancement of intelligent transportation systems, accurate identification of driver abnormal behavior is crucial for enhancing road safety. However, the limited computing power of vehicular systems poses a challenge for running efficient and explainable behavior recognition models. This paper proposes a lightweight and explainable driver abnormal behavior recognition model based on an improved You Only Look Once version 8 (YOLOv8). Firstly, a Spatial and Channel Reconstruction Convolution (SCConv) module is introduced to optimize the Convolution to Feature (C2f) structure, enhancing the model's feature extraction capabilities while reducing parameter redundancy. Secondly, a Spatial Pyramid Pooling with Fast Large Separable Kernel Attention (SPPF-LSKA) module is designed to better capture image context and integrate global information. Additionally, a Dynamic upsample (Dysample) module is introduced to improve the model's ability to capture subtle driver movements. Lastly, a Lightweight Shared Group Normalization Convolution Detection Head (LSGCDH) is designed to enhance the model's generalization ability, significantly reducing the model's computational load, parameter count, and size. Experimental results demonstrate that our approach has significant advantages for edge device deployment compared to mainstream algorithms. The visualization results effectively corroborate the role of each improved structure, enhancing the explainability of the abnormal behavior recognition model, which is beneficial for deployment in vehicular systems and contributes to improving road traffic safety.