Federated Learning (FL) is a key paradigm for distributed, data rich environments, leveraging devices across the Cloud-to-Edge continuum to minimise latency and bandwidth usage while preserving privacy. However, testing these systems (IoT, 5G/6G) with FL capabilities in physical testbeds is challenging due to high costs and complexity, making simulation tools essential. In this paper, we introduce a simulation framework for modelling FL workflows in Cloud-to-Edge scenarios. Implemented as an extension to the DISSECT-CF-Fog discrete-event simulator, our module reuses its workflow scheduling and execution functionalities to model FL processes with centralised orchestration, diverse network topologies, varying device capacities, and dynamic workloads. It abstracts low-level Machine Learning mechanics to focus on first-order system effects such as latency, network traffic, and energy use. We model a Holographic Type Communications (HTC) use case as a baseline scenario (S0) to orchestrate FL deployments and quantify trade-offs in round latency, traffic, and energy across four operational scenarios: scalability (S1), aggregation policies (S2), pacing strategies (S3), and privacy and compression (S4). Each scenario tracks a synthetic learning progress proxy (e.g., rising from ≈0.50 to >0.9 over 20 rounds in S0) and per-round aggregator energy (e.g., ≈1kJ for early aggregation to ≈10kJ for timeout rounds in S0). Ultimately, the module exposes how operational variations shape system costs. For example, aggressive model compression (e.g., 0.2× in S4) reduces per-round traffic proportionally, yet the corresponding energy saving is bounded by fixed per-round infrastructure costs that dominate when rounds close on a timeout rather than on transmission completion.
Accurately characterizing the interaction between an oil jet and a rotating gear is essential for predicting load-independent losses in high-speed transmissions. CFD is increasingly employed to analyze oil-jet lubrication, yet the literature still provides only a limited amount of experimental data acquired under controlled conditions, leading to persistent uncertainties regarding the full predictive maturity of available numerical approaches. This study investigates the jet-gear interaction through a dedicated experimental setup specifically designed to measure the torque generated by an impinging oil jet on a spur gear. Measurements were conducted over multiple operating conditions using a differential procedure that isolates the net effect of the jet from other mechanical contributions.In parallel, a detailed three-dimensional CFD model was developed in SimericsMP+, incorporating a refined multiphase representation of the jet and a moving-grid treatment of gear rotation. A separate model of the injection nozzle was used to determine the effective jet velocity corresponding to the measured flow rates. The combined experimental and numerical analysis enables assessing the capability of the solver to reproduce the resisting torque as a function of flow rate and rotational speed.The comparison between simulations and measurements shows consistent agreement across the investigated range, confirming the reliability of the proposed modelling strategy for analyzing jet lubrication phenomena. The validated methodology provides a transferable basis for future studies involving entire gear trains and more complex lubrication configurations.
Autonomous vehicles (AVs) promise transformative safety and mobility benefits, yet their deployment and operation in urban environments remain constrained by persistent safety concerns, especially at signalized intersections with complex pedestrian interactions. Right-turn maneuvers are particularly challenging due to occlusion-induced visibility limitations caused by urban infrastructure elements such as buildings, vegetation, and parked vehicles, which restrict the AV’s surrounding surveillance and hinder timely pedestrian detection. While vehicle-based sensors like LiDAR and cameras provide foundational awareness, they fall short under occluded conditions. Vehicle-to-Everything (V2X) communication, encompassing Vehicle-to-Vehicle (V2V), Vehicle-to-Infrastructure (V2I), and Vehicle-to-Pedestrian (V2P), offers a cooperative perception solution, allowing AVs to access critical information beyond their direct sensor field. This paper presents a MATLAB-based simulation framework designed to evaluate AV-pedestrian interactions at signalized intersections under diverse V2X communication and visibility conditions, replicating a real-world intersection in Jersey City, NJ. The framework enables real-time data sharing between AVs, infrastructure, and pedestrians to simulate detection performance and warning message exchange during right-turn scenarios. Results indicated that V2P enables early alerts with signal strength between 80%–100% allowing the ego-vehicle to decelerate safely to 5 m/s. V2I and V2V supported early detection through Road-Side-Unit (RSU) and other vehicles, speed reduction, and ensured timely response through shared warning messages. While the framework does not propose new algorithms, its primary contribution lies in the integrated simulation design that simultaneously evaluates all three V2X communication types. These findings provide actionable insights into enhancing transportation safety policy to improve AV pedestrian safety.
Distributed bearing-based target tracking is vulnerable to cyber-attacks on the communication links between sensor nodes and the fusion center, since corrupted azimuth/elevation measurements can significantly distort the geometric fusion process. This paper proposes a trust-aware distributed tracking framework that combines local Smoother Integral Sliding Mode Filter (SISMF)-based bearing estimation with an online trust-weighted fusion mechanism. In the proposed architecture, SISMF is used at each geographically separated sensing node to obtain smooth and noise-robust azimuth/elevation estimates, while cyber-resilience is introduced at the fusion center through dynamic node trust factors. Each trust factor is updated using innovation consistency, normalized innovation squared (NIS)-based statistics, log-likelihood-ratio weighting, hysteresis, hard isolation, and replay-oriented staleness detection. The resulting trust values are incorporated into weighted line-of-bearing geometric fusion and Kalman filtering so that measurements from suspicious nodes are adaptively down-weighted during tracking. Spoofing, replay/delay, and additive bias-injection attacks are simulated under a single-compromised-node switching scenario, which is motivated by the geographically separated node deployment and independent node–center communication links. The proposed approach is compared with equal-weight baseline fusion, IRLS-only robust fusion, Trust-WLS, and Trust-IRLS variants. In addition to representative trajectory and angular-error plots, Monte Carlo simulations, ablation analysis, attack-detection metrics, parameter sensitivity analysis, and runtime/scalability evaluation are provided. The results show that trust-aware fusion substantially reduces tracking error compared with non-trust baselines, with Trust-WLS achieving the lowest Monte Carlo RMSE in the tested scenario. The detection results further demonstrate that the learned trust factors provide interpretable attacked-node indicators, while the runtime analysis supports the real-time feasibility of the proposed framework for small- and medium-scale distributed sensing networks.
Current fire safety protocols in nursing homes face critical challenges due to elderly residents’ physiological vulnerabilities and limited empirical evidence on caregiver decision-making during multi-floor fire emergencies. Traditional drills and conventional computational simulations often fail to capture vertical evacuation complexity, elderly resident heterogeneity, and fine-grained caregiver-resident interactions. To address these issues, this research develops a desktop-based Virtual Reality (VR) evacuation simulation model. The simulation model enables safe and repeatable interactions between real caregivers and virtual elderly residents, while automatically recording high-resolution behavioral trajectories under controlled conditions to simulate and analyze the effects of caregivers’ initial floor location, the proportion of self-evacuating elderly residents, and the behavioral patterns of high-performance caregivers on evacuation performance. Based on extensive behavioral data generated from 270 simulation trials, multidimensional performance metrics are constructed to distinguish high- and low-performance caregivers. Subsequently, multiple analytical methods are employed to analyze these simulation data and reveal the mechanisms driving performance differences. The simulation results reveal significant floor-level differences in evacuation efficiency, indicating that the second floor (2F) provides the most effective initial deployment location under the spatial configuration examined in this research. Furthermore, a potential threshold-like pattern is observed: when the proportion reaches approximately 30%, caregivers’ cognitive and physical workloads are substantially alleviated within the observed experimental range. Spatiotemporal analysis of trajectories shows that high-performance caregivers adopt proactive call mobilization and systematic, low-tortuosity search paths. The Bayesian causal mediation model further shows that technical proficiency improves evacuation outcomes primarily through indirect pathways, particularly time-cost reduction and efficient return-to-assist organization. These findings provide quantitative and data-driven support for optimizing staffing strategies and emergency response planning, as well as for incorporating behavior-informed rules into evacuation simulation models. Collectively, they highlight the value of the proposed VR-based framework for evacuation behavior analysis and supporting emergency planning in nursing homes.
This study investigates the effect of misinformation on evacuee behaviour and evacuation outcome using an agent-based modelling approach. Through simulating emergency scenarios with varying proportions of misleading agents and different levels of gullibility of evacuating agents, we analyse the effect of misinformation on evacuation time, decision-making, and success rates. The model shows that even a small minority of misleading agents can significantly disrupt the behaviour of the majority, leading to poorer collective outcomes. The findings suggest the presence of threshold conditions under which disruption becomes more pronounced, highlighting a potential sensitivity of evacuation dynamics to misinformation within the simulated environment. Within the scope of this stylised model, the results provide qualitative insights into the interaction between misinformation processes and evacuation dynamics. The study focuses on isolating and examining underlying behavioural mechanisms in a controlled setting. The framework integrates concepts from misinformation propagation and evacuation modelling, and provides a basis for future work extending the analysis to more complex and realistic spatial environments.
Autonomous Vehicles (AVs) environments require task execution under continuously changing mobility, communication, and infrastructure conditions. Runtime offloading decisions depend on communication continuity, queue dynamics, resource availability, and execution feasibility during vehicle movement. Existing vehicular simulators commonly model mobility, wireless networking, or task offloading as separate subsystems, which limits unified evaluation of mobility aware vehicular computing and Artificial Intelligence (AI) driven scheduling behavior. This paper presents DriveNetSim, an integrated simulation framework for mobility aware and AI driven vehicular task offloading. The framework combines mobility aware communication and queue aware multi tier offloading within an integrated simulation. DriveNetSim supports Vehicular Edge (VE), Base Station (BS), Collaborative Vehicular Execution (Collaboration), Metro Edge (ME), and Cloud Computing (CC) tiers under heterogeneous Ultra Reliable Low Latency Communications (URLLC), Massive Machine Type Communications (mMTC), and Enhanced Mobile Broadband (eMBB) workloads. The simulator supports runtime feasibility analysis, collaborative execution control, oracle aligned diagnostics, regret analysis, and AI compatible scheduling evaluation. The simulator records queue dynamics, latency progression, mobility transitions, resource utilization, and scheduling behavior through reproducible runtime diagnostics and structured evaluation outputs. Experimental case studies demonstrate unified evaluation of mobility, communication, computation, collaboration, and AI driven scheduling within a reproducible vehicular computing environment.
Contact tracing, early diagnosis, and social distancing are crucial to contain outbreaks of infectious diseases, especially in the presence of asymptomatic infectives. Therefore, identifying the individuals at the highest risk of infectious disease and prioritizing them for testing is necessary to ensure both the public health and the cost-effectiveness of screening strategies. In this paper, we present an efficient quantitative approach to predict the spread of infectious diseases within a cluster by exploiting a stochastic model of disease evolution in an individual alongside observations of contacts, symptoms, and results of diagnostic tests. Specifically, we present an iterative solution technique to estimate the probability that a subject is infectious and not isolated over time given the observations acquired up to that time, achieving computational efficiency by disregarding dependencies among observations. We generated multiple synthetic data sets of observations, featuring different social network topologies, numbers of individuals, and densities of internal contacts, while accounting for noisy, erroneous, and missing observations. For each data set, we performed extensive experiments by varying the stochastic parameters of our approach from those used to derive a ground truth by performing stochastic simulation. Experimental results demonstrate that our approach effectively ranks subjects according to their probability of being infectious and not isolated, achieving high accuracy with respect to the ground truth. Notably, our approach outperforms an alternative baseline that accounts for dependencies among observations while maintaining a comparable runtime.
Accurate prediction of powder mechanical behavior is essential for pharmaceutical tablet manufacturing, yet conventional experimental campaigns and discrete element method (DEM) simulations remain costly and computationally intensive. This limits rapid formulation screening, parameter exploration, and simulation-assisted decision support. We present CompactAI, a structured deep-learning surrogate framework for DEM-based stress–strain prediction in pharmaceutical tablet compaction. CompactAI predicts loading and unloading responses under uniaxial compression from four physically interpretable DEM descriptors: loading stiffness, unloading stiffness, adhesion stiffness, and plasticity depth. The framework combines neural coefficient prediction, equation-based curve reconstruction, and collapse classification for catastrophic unloading cases associated with severe plastic deformation.We evaluate CompactAI on CompactAI-26, a DEM-generated dataset of 4796 tablet-compaction simulations, using held-out testing, five seeded train–validation–test splits, baseline comparisons, objective ablations, and curve-level error analysis. CompactAI achieves 0.1023 nMAE for loading reconstruction and 0.1949 nMAE for non-collapse unloading reconstruction on the original held-out test set, while producing end-to-end predictions in approximately 0.593 ms, compared with DEM runtimes of roughly 20 min to 2 h per simulation. Across the five seeded splits, CompactAI outperforms mean-curve, nearest-neighbor, and PCA–ridge baselines for loading reconstruction. Unloading reconstruction remains more challenging, with nearest-neighbor retrieval serving as a competitive local-interpolation baseline. These results validate one CompactAI surrogate instance for a single DEM configuration and illustrate a lightweight framework for instantiating configuration-specific surrogates from DEM-generated tablet-compaction datasets. CompactAI-26 is provided as a benchmark dataset for future work on predictive tablet-compaction simulation.
Securing real-time visibility in manufacturing is critical, yet conventional video surveillance, such as Closed-Circuit Television (CCTV), suffers from blind spots and a lack of semantic insight. To address these limitations, this paper proposes a Discrete Event Simulation (DES)-based Context-Aware Visual Digital Shadow (CAVDS). The CAVDS provides “augmented monitoring” capabilities through spatial, semantic, and temporal augmentations. We present a bi-modal DES framework comprising two distinct models: Simulation Mode Model (SMM) and Digital Shadow Mode Model (DSMM). This framework uniquely addresses the prevalent lack of model verification in existing digital twin studies by utilizing the SMM for rigorous structural verification prior to real-time deployment, while simultaneously minimizing development time and cost. Subsequently, the DSMM integrates with the physical system to provide validated augmented monitoring. To ensure ‘operational fidelity’, we propose a quantitative evaluation method using the Mean Absolute Percentage Error (MAPE) metric applied to both verification and validation purposes. The proposed framework is validated using a miniature automated production line. Experimental results demonstrate that: (1) augmented monitoring can be implemented efficiently using the bi-modal DES framework; (2) the systematic integration of SMM and DSMM enables rigorous verification and validation of the CAVDS; and (3) system configurations can be optimized to enhance physical synchronization performance. Ultimately, this study demonstrates the practical feasibility of the CAVDS system as an intelligent virtual observer that overcomes the physical constraints of traditional monitoring.
This study presents a hybrid peridynamics-finite element (PD-FEM) framework for fracture simulation of concrete beams in LS-DYNA, focusing on predictive accuracy and computational efficiency. Three-point bending tests with different notch positions were simulated, representing pure Mode I and mixed mode conditions. The peridynamic formulation was restricted to fractureprone regions, while the remaining domain was modeled using conventional FEM. The prototype microelastic brittle (PMB) material model was adopted for the peridynamic domain. The hybrid PD-FEM approach reproduced crack trajectories consistent with experimental observations and fully peridynamic results reported in the literature, particularly for Mode I dominated cases. For the pure Mode I configuration of gamma = 0 (KII/KI = 0), the predicted crack path matched the experimental result. under the mixed mode condition of gamma = 0.5 (KII/KI = 0.19), the predicted crack angle differed by only 1 degrees from the experiment, while under the higher mixed mode condition of gamma = 0.72 (KII/KI = 0.29), the deviation increased to 14 degrees This larger discrepancy is mainly attributed to the limitations of the PMB formulation in representing shear-dominated fracture behavior. Runtime analysis showed that, for the central notch configuration (gamma = 0), the hybrid PD-FEM model required 63% of the runtime of the fully FEM model and approximately 13% of that of the fully peridynamic model. For the other two cases, involving larger PD regions (gamma = 0.5 and 0.72), the runtime increased and exceeded that of the fully FEM model. However, in all three hybrid models, the computational cost remained much lower than that of the fully peridynamic approach. Therefore, the proposed framework provides a practical and computationally efficient alternative to fully peridynamic modeling for concrete fracture problems governed primarily by Mode I dominated crack propagation.
IoT-based Wireless Sensor Networks (WSNs) are vital, enabling real-time environmental monitoring with minimal human intervention at low cost. However, energy conservation and network longevity remain key challenges. The Low-Energy Adaptive Clustering Hierarchy (LEACH) protocol, although widely adopted, relies solely on random Cluster Head (CH) selection, which may select low-energy nodes far from the Base Station (BS) as CHs, leading to non-uniform CH distribution, uneven energy depletion, and premature node death. This work presents a simulation-based hierarchical clustering modeling framework implemented in NS-3 (version 3.43), centered on the proposed Hybrid Node Rank LEACH (Hybrid-NR-LEACH) to enable systematic evaluation of CH selection mechanisms under realistic wireless channel conditions. To mitigate randomness in CH selection, Hybrid-NR-LEACH incorporates normalized Node Rank (NR)-based CH selection, combining residual energy and distance metrics to balance randomness and determinism. In addition, we implement an Enhanced IBRE-LEACH (E-IBRE-LEACH) as an intermediate framework stage to support comparative evaluation. As a supporting analytical contribution, we investigate the necessity of hierarchical levels in LEACH by deriving an optimal transmission range, r0, using the Log-distance propagation loss model with device-specific parameters. Furthermore, to improve channel realism, we incorporate a Jakes propagation-loss model during the CH advertisement phase to capture multipath fading and temporal variations in RSSI. Simulation results on real-world datasets demonstrate that Hybrid-NR-LEACH achieves nearly 4 & times; higher Aggregation Compression Efficiency (ACE) and CH participation rate, along with nearly 25 & times; higher throughput than E-IBRE-LEACH, while maintaining competitive energy efficiency, network longevity, and high stability.
In this study, appropriate car-following models are selected to simulate the following behaviors of manual driving vehicles and intelligent driving vehicles. Based on the HighD dataset, the parameters of the manual driving vehicle are calibrated using the VISSIM-MATLAB joint simulation environment, and the parameters of the car-following model of the intelligent driving vehicle are calibrated using the OpenACC dataset and genetic algorithm. To realistically capture the operational characteristics of intelligent truck platooning (ITP) under mixed traffic conditions, a statebased platoon formation mechanism is incorporated to describe dynamic transitions among different vehicle roles. This study comprehensively analyzes the stability and safety of ITP and their impact on road capacity. Findings indicate that ITP maintains good stability at high speeds but experiences instability at low speeds. During the emergency braking situation, it is crucial to enhance the safety measures of ITP to reduce the risk of collision. Through the analysis of the fundamental diagram and time-space diagram, it is found that when the proportion of trucks is low and moderate, ITP can improve the road traffic capacity. Nevertheless, when the proportion of trucks and the proportion of intelligent driving trucks among them are high, they instead constrain road traffic capacity. Because these results are derived from a simulated environment with specific modeling abstractions, the conclusions should be interpreted as mechanism-oriented insights rather than direct quantitative predictions. However, the research results still provide theoretical support and practical references for improving the stability and safety of intelligent driving truck platooning.
Rotating machinery in industrial applications is often subjected to multiple simultaneous fault conditions, whose combined effects on system dynamics are not fully understood. This study investigates the nonlinear vibration response and instability of a rotor-bearing system supported by hydrodynamic journal bearings, considering the combined effects of bearing misalignment, rotor unbalance, and a transverse breathing crack. A high-fidelity model is developed that couples shaft vibration with nonlinear hydrodynamic bearing forces. The finite element method is employed to model the rotor-bearing system, and the shaft crack is modeled using the Strain Energy Release Rate (SERR) approach, combined with the Crack Closure Line technique, to account for crack breathing. Simultaneously, the nonlinear bearing forces are obtained from the solution of the Reynolds equation at each time step using the finite difference method. The dynamic response is calculated using the Newmark integration scheme. The results reveal how the combined presence of multiple fault conditions alters the dynamic behavior, amplifies super-harmonic components, and changes the instability threshold. Nonlinear vibration analyses, based on orbit plots, frequency spectra, Poincare maps, and waterfall diagrams, reveal that the interaction between multiple fault conditions produces characteristic dynamic signatures. The findings provide valuable insight into the interacting mechanisms between hydrodynamic lubrication phenomena, structural faults, and fluid-induced instabilities. This deeper understanding could support the development of more reliable fault diagnosis systems and contribute to informed design optimization of rotor-bearing systems operating under multi-fault conditions.
This paper investigates into machine learning models and transaction processing in NoSQL databases - an area that has yet to be researched. It designs and implements a novel integrated platform that leverages the power of machine learning in transaction processing in order to predict throughput, data consistency and the risk of data inconsistency in NOSQL databases. The platform erects a ML-based pipeline for data processing, features extraction, model training and execution behaviour of transactions in NoSQL databases. The platform simulates a real life application of London Bus Service in order to generate and define datasets and extract relevant features. It employs Spearman rank correlation method to identify correlation between transaction features. It implements two widely used machine learning models, K-fold CV and KNN, that predict throughput, data consistency and the risk of data inconsistency. The proposed platform is evaluated through a series of experiments. The experiments evaluate the performance and accuracy of the machine learning models. The results provide useful insights into the execution of transactions and the capability of NoSQL databases. Knowing throughput, in advance, can assist in identifying the capacity of a NoSQL database in relation to processing of big data under different conditions, e.g., failure and recovery. Moreover, knowing the risk of data inconsistency in advance can help in preventing database from going into inconsistent state.
People face numerous security risks in the era of modern digital advancement due to the existence of escalating network viruses in communication systems of information technology that become more challenging with the proliferation of the Internet. Effectively detecting and defending against network viruses is always an essential and critical concern in the design of a protected computer security environment. In this communication, the distinct structure of a group of devices connected to the server responsible for replicating the viruses is investigated. The dynamical response patterns of virus proliferation are examined through the implementation of intelligent computational knowledge-driven predictive networks. The nonlinear differential system is characterized by the uninfected, internally, and externally infected states of the computers. The reference solutions for studying the dynamics of all three classes in the model across different attack routes are determined for various scenarios by varying detaching and recruiting rates for both old and new computers, the rate of infection-free computers, the rate of internally affected computers, and the rate of external computers with antivirus capability employing the RungeKutta method. The designed methodology is based on multilayer feed-forward networks optimized by means of the Bayesian regularization technique. The generated data are utilized as inputs and targets for training, and testing samples to create an approximate predictive model for all classes in the proposed nonlinear system. Comprehensive simulations are performed to measure the proposed model's accuracy in sophisticated persistence virus attacks that target critical network routes. The presented results closely match the observed data on the basis of mean squared error convergence, histograms, and analysis of regression indices that validate the efficacy, stability, and significance of the proposed approach. The proposed NNBR methodology is further compared with advanced machine learning and deep learning techniques, and the results demonstrate the optimum performance of the presented technique over all the competing methods.
The rapid expansion of Internet of Things (IoT) and data-intensive applications has accelerated the evolution of the compute continuum, where edge and cloud resources collaboratively execute complex workflows. Efficient workflow scheduling in such environments is challenging due to heterogeneous resources, dynamic workloads, communication delays, and varying security risks across distributed infrastructure. Most existing scheduling approaches primarily optimize performance metrics such as latency or makespan while overlooking the security implications associated with task placement decisions. To address these challenges, this paper proposes a Security-Risk-Aware Federated Multi-Agent Reinforcement Learning framework with Decentralized Execution (SR-DMARL) for workflow scheduling in compute-continuum environments. Here, decentralization refers to the execution phase, where agents make local scheduling decisions without a centralized scheduler, while training uses periodic FedAvg-based parameter synchronization. The novelty of SR-DMARL lies in three integrated mechanisms: a hierarchical task-node security compliance model, a communication-exposure-aware workflow risk formulation, and a decentralized credit-assignment mechanism that links each agent's policy update to its local contribution to security and scheduling performance. The scheduling problem is formulated as a multi-objective optimization task that jointly minimizes workflow completion time, communication overhead, and security risk while improving task success rate and resource utilization. Unlike generic MARL-based schedulers that rely mainly on shared global rewards or performance-oriented objectives, SR-DMARL embeds security compliance, sensitive-task exposure, and cross-node communication risk directly into the state, reward, and local credit-assignment process. Extensive simulation experiments demonstrate that the proposed approach improves scheduling efficiency and reduces security risk compared with existing baseline methods. In the security-performance trade-off analysis, SR-DMARL achieves a makespan of 235 s, a security-risk score of 95, and resource utilization of 91%, compared with 455 s, 315, and 63%, respectively, for the weakest baseline. Under workload scaling, SR-DMARL maintains communication overhead around 22 MB at 500 tasks and reaches approximately 97% resource utilization. Statistical analysis using Friedman and Wilcoxon signed-rank tests further confirms that the observed improvements are significant at alpha = 0.05. These results indicate that SR-DMARL provides an effective and scalable mechanism for security-risk-aware workflow orchestration across heterogeneous edge-cloud compute-continuum environments.
