The persistent trend of global urbanisation presents a dual challenge: accommodating population growth while addressing significant environmental and social pressures. In response, the 'smart city' concept, underpinned by the Internet of Things (IoT), has become a leading vision for future urban development. This literature review systematically synthesises and critically assesses the existing academic discourse on the relationship between IoT-enabled smart city solutions and urban sustainability. It begins by establishing the conceptual foundations, exploring the evolution of urban sustainability and smart city paradigms, and positioning IoT as a vital enabler of infrastructure. The review then thematically examines the application of IoT across key urban sectors, such as energy, water, mobility, waste, and the built environment, analysing contributions towards sustainability goals, including resource efficiency and emissions reduction. Moving forward, the review also scrutinises a broad body of critical literature, highlighting ongoing challenges related to techno-solutionism, data governance, social equity, and barriers to implementation. Through this synthesis, a notable research gap emerges: a deficiency of integrated, socio-technical frameworks that guide the deployment of IoT solutions to ensure they achieve verifiable sustainability outcomes in an equitable way. The review concludes by emphasising the necessity for future research to shift focus from technological potential assessments to empirical studies of real-world implementation processes and comprehensive impact evaluations.
The increasing penetration of grid-connected photovoltaic (PV) systems introduces significant power quality challenges due to fast irradiance variations, intermittent generation, and nonlinear load interactions. Conventional control and signal-processing-based power quality enhancement techniques often suffer from limited response speed and reduced effectiveness during high-frequency transients. This paper proposes a Spiking Neural Network (SNN)-based fast transient detection and control framework for power quality improvement in grid-connected PV systems. The proposed approach employs an event-driven neuromorphic SNN to detect voltage and current transients with ultra-low latency, enabling rapid identification of harmonics, voltage sags, swells, and sudden load disturbances. Unlike traditional artificial neural networks, the SNN processes information in the form of discrete spikes, significantly reducing computational complexity and enhancing real-time responsiveness. The detected transient features are directly integrated with a distributed active power filter control strategy to generate adaptive compensating current references. Simulation studies carried out in MATLAB/Simulink under varying irradiance, nonlinear load, and grid disturbance conditions demonstrate that the proposed SNN-based controller achieves faster transient detection, lower total harmonic distortion, and improved power factor compared to conventional PI- and ANN-based controllers. The results confirm the effectiveness of neuromorphic intelligence in enhancing dynamic power quality performance, making the proposed method a promising solution for next-generation smart PV-integrated power systems.
Accurate segmentation of brain tumours from magnetic resonance imaging (MRI) depends critically on the quality and consistency of acquired scans; however, MRI data are often affected by artefacts, intensity inhomogeneity, and inter-scanner variability, which degrade the performance of deep learning–based models. To address this, the present study proposes a denoising-guided enhancement pipeline designed to improve anatomical and geometric consistency before segmentation. The preprocessing framework integrates Gibbs ringing artefact suppression, N4ITK bias field correction, Z-score–based intensity normalisation, adaptive histogram equalisation, along with spatial alignment and resolution standardisation, applied in an empirically optimised sequence across multimodal MRI inputs (T1, T1c, T2, FLAIR). Tumour segmentation is performed using an Attention Residual U-Net (ARU-Net), enabling improved feature localisation and boundary delineation. Experimental evaluation on BraTS 2015 and BraTS 2018 datasets demonstrates enhanced performance, achieving Dice scores of 0.93 for whole tumour, 0.85 for tumour core, and 0.74 for enhancing tumour. Ablation analysis confirms that both the inclusion and ordering of preprocessing steps significantly influence segmentation accuracy, while statistical validation using corrected resampled t-tests establishes the robustness and significance of the observed improvements. The proposed pipeline is modular, reproducible, and suitable for integration into clinical imaging workflows, where denoising-guided preprocessing ensures early removal of acquisition-related artefacts without compromising fine anatomical details.
The squeeze film behaviour of poroelastic bearings with rough surfaces and couple stress fluids is studied using a simplified model in which the action of the couple stress synovial fluid in lubricating the hip joint is examined. The articular cartilage. The layer is modelled as a biphasic poroelastic matrix material. A modified average Reynolds equation is derived, which accounts for the couple stress effects, random surface roughness, and the elastic nature of the cartilage-bearing surface. Two types of one-dimensional random roughness patterns, longitudinal roughness and transverse roughness, are presented using Christensen's stochastic theory. By using a domain transformation, the reduced governing equations can be mapped onto the unit square and solved numerically using the sine-cosine wavelet operational matrix of integration method. Uniqueness, uniform convergence, convergence of the partial sums to the exact function, and commutation of the integration and limit operations are guaranteed by proving some properties of the wavelet approximation. The numerical results indicate that, although couple stresses can improve the performance of the joint as a whole, the effect on the squeeze film performance of surface roughness must be considered, depending on the patterns of surface roughness. The wavelet-based method proposed in this paper is accurate and efficient.
This paper presents the design, optimization, and experimental validation of a triband star-shaped microstrip patch antenna for Ku-band applications. The design process begins with a single star-shaped patch antenna, which is progressively improved and extended to form a multiple-input multiple-output (MIMO) configuration. The final optimized MIMO antenna operates within 12–16 GHz, achieving resonances at 12.3 GHz, 14.2 GHz, and 15 GHz. The star-shaped geometry is employed to enhance the impedance bandwidth and improve gain performance. Simulated and measured results show good agreement in terms of return loss, impedance bandwidth, voltage standing wave ratio (VSWR), and radiation patterns. Moreover, diversity performance metrics, including the envelope correlation coefficient (ECC) and diversity gain (DG), confirm low correlation and near-ideal diversity across the three operating bands. The proposed triband MIMO antenna exhibits stable radiation performance and satisfactory gain, making it a strong candidate for 5G and satellite communication systems in the Ku-band.
Network protection relies on machine learning-based systems that detect intrusions. The detection systems lose their effectiveness because they use multiple duplicate features, and their performance depends on the specific network traffic patterns and system operational requirements, which prevent real-time functioning. The research presents a PSO-DBF intrusion detection framework, which begins with Distributional Boosting Forest (DBF) as its first step to create two groups (C1 and C2) that display similar probabilistic characteristics through network traffic clustering. The research team uses Particle Swarm Optimization (PSO) to process each cluster when they complete their clustering process because the method helps them find the most valuable network attributes, which decrease feature duplication while enhancing the ability to distinguish different features. K-Nearest Neighbors (KNN) provides the best performance when conventional machine learning classifiers use optimized feature subsets for intrusion detection. The proposed framework demonstrated its efficiency through experiments that utilized recognized IDS datasets. PSO removed almost 50% of the initial features while keeping 18 features from NSL-KDD and 21 features from UNSW-NB15, achieving reduction rates of approximately 56 percent and 57 percent. The proposed PSO–DBF with KNN framework achieved 99.36% accuracy on NSL-KDD and 99.89% accuracy on UNSW-NB15, exceeding the performance of Support Vector Machine (SVM), Naive Bayes (NB), Quadratic Discriminant Analysis (QDA), deep neural models, and recent hybrid metaheuristic-based IDS frameworks. The main improvement of the proposed method comes from its ability to reduce detection times, which drop from 0.44 milliseconds to 0.29 milliseconds. The DBF-PSO framework achieves its optimal performance for intrusion detection in enterprise cloud and edge-network security environments because of its detection accuracy and energy efficiency.
This study develops a broad extension of logarithmic series and presents exact formulas for their sums. By reformulating the series through suitable integral and functional representations, the work uncovers direct links between these generalized series and polylogarithmic functions. The approach yields several transformation identities that streamline the evaluation of such series and reveal a unified structure underlying many classical logarithmic and alternating forms. Illustrative special cases and numerical checks highlight the accuracy and versatility of the derived results, demonstrating their usefulness in analytic methods and computational applications
Hyperparameter tuning remains a major computational challenge in the field of machine learning. Traditional methods (grid search, random search, Bayesian optimization) are constrained by high dimensionality and complex parameter dependencies. This article explores constraint programming (CP) as a promising alternative, leveraging its ability to handle complex constraints and efficiently reduce the search space. We systematically compare CP methods to standard methods across different data types and learning algorithms. Performance metrics include accuracy, computational efficiency, convergence time, and the number of required evaluations. The results highlight the superior advantages of CP for complex hyperparameter dependencies and constrained search spaces, while also identifying scenarios where traditional methods remain preferable. This study contributes to the field of Automated Machine Learning (AutoML) and provides concrete recommendations for hyperparameter tuning.
This study presents a transient thermo-mechanical finite element framework for high-speed end milling of AISI 4340 steel. The model couples moving heat sources, rate- and temperature-dependent plasticity, and adaptive mesh refinement (AMR) triggered by temperature gradient, plastic strain rate, and contact pressure. It is integrated with a design of experiments/response surface methodology using cutting speed (VC), feed per tooth (fZ), radial depth of cut/width of cut (ae), axial depth of cut (ap), and coolant mode. Responses include peak interface temperature per tooth (Tpeak), predicted surface residual stress (?_xx^"surf" ), and depth of compressive residual stress layer (dcomp). Experiments provide X-ray diffraction-based surface/depth profiles and arithmetic mean surface roughness (Ra). AMR is applied in this study to minimize the cut compute cost by 41-52% and error by 35-45%. Across 12 validation cuts, root mean square errors were 24 °C of Tpeak, 33 MPa of ?_xx^"surf" , 0.07 µm of Ra, and 22 MPa of dcomp. The response surface methodology and analysis of variance identified VC as the main driver of thermal load, while fZ, ae, and ap controlled the sign and depth of the residual field; coolant modified heat partition. Multi-objective desirability optimization with a material removal rate constraint yielded a balanced minimum quantity lubrication. Overall, exit-edge cooling and subsurface plasticity jointly set residual sign and magnitude; AMR is essential to resolve these gradients efficiently. The framework offers a reproducible route for residual stress-aware process planning in fatigue-critical AISI 4340 components while preserving throughput and is readily transferable to allied high-strength steels.
The present study examines the impact of hydrogen induction on the performance and emission attributes of a Reactivity Controlled Compression Ignition (RCCI) Engine operating on a microalgae biodiesel–diesel blend (B20D80) with a constant injection timing of 23° BTDC and an injection pressure of 200 bar. Experiments were conducted with hydrogen induction at flow rates of 3, 6, and 9 lpm, referred to as B20D80 + H? 3 lpm, B20D80 + H? 6 lpm, and B20D80 + H? 9 lpm, respectively. Among the tested fuel combinations, the B20D80 + H? 9 lpm blend showed improved performance under these fixed injection conditions, achieving a 20.8% enhancement in brake thermal efficiency and a 28.1% decrease in brake specific fuel consumption compared to conventional diesel operation. Emission analysis indicated that hydrogen enrichment led to substantial reductions in major pollutants, carbon monoxide, and smoke opacity, decreasing by 24% and 25%, while carbon dioxide and hydrocarbon emissions were reduced by around 8.6% and 35%, owing to the carbon-free nature of hydrogen and the oxygenated structure of biodiesel. However, nitrogen oxide emissions increased moderately by 22.8%, which is due to higher in-cylinder temperatures resulting from enhanced combustion. Overall, the results demonstrate that hydrogen-assisted microalgae biodiesel operation significantly improves combustion efficiency while effectively reducing most exhaust emissions, highlighting its viability as a cleaner, more efficient dual-fuel strategy for RCCI engines.
New developments in the Brain-Computer Interface (BCI) technology have increased the rate at which research has been done on precise and quick electroencephalography (EEG)-based signal classification models. This review analyses new trends, procedures, problems, and gaps in research on EEG signal classification in three large cognitive paradigms: Steady-State Visual Evoked Potential (SSVEP), detection of the attention focus, and motor imagery (MI). These paradigms form the focus of real-time BCI applications, e.g., assistive technologies, neurorehabilitation, adaptive learning, and augmented interaction systems. The analysis presented in the paper on the development of the traditional machine learning (ML) and the modern deep learning (DL) models of the EEG interpretation systematically reviews the progression of the original ideas in the EEG interpretation field. Power spectral density analysis, Common Spatial Patterns (CSP), wavelet transform, and empirical mode decomposition (EMD) techniques of feature extraction, and Support Vector Machine (SVM), Random Forest (RF), Convolutional Neural Network (CNN), and Long Short-Term Memory (LSTM) techniques are critically examined. Some of the performance evaluation metrics that are widely employed in the literature are also addressed. Special attention is paid to the real-life issues that accompany real-world EEG data, such as low signal-to-noise ratio, artifact contamination, inter-subject variability, limited diversity of datasets, and bad model interpretability. It is believed that such public benchmark datasets as BCI Competition datasets, PhysioNet, and other multi-subject repositories can be used to support comparative analysis. Additional requirements of unified evaluation frameworks, real-time system-aware assessment, hybrid models, multimodal fusion strategies, transfer learning, and explainable AI have been identified in the review in an attempt to enhance the accuracy, robustness, and trustworthiness of EEG-based cognitive systems. On the whole, the given study can be used as a consolidated basis for the creation of future-generation EEG-based BCI frameworks.
In this work, we use the Mann iteration process rather than the conventional Picard operator to extend fixed point findings in G-metric spaces. Mann iteration is known to provide better convergence properties and stability in fixed point approximations, particularly in cases where Picard iteration fails due to weak contractive conditions. We present a new family of?rational-type contractive conditions and prove the existence and uniqueness of fixed points of single-valued mappings in G-complete G metric spaces. Specifically, we improve upon existing theorems in the literature both by generalizing their?statements as well as strengthening their use through an improved iterative scheme.
Background: The growing need for secure communication on resource-constrained systems, such as those in the Internet of Things (IoT), has led to a significant increase in demand for lightweight symmetric ciphers. Nevertheless, different techniques and implementations exist, making the selection of the optimal security solution for a particular application challenging. Objective: This study primarily focuses on implementing an optimised Lightweight Advanced Encryption Standard (LAES) algorithm in hardware to address the critical need for energy-efficient security solutions for IoT devices. Methods: This study implements LAES to meet the security requirements for IoT devices. The Kintex 7 and Spartan 7 FPGAs (Field Programmable Gate Arrays) are utilised for implementation, with critical performance metrics such as hardware area utilisation used to evaluate performance. The algorithm eliminates the computationally expensive MixColumns operation from standard AES while maintaining essential security transformations. Performance evaluation focused on hardware resource utilisation (LUTs, FFs, IO) and power consumption across clock frequencies ranging from 1 ns to 20 ns. Results: The results indicate significant advancements in achieving area and power-efficient designs. Our findings show that the reduction in power consumption is by 95.29% and 92.07% as compared to existing models. The area consumption, such as LUTs, FFs, and IO, has also been significantly decreased compared to existing models. Conclusions: The proposed LAES architecture demonstrates that strategic algorithm optimisation can yield substantial improvements in both power efficiency and hardware utilisation without compromising security, making it highly suitable for IoT deployment.
Wireless Sensor Networks (WSNs) are critical to modern IoT applications, yet their deployment is often constrained by limited energy, dynamic topologies, security vulnerabilities, and stringent Quality-of-Service (QoS) requirements. While existing approaches frequently address these challenges in isolation, this paper introduces a holistic routing framework that synergistically integrates an Improved Fuzzy Logic System (IFLS) with Particle Swarm Optimization (PSO) to balance multiple performance metrics in real time. Our hybrid model dynamically tunes routing parameters and fuzzy rules based on network state—including energy levels, congestion, node density, mobility, and security threats—thereby optimizing cluster-head selection, path stability, and trust-aware communication in UAV-assisted WSNs. Extensive simulations demonstrate that the proposed system achieves a 94.2% packet delivery ratio, reduces energy consumption by 48%, and extends network lifetime by 97% compared to contemporary fuzzy-based and trust-aware routing protocols. The work thus offers a scalable, adaptive, and energy-efficient routing solution suitable for large-scale, resource-constrained, and mobility-prone sensor networks. We also provide complete algorithmic specifications and reproducible simulation setups to facilitate validation and further research.
The comprehensive study investigates the application of cutting-edge machine learning algorithms and advanced image processing techniques for the early detection of lumpy skin disease in cattle. The proposed robust analytical framework that evaluates multiple predictive models using comprehensive performance metrics, including F1 scores ranging from 0.87 to 0.97, precision up to 0.984, recall up to 0.963, and accuracy peaking at 97.77%. The novel approach incorporates pixel-level analysis to quantify disease severity through the ratio of affected to healthy tissue, complemented by processing speed delays between 5.54ms and 20.95ms. The research demonstrates significant improvements over traditional diagnostic methods, with particular emphasis on the model's ability to identify high-risk cases requiring immediate intervention. These findings have substantial implications for veterinary medicine, agricultural technology development, and livestock management policies, potentially revolutionizing disease surveillance systems in the agricultural sector.
This study investigates the issues of energy usage in multipath wireless networks utilizing the Multipath Transport Control Protocol (MPTCP) under application-level timing knobs implemented in socket logic, which allows numerous TCP connections via different pathways. Due to route heterogeneity, MPTCP consumes more energy. Currently, many research works have provided several techniques to optimize energy efficiency; however, they focused on individual systems rather than total performance. This work proposed a stochastic multipath scheduling technique that considers the fluctuations in data transmission rate and path capacity. The scheduling mechanism is associated with the optimization problem to achieve the objectives of maximizing throughput, avoiding congestion, and improving stability. An algorithm is developed to solve multipath data transmission issues by utilizing the drift-based constraints. Simulations are performed to generate results for the comparison of three different optimized MPTCP schemes in the application layer with baseline and conventional protocols. The results are showing considerable improvements in throughput and end-to-end latency
Drying is a critical post-harvest operation for agricultural and herbal products, but remains energy-intensive and performance-limited under humid tropical conditions. This study experimentally investigates the energy performance and drying kinetics of a photovoltaic–thermal (PVT) solar dryer integrated with porous absorber materials to enhance heat and mass transfer. The developed system consists of a 150 W PVT panel, a double-pass solar collector, and porous steel mesh absorbers with porosities of 0.98, 0.97, and 0.96. Experiments were conducted under tropical climatic conditions in Thailand at a low air velocity of 0.07 m/s using kaffir lime leaves as a representative leafy material. The results show that drying occurred entirely in the falling-rate period and was governed by internal moisture diffusion. Porous integration improved thermal uniformity and airflow turbulence, increasing the drying rate by 43.73–56.01% and reducing specific energy consumption (SEC) by 30.44–36.01% compared with the non-porous configuration. Optimal performance was achieved at a porosity of 0.96, yielding the highest Deff (8.59x10-13 m²/s) and the lowest SEC of 21.84 MJ/kg. Thin-layer analysis confirmed that the Page model best described the drying kinetics (R2 = 0.980–0.999). Statistical analysis (ANOVA) verified that porosity is a dominant factor influencing performance (p < 0.01). Overall, the integration of porous materials into a PVT solar dryer significantly enhances drying performance and represents an energy-efficient approach for drying leafy agricultural products in humid tropical regions.
Introduction: The widespread adoption of Internet of Things (IoT) devices has transformed multiple industries, enhancing operational efficiency and convenience. However, the rapid expansion of IoT ecosystems also brings forth significant security challenges. Traditional security frameworks often fail to adequately protect these systems due to their large scale, diversity, and limited resources. In response, cloud-based security solutions have emerged as a promising alternative, offering centralized management, advanced authentication techniques, and real-time threat monitoring. Problem Statement: IoT environments are vulnerable to various security risks, including unauthorized access, data breaches, and device manipulation. Existing security mechanisms often fall short when it comes to defending against sophisticated cyber-attacks targeting IoT devices and networks. The resource-constrained nature of many IoT devices further limits the implementation of robust local security measures. As a result, there is an urgent need for effective, cloud-based security solutions designed specifically for the unique demands of IoT systems. Objective: This research aims to explore the effectiveness of cloud-based security solutions in mitigating the security challenges faced by IoT environments and devices. The study focuses on evaluating the performance of cloud-based authentication mechanisms, intrusion detection systems, and encryption techniques in strengthening the security and privacy of IoT ecosystems. Methodology: A comprehensive approach is employed, combining a literature review, case studies, and empirical research to assess the current landscape of IoT security in smart environments. Data collection includes unstructured interviews with industry experts and stakeholders, offering insights into current practices and emerging security trends. The research framework incorporates threat modeling, risk assessments, and the development of proactive security strategies. Results: Initial findings indicate that cloud-based security solutions offer several benefits for protecting IoT environments and devices. Centralized management enhances integration and scalability, while advanced authentication methods, such as multi-factor and biometric authentication, improve access control. Real-time threat detection and response capabilities further bolster security by enabling timely interventions to prevent breaches and attacks. Conclusion: Cloud-based security solutions present a highly effective approach to addressing the unique security concerns of IoT environments and devices. By leveraging the scalability, flexibility, and computational power of cloud platforms, organizations can enhance the resilience of their IoT deployments against evolving cyber threats. However, further research is needed to optimize cloud-based security tools to better serve diverse IoT applications and use cases.
Introduction: The rapid expansion of Internet of Things (IoT) devices, combined with the widespread adoption of cloud computing, has led to an interconnected digital environment. However, the lack of interoperability and standardization between IoT and cloud systems presents significant challenges. This study explores the importance of addressing interoperability issues and establishing standardized practices to facilitate smoother integration between these technologies. Problem Statement: The diverse range of IoT devices and cloud platforms has created a fragmented ecosystem where interoperability challenges impede the seamless exchange of data and functionality. The absence of universally accepted protocols further complicates compatibility, leading to performance inefficiencies, increased development complexity, and potential security risks. Resolving these issues is essential for unlocking the full capabilities of IoT and cloud integration. Objective: The goal of this research is to examine the current state of interoperability and standardization in IoT and cloud integration. The study aims to identify existing challenges, evaluate current standards and protocols, and propose solutions to enhance interoperability and standardization, fostering a more cohesive and efficient integration between these technologies. Methodology: This study uses a multi-method approach, including a thorough literature review, case studies of existing IoT and cloud integration efforts, and unstructured interviews with industry experts. The analysis focuses on identifying recurring interoperability challenges, evaluating the effectiveness of existing standards, and reviewing successful integration strategies. This comprehensive approach provides a detailed understanding of the complexities surrounding IoT and cloud interoperability. Results: The research identifies key challenges affecting the interoperability of IoT and cloud systems. Through detailed analysis of current standards and successful integration cases, the study offers insights into effective strategies for overcoming these barriers. The results provide actionable recommendations for enhancing interoperability and achieving smoother integration across various IoT and cloud environments. Conclusion: This study emphasizes the urgent need for improved interoperability and standardization in IoT and cloud integration. The research findings, along with proposed solutions, offer valuable direction for industry professionals, policymakers, and researchers working towards creating a more interconnected and efficient digital ecosystem. As IoT and cloud technologies continue to advance, establishing strong, standardized frameworks is crucial to realizing the full potential of these transformative technologies.
Modern Industrial Control Systems (ICS) and Supervisory Control and Data Acquisition (SCADA) networks face a growing class of logic-layer attacks in which adversaries silently manipulate configuration or project files instead of deploying traditional malware. Existing defences, such as network intrusion de-tection systems and machine-learning-based anomaly detectors, struggle to ob-serve these pre-deployment logic changes and often incur high operational complexity. This paper presents a lightweight, host-based framework that uses YARA, a rule-based pattern-matching engine, to perform static inspection of XML configuration files generated by SCADA engineering tools. The proposed system is implemented on a Windows 10 engineering workstation using Mod-busPal as a Modbus TCP simulator, Python for file monitoring and GUI devel-opment, and YARA CLI/Python bindings for rule execution. Custom YARA rules are crafted to detect unauthorized Modbus function code 5 (Write Single Coil) operations targeting critical coil addresses, modelling malicious logic injections such as covert actuator activations. In a controlled lab environment, using a va-riety of ModbusPal project files, a combination of benign (no infiltration) and tampered project files, as well as our detection framework, achieved less than 200 milliseconds of latency for detecting true positives (and 0 false positives and 0 false negatives) for the defined ruleset and under a negligible resource over-head. These findings indicate that static logic validation at the host-level would fulfil an effective integrity pre-deployment check for PLC logic in addition to current network-based and behaviour-based ICS security mechanisms, without requiring modification of the installed PLC hardware and network protocol.
