Mixing parameters significantly influence asphalt-mixture performance, yet mixing speed, time, and temperature are commonly selected from general practice rather than optimized against performance criteria. This study evaluated the effects of these parameters on Marshall stability and semi-circular bending (SCB) fracture load. A balanced 3 × 2 × 2 full-factorial design was used, comprising three temperatures (135, 150, and 165 °C), two mixing times (2 and 7 min), and two mixing speeds (1500 and 3000 rpm), with three replicate batches per treatment. The optimum binder content was 3.5% of the total mixture mass, as determined using the Marshall mix design method. The best overall performance occurred at 165 °C, 7 min, and 3000 rpm, yielding a Marshall stability of 12.9 kN and an SCB fracture load of 57.5 kN. Factorial ANOVA showed that mixing temperature, time, and speed significantly affected both responses (p < 0.001). Exploratory machine-learning models were developed to predict performance from the mixing parameters. Random Forest outperformed Gradient Boosting, with R² values of 0.794 for Marshall stability and 0.828 for SCB fracture load. The results demonstrate that coordinated optimization of mixing temperature, time, and speed can improve asphalt-mixture stability and fracture resistance, while data-driven models can support laboratory-scale protocol selection.
Large language models have transformed conversational artificial intelligence; however, their performance remains weak for languages with extreme resource scarcity. Brahui is a typologically distinct Dravidian language with limited annotated data and minimal representation in multilingual resources. This study presents a transformer-based chatbot architecture designed for Brahui and introduces an embedding fusion mechanism to address data scarcity. The mechanism combines multilingual embeddings with Brahui-specific embeddings, enabling the model to retain cross-lingual knowledge while learning language-specific semantic and morphological features. Experiments show substantial improvement over the baseline: semantic similarity increases from 79.91 to 98.23, perplexity decreases from 94.27 to 14.64, and the system achieves 76% accuracy in correctly answering questions. The proposed framework provides a robust and parameter-efficient approach for developing fluent generative dialogue systems in severely under-resourced linguistic contexts.
The increased adoption of renewable energy sources in DC microgrids has improved the sustainability and has also posed some challenges associated with the effective coordination of converters and minimization of power losses. Uncoordinated operation of the Modular Multilevel Converters (MMCs) can result in higher conduction and switching losses especially those in high-voltage DC applications. The overall goal of this paper is to work out an integrated energy management system that reduces the overall converter losses and provides equal work of MMC-based DC microgrids. A Particle Swarm Optimization (PSO)-based approach is suggested to meet this goal by optimal distribution of converter output currents and submodule utilization ratio within the operating limits set in steady-state operation. An equivalent resistance converter loss model, which includes a constant loss element, constant loss elements and factors of module utilization are created and integrated into the optimization problem. The suggested framework is run in MATLAB and tested in an experimental circuit of a DC micro-grid including grid, photovoltaic, and battery-interfaced MMCs. It has been shown in simulation that the PSO algorithm converges rapidly and total system losses are reduced by about 2.28 kW over the conventional uncoordinated operation, but does not decrease the DC bus power balance or violate operational limits. The findings verify that converter-aware optimization has the potential of a great contribution towards energy efficiency and coordination of MMC-based DC microgrids, and forms a computationally-efficient basis on which to build future extensions to adaptive and decentralized approaches to control.
Smart cities are increasingly leveraging technology to improve urban living and meet the growing demands of their populations. The adoption of innovative technologies, such as mobile-based smart urban mobility services, plays a pivotal role in this transformation. This research aims to apply the Technology Acceptance Model (TAM) to examine the key factors influencing citizens' intentions to adopt mobile-based mobility service apps in Lahore, Pakistan. A paper-based survey was conducted across six neighborhoods representing high-, medium-, and low-income groups, yielding 500 responses. Structural Equation Modelling (SEM) via SmartPLS4 indicates that Perceived Usefulness, Perceived Ease of Use, Habit, and Perceived Security Risk collectively explain 74.7% (R2 = 0.747) of user intention of citizens to adopt Smart Mobility Services (e.g., Mobile Application-based ride hailing) with all paths statistically significant (p < 0.001). Furthermore, the relationships among the variables have been explained using structural equation modelling (SEM). Based on these findings, this study provides valuable insights into the proposed approach to build smart citizens and increase well-being and quality of life. Findings provide useful insights for developing marketing strategies to improve the uptake of mobile-based mobility service applications among citizens.
This study develops and empirically evaluates an Adaptive Gamification Design Model (AGDM) to address programming learning difficulties (PLDs). A three-phase mixed-methods design was employed. Phase I used a PRISMA-guided systematic review of studies published from 2010 to 2025 (N = 112) to construct a multidimensional Programming Learning Difficulty Taxonomy (PLDT) encompassing cognitive, affective, and instructional challenges. In Phase II, the taxonomy was validated through exploratory and confirmatory factor analyses of survey data from 842 undergraduate computer science and software engineering students at four public universities. Phase III comprised a 14-week quasi-experimental intervention involving a control group (n = 142) and an adaptive-gamification group (n = 144). Cognitive load, syntax anxiety, and self-efficacy deficit significantly predicted course failure. Compared with traditional instruction, the AGDM environment produced higher academic performance and intrinsic motivation and reduced the dropout rate from 34% to 12%. Structural equation modeling indicated that engagement and self-efficacy mediated the relationship between adaptive gamification and academic performance. The framework is operationalized through the Adaptive Gamification Optimization Algorithm (AGOA), which dynamically adjusts task difficulty, feedback scaffolding, and motivational incentives according to each learner's PLDT profile. The study contributes a validated taxonomy and a scalable adaptive-gamification framework that can be integrated into computing curricula to support competence, retention, and personalized learning.
This research paper describes the detailed design of a hybrid system comprising of Light Fidelity (LiFi) communication and Radio Frequency energy harvesting (RFEH) to serve the Internet of Things (IoT) devices. The system developed is aimed at resolving the essential concerns of the viable transmission of data, energy efficiency, and sustainable functioning of extensive IoT systems. This study tested the main performance indicators, including the Signal-to-Noise Ratio (SNR), Bit Error Rate (BER), energy harvesting (EH) efficiency, and power sharing strategies of several IoT devices, through MATLAB simulations. Solar EH also promotes the ability of the system to satisfy the requirements of the IoT power needs. The comparative analysis to industry standards shows significant performance improvements, which makes the system a sustainable solution in the case of IoT implementation in energy-limited environments. The results validate the design by achieving improved BER performance, enhanced data rate adaptability, and reliable IoT operation under varying energy conditions.
Ensuring the mechanical integrity and service life of rolling-element bearings requires a clear understanding of contact stresses under operational loading. This study presents a finite element analysis of a deep-groove ball bearing in ANSYS, where the load is transferred from the shaft through the rolling elements to the outer race. The Goodman and Soderberg mean-stress criteria were integrated with the contact model to compare fatigue predictions. A mesh-independence study used 10,000, 25,000, and 70,000 elements. At a radial pressure of 100 MPa, the maximum von Mises stress was approximately 496 MPa, and local factor-of-safety values were below 1.0, indicating a risk of local yielding and premature fatigue failure. Reducing the pressure to 10 MPa increased the minimum Goodman factor of safety to 1.73 and extended the predicted fatigue life to 1 × 10⁶ cycles. The Soderberg criterion consistently produced lower safety factors and shorter life estimates than the Goodman criterion under the same loading conditions. The integrated framework generates spatially resolved stress, safety-factor, and fatigue-life maps that can support bearing-design optimisation and predictive maintenance of rotating machinery.
Reclaimed Asphalt Pavement (RAP) offers a practical route to reduce material consumption and construction cost in asphalt paving, yet its adoption in Pakistan remains constrained by limited local performance–cost evidence. This study investigated the mechanical and economic feasibility of RAP-incorporated hot mix asphalt using locally sourced materials and market-based cost data. Mixtures containing 0%, 25%, 50%, and 75% RAP were designed using the Marshall method, and their performance was evaluated through volumetric analysis, Marshall stability, and semi-circular bending (SCB) peak load testing. Economic assessment was conducted using contractor quotations cross-checked with the Market Rate System (MRS-2024). The optimum binder content was 4.40%. Among all mixtures, the 25% RAP mixture delivered the best overall balance of structural response and cost efficiency, increasing Marshall stability from 9.40 kN to 14.28 kN, equivalent to a 51.9% improvement, while reducing estimated construction cost by 13.8% relative to the virgin mixture. Higher RAP contents showed lower SCB peak load, indicating that excessive aged binder increased mixture stiffness and brittleness, reduced effective blending with the virgin binder, and limited the mixture’s ability to dissipate fracture energy. For the materials and testing conditions investigated, 25% RAP is identified as the most practical and sustainable replacement level for local road applications.
The Buildings at the Mehran University of Engineering and Technology (MUET), Jamshoro, pose a major concern because of the critically low safety conditions. This study analysed the stability of the slopes around Building A and Building B with PLAXIS-2D (Finite Element Modelling Software). The analysis showed that the slopes of Buildings A and B are unstable under natural conditions, with FOS values of 1.099 and 1.002, respectively. Moreover, the slopes were analysed under erosion conditions, and the result showed that on 15° erosion condition, instability increases, reducing FOS to 0.978 and 0.874, respectively. Analysis with building offsets of 0 m, 1 m, and 2 m was also carried out, and it indicated that 0 m offset is most critical. Steep slope angles of 65.7° (Slope A) and 68.2° (Slope B), along with Rocky clay soil of this region, further increase landslide risk, highlighting the adverse influence of slope geometry. Therefore, concrete piles were used as a remedial measure to resist the slope failure. Numerical modelling showed that concrete piles are effective in enhancing slope stability. From the various experimental models, a 1-meter-diameter and 25-meter length (12-meter above crest and 13-meter-long pile embedded at the top of the slope) gave the most promising results. The FOS rose to 2.1 for Building A and 1.8 for Building B, respectively, which reflected a stable slope condition. Piles installed along the slope crest acted as a barrier to lateral soil movement, increasing resistance to sliding forces and thereby reducing the risk of slope failure and potential damage to nearby structures. This approach provides a safe and economic solution. The results of this study indicate that the installation of concrete piles is an effective method for slope stabilization in such geotechnical conditions.
Human Action Recognition (HAR) is vital in surveillance, sports analysis, and Human–Computer Interaction (HCI). Publicly available benchmarks mostly suffer from limited sample sizes, class imbalance, occlusions, and viewpoint variability, which may affect model generalization. Most existing HAR augmentation pipelines rely on generic or heuristic transformations that do not explicitly model camera-induced variations. To address these challenges, we propose an Augmented Data Generator Framework (ADGF) for HAR that integrates multiple parameter-controlled augmentation strategies designed to approximate realistic camera-induced variability while preserving temporal consistency. Our method integrates four transformation strategies. Each augmentation technique is parameterized across validated variation ranges, realistically simulating camera viewpoint changes, motion diversity, and partial occlusions, while preserving frame-to-frame temporal coherence, compared with uniformly applied conventional augmentations. ADGF distributes controlled transformations to maintain a balance and preserve action semantics. We evaluate performance with standard metrics, and perform an ablation study to measure each augmentation's contribution. With the combined augmented set, a top-1 accuracy on HMDB51 rises from 87.5% (baseline) to 97.7%, and final training loss falls from 0.21 to 0.12. Required epochs drop from 38 to 25 (≈34% fewer epochs). Cross-dataset testing on UCF101 demonstrates improved robustness, with accuracy increasing from 85.1% to 96.5%. The ablation study shows that each augmentation yields a positive gain with viewpoint variation, that occlusion simulation produces the largest single-technique improvements, and that the full combination produces the strongest transferability. These results indicate that a balanced structured augmentation strategy substantially improves both in-dataset performance and cross-dataset robustness. Implementation details, augmentation configurations, and experimental settings will be made publicly available upon publication to support reproducibility.
This study investigates the prediction and multi-objective optimization of tensile strength in woven twill fabrics using artificial neural networks (ANNs). A dataset of 135 fabric samples was developed using yarn tensile strength in the warp and weft directions, ends per inch, and picks per inch as predictors, with fabric tensile strength in the warp and weft directions as responses. Feed-forward ANN models with multiple hidden layers were trained using backpropagation and Bayesian regularization. The optimized models showed strong agreement with laboratory measurements. For warp-direction tensile strength, the correlation coefficients were 0.9998 for training, 0.9991 for testing, and 0.9997 for the complete dataset. For weft-direction tensile strength, the corresponding values were 1.0000, 0.99907, and 0.99994. The maximum and mean absolute errors were 8.4093 and 1.7588, respectively, for the warp direction, and 6.4159 and 0.37005 for the weft direction. The maximum percentage errors were 2.8719% and 4.28048%, respectively. A minimax-based multi-objective optimization identified the optimum manufacturing conditions as warp- and weft-yarn strengths of 495.36, 81 ends/in, and 81.9 picks/in. Experimental validation under these conditions produced tensile strengths of 585 in the warp direction and 660 in the weft direction. The results demonstrate that Bayesian-regularized ANN models can accurately predict twill-fabric tensile strength and provide practical manufacturing parameters for quality improvement and structural optimization.
South Asia is experiencing increasing energy demand and growing environmental pressure because of continued reliance on fossil fuels. This paper presents a data-driven evaluation of renewable energy deployment in eight South Asian countries, including Afghanistan, Bangladesh, Bhutan, India, Maldives, Nepal, Pakistan, and Sri Lanka. The study analyzes renewable energy share, installed capacity by source, CO₂ emissions, and technical potential for solar, wind, and biomass resources using harmonized datasets from international energy agencies. The assessment compares regional disparities and trend-based information to generate quantitative insights for policy implementation. Regression analysis indicates that the relationship between renewable penetration and emissions is statistically significant and negative (-0.68, 95% confidence interval: -0.82 to -0.54, R² = 0.72). The findings show that high renewable penetration is concentrated in hydropower-based countries, including Bhutan (82%) and Nepal (74%), whereas countries with lower renewable shares continue to face higher carbon impacts as energy demand grows. Solar energy appears to be the most predictable regional option, whereas wind and biomass potential remain location specific. The paper offers empirical data to support varied renewable energy policies and provides a policy-based analytical structure for sustainable energy planning in the developing world.
The accurate and rapid detection of pathogens is important for maintaining infection control, ensuring food safety, supporting environmental monitoring, and supporting public health surveillance. Relying on traditional methods is time-consuming, which typically takes 24-72 hours to generate results. This study presents a machine-learning-based bacterial detection system that integrates smartphone imaging with convolutional neural networks to classify Escherichia coli, Salmonella typhi, and Staphylococcus aureus. In this study, a portable smartphone-compatible microscope with a magnification of 750x and a resolution of 1 µm was used, enabling high-quality bacterial imaging directly at the sampling site. To train three CNN architectures, namely, ResNet50, VGG16, and InceptionV3, hundreds of microscopic images were obtained by preparing, scanning stained bacterial slides, and then pre-processing them for this study. Furthermore, these models were assessed using various bacterial pair combinations to evaluate the classification accuracy. The results revealed that the system achieved 98.6% accuracy, demonstrating robust classification performance. The developed system was shown to outperform traditional cultural laboratory methods, enabling it to transform an ordinary smartphone into a handheld bioresearch machine that provides diagnostic results within minutes. The novelty lies in the comparative assessment of three CNN architectures across multiple bacterial class combinations, enabling robust assessment of model performance in distinguishing morphologically similar and distinct bacterial species. Through the combination of optical imaging, convolutional neural networks, and adaptive mobile computing, the investigation gains a significant decrease in the turnaround and operational cost of the assay and maintains the same level of accuracy as gold-standard assays.
The housing sector in Pakistan has a documented backlog of 10.3 million units resulting from multiple social and economic challenges. The crisis is particularly severe in Sindh Province because of extreme climatic conditions and the vulnerability of informal settlements. This situation requires a robust engineering approach to deliver housing that is both resilient and affordable. Accordingly, a paradigm shift is needed to reconcile cost sensitivity with low-carbon sustainability. This study addresses this need by developing an optimized housing model based on standardized life cycle assessment (LCA) and multi-criteria decision-making (MCDM) techniques. A hybrid decision-support tool combining the analytic hierarchy process (AHP) and the technique for order preference by similarity to ideal solution (TOPSIS) was developed. In total, 13 alternative materials and 14 construction strategies were evaluated against environmental, economic, and technical criteria. Qualitative insights from industry professionals were cross-validated for consistency and statistical significance using Kendall's coefficient of concordance (W), whereas the quantitative assessment comprised a cradle-to-gate embodied-carbon inventory and standardized cost modeling. The findings indicate that compressed stabilized earth blocks (CSEBs) and precast roofing form the optimal configuration for the local context. Compared with a conventional brick structure, the proposed design reduced embodied emissions by 30% and construction costs by 13%. Strong expert consensus (W = 0.82, p < 0.001) also supported the model's implementation feasibility. The study therefore presents an actionable pathway for resilient and affordable housing by integrating rigorous decision-support tools with locally measured environmental, economic, and technical performance. It also contributes to the achievement of UN Sustainable Development Goals 9, 11, and 13.
Silver nanoparticles have been synthesized using physical, chemical, and biological methods because of their distinctive physicochemical properties. Among these approaches, chemical reduction is efficient, rapid, reproducible, and cost-effective. This study synthesized environmentally friendly, stable, and economical ofloxacin-capped silver nanoparticles (OFX-AgNPs) and evaluated them as a colorimetric sensor for the selective detection of Vermox in biological and environmental samples. The nanoparticles were characterized using UV-visible spectroscopy, X-ray diffraction (XRD), dynamic light scattering (DLS), zeta-potential analysis, and transmission electron microscopy (TEM) to assess their optical properties, structure, particle size, morphology, and stability. The OFX-AgNPs were uniformly dispersed, spherical, narrowly distributed in size, and highly stable. UV-visible spectroscopy confirmed the interaction between ofloxacin and the nanoparticle surface, while TEM analysis showed a uniform morphology and size distribution. The OFX-AgNPs exhibited high sensitivity and selectivity for Vermox, demonstrating their potential as an efficient colorimetric probe. Vermox was detected in wastewater, urine, and blood samples through a distinct optical response. The chemical reduction method was selected because of its simplicity, rapid synthesis, cost-effectiveness, and reproducibility compared with biological and physical methods.
Combustion instability in solid rocket motors (SRMs) arises from the coupling of unsteady pressure oscillations with chamber acoustic modes and can threaten structural integrity and operational reliability. This study examines how replacing a conventional axial inlet with a radial inlet affects internal flow dynamics, vortex-acoustic coupling, and pressure fluctuations in an SRM combustion chamber. Radial inflow introduces a circumferential velocity component that increases tangential shear near the walls, promotes earlier vortex roll-up around the inhibitor, and produces denser, more three-dimensional vortical structures, thereby strengthening vortex-acoustic interactions. Two-dimensional axisymmetric simulations were performed using scale-resolving simulation (SRS) and unsteady Reynolds-averaged Navier-Stokes (URANS) approaches. Pressure histories recorded at two monitoring points were analyzed using the fast Fourier transform (FFT) to identify dominant acoustic frequencies and normalized amplitudes. The SRS models agreed closely with the experimental data, accurately capturing the first four acoustic modes and the monotonic decrease in normalized amplitude with increasing mode number; LES with the WALE model produced the most reliable overall results. By contrast, the URANS models smoothed small-scale fluctuations, underpredicted higher-frequency modes, and misrepresented amplitude levels, particularly when the scale-adaptive simulation model was used. These findings show that the radial-inlet configuration substantially alters SRM flow physics and instability characteristics and demonstrate the value of advanced SRS approaches for inlet-specific designs intended to mitigate combustion instability.
nxiety and depression are crucial issues related to human mental health globally and are especially prevalent among users of social networking and online platforms such as X and Reddit. Roman Urdu has become a dominant language for expressing and communicating personal feelings on these platforms, making it a valuable resource for detecting early signs of depression. This study focuses on the automatic detection of depression in Roman Urdu social media posts using machine learning methods. A human-annotated corpus of 6,000 posts was created, and four machine learning classifiers, namely Naive Bayes (NB), Support Vector Machine (SVM), Random Forest (RF), and K-Nearest Neighbors (KNN), were trained using natural language processing features such as TF-IDF. The accuracy, training time, and testing time of these models were compared. The results demonstrate that the Random Forest classifier outperforms the other models, achieving 93% accuracy with high efficiency. The study highlights the effectiveness of scalable machine learning-based early detection methods for underrepresented languages such as Roman Urdu.
This research addresses the growing challenge of supply chain digitalization by emphasizing the importance of information exchange and cooperation. By extending the concept of digital twins, the study reviews, develops, and conceptualizes the emerging notion of Digital Supply Chain Twin (DSCT). While research on digital twin technology is continuous, data from many industry projects is easily accessible; its implementation within supply chains is growing as a multidisciplinary field, evolving from decades of advancements in supply chain digitization. While supply chains have employed digital twins, the idea of digitizing supply chains is still relatively emerging, with imprecise notions that go beyond important metaphors (such as visibility and provenance). The primary objective is to establish organizational and technological systems that enable supply chains to be monitored and controlled from integrated digital interfaces (such as a dashboard screen). Following an extensive literature review on the interactions between digital twins, digital supply chains, and digital technologies (such as blockchain), we propose a Digital Supply Chain Twin System (DSCTS) framework, specify an important gap (which we refer to as identifiability), and frame our findings in the absence of practical use cases. In the end, the proposed framework is applied to the expired medicine recycling of a simple reverse pharmaceutical supply chain.
Software-Defined Networking (SDN) has changed the way networks are built in modern days by separating the control and data planes which thus made it possible to have centralized programmability and flexibility. Although these security mechanisms have developed, SDN security management still depends on static or manually configured policies which are not sufficient anymore for the dynamic and sophisticated cyber threats. AI, especially, Reinforcement Learning (RL), has now become a reliable option to automate SDN security policies and thus allow quick adaptive response to changing threats. In this article, a complete theoretical analysis has been conducted where SDN security policy management through RL-based dynamic automation is juxtaposed against traditional methods. By combining recent studies outcomes, the current work delineates the pros, cons, and practical applicability of both methods. A proposal of a conceptual framework for the implementation of RL into SDN controllers is made which would subsequently help to connect the present gaps and guide future research in both academic and industrial sectors. The paper wraps up with the identification of research challenges, posing of questions that need to be answered, and giving suggestions for the further development of automated, intelligent SDN security architectures
Accurate segmentation of skin lesions is an important step in diagnosing skin cancer. However, most segmentation methods based on deep learning rely heavily on large volumes of manually annotated images, which are both costly and difficult to obtain. This paper proposes a new framework for semi-supervised learning that effectively links an abundance of unlabelled thermoscopic images with a small, labelled dataset to improve the performance of segmentation. The framework integrates an architecture of Teacher-Student, where the Teacher network builds pseudo-labels for the unlabelled samples. The refinement of pseudo-labels is achieved through uncertainty-based filtering, which discards unreliable areas, and is further enhanced by Conditional Random Field (CRF) post-processing, which sharpens the lesion boundaries. The refined pseudo-labels are then utilised for training the student network, allowing for robust learning through both unlabelled and labelled data. Experiments on the HAM10000 and ISIC 2018 datasets indicate that the proposed model achieves an IoU of 0.64 and a Dice Coefficient of 0.74 with only 10 percent of labelled data. It was comparable to fully supervised training with 100 percent annotations. Qualitative outcomes suggest improved delineation of lesion boundary and decreased false segmentation. This research work demonstrates the potential of pseudo-label refinement as an effective strategy for addressing the scarcity of annotations in medical image segmentation.
