The reusage of gypsum-based construction materials in geotechnical applications provides a sustainable alternative for soil stabilization. In this study, the effects of synthetic gypsum (CaSO4·½H2O) on the physical and mechanical properties of a low-plasticity silt (ML) type kaolin soil were experimentally investigated. Atterberg limit, standard Proctor and unconfined compressive strength (UCS) tests were applied on untreated and treated soil samples by rates of 2%, 5%, 8%, and 11% gypsum by dry weight for curing periods of 1 and 15 days. The results of the experiments revealed a significant change in the mechanicai behavior of the soil by the usage of gypsum treatment. The soil plasticity decreased by low to moderate gypsum additive while an improvement was observed for the compaction characteristics pointing out a denser and more stable particle structure. A clear improvement has been observed for strength properties of the soil with increasing gyspum content up to an optimum level, pointing out an enhancement for interparticle bonding and pozzolanic interactions in the soil matrix. Besides, higher gypsum rates resulted in a strength reduction by increased deformability, pointing out a distortion in the soil fabric and the formation of less effective bonding at excess rate of binder contents. Consequently, the findings revealed that construction material type gypsum can serve as an effective stabilizing binder for ML-type soils when applied at moderate rates. However potential durability problems and long-term performance should be carefully considered in practical applications.
Urban air mobility increasingly relies on autonomous multicopter fleets whose operational sustainability remains constrained by the absence of intelligent recharging infrastructures. This study introduces a simulation based decision intelligence model designed to evaluate six multicopter charging station archetypes under smart city conditions. The proposed framework integrates five normalized evaluation factors, namely Security, Infrastructure Cost, Logistics Compatibility, Smart City Integration, and Sustainability, within a transparent and auditable multi-criteria decision framework. Two complementary evaluation modes are developed to ensure analytical rigor and interpretability. The first mode, Mode A, represents a reproducible baseline configuration that employs equal weighting to retained methodological clarity. The second mode, Mode B, functions as a bounded coordination operator that establishes a controlled relationship between infrastructure capacity and logistics flow, enabling interaction informed evaluation without altering the ranking logic. Synthetic decision data are generated through Latin Hypercube Sampling, while bootstrap resampling is used to quantify uncertainty. The stability of both modes is analytically verified, showing that Kendall’s τ exceeds 0.90 and Top-k retention remains above 95 percent. These results demonstrate that introducing interaction awareness refines interpretability while maintaining analytical consistency across uncertainty ranges. The findings reveal that Last Mile and First Mile stations maintain the highest composite efficiency scores, 0.82 and 0.80 respectively, across various urban morphologies. Roof and Electric Vehicle Coupled configurations also display competitive scalability and improved performance when aligned with renewable energy scenarios. The overall framework provides a reproducible, policy aligned, and scientifically traceable foundation for the planning, deployment, and empirical calibration of urban drone charging networks. It further establishes a consistent methodological pathway for decision making in data scarce environments, ensuring that analytical transparency and operational relevance are sustained throughout future pilot implementations.
This study presents a comparative analysis of Principal Component Analysis (PCA) and ANOVA-based feature selection methods for Android malware detection, evaluating their impact on classification accuracy and computational efficiency. Three preprocessing scenarios were examined: using the original dataset with 241 features, applying PCA for feature extraction (retaining all components due to variance thresholds), and employing ANOVA to reduce the feature set to 120. Support Vector Machines (SVM), Wide Neural Networks, and Logistic Regression classifiers were trained on these datasets, with hyperparameters optimized via 5-fold cross-validation. Results demonstrated that SVM consistently achieved the highest accuracy across all scenarios, peaking at 99.25% with PCA. However, PCA failed to reduce dimensionality of models and increased training times for SVM compared to the original dataset. In contrast, ANOVA effectively reduced the feature count, lowering SVM training time to 4.81 seconds while obtaining 98.95% accuracy. These findings highlight ANOVA as a computationally efficient method, balancing high detection performance with reduced resource demands. While PCA marginally improved accuracy, its computational cost renders it less practical for real-time applications. The study concludes that feature selection via ANOVA offers a superior trade-off for Android malware detection, prioritizing both accuracy and efficiency. Future work should explore advanced feature selection techniques and validate models on diverse datasets to enhance generalizability and address evolving malware threats.
The diversity and sophistication of malicious content has significantly impacted end-users of Information and Communication Technologies. In order to mitigate the impact of malicious content, automated deep learning-based techniques have been developed to proactively defend user systems against malware. In this study, we implement a hybrid model for malware detection and classification using the MaleVis dataset. First, feature extraction from the dataset is performed using DenseNet-121, EfficientNet-B0 and ResNet-50 models. These models are deep learning architectures that have been trained on large datasets and are known for their powerful feature extraction capabilities. Each model was used to extract feature vectors from the images in the Malevis dataset. These feature vectors were then merged. The combined feature vectors were used for classification using XGBoost, a powerful classification algorithm. This hybrid model approach combines the feature extraction capabilities of deep learning models with the classification capability of XGBoost to detect malware. Experimental results show that the proposed hybrid model achieves high accuracy rates on the MaleVis dataset. The study shows that combining the feature extraction capabilities of different deep learning models and using these features with a classifier such as XGBoost can provide significant improvements in malware detection and classification.The results demonstrate the model’s potential for integration into real-world threat detection systems.
This paper presents a comparative evaluation of two integration strategies for the Xilinx Zynq-7000 System-on-Chip (SoC): an Advanced eXtensible Interface-Direct Memory Access (AXI-DMA)-based architecture and a Block RAM (BRAM)-based architecture. Both designs employ a custom processing element (PE) for arithmetic operations, yet they differ significantly in data transfer and buffering mechanisms. In the AXI DMA design, communication between the processing system (PS) and programmable logic (PL) is achieved via an AXI4-Stream interface controlled by a DMA engine. In contrast, the BRAM-based design uses dual-port block memories via AXI BRAM controllers, enabling direct operand access. Implementation results indicate that both designs comfortably meet the resource constraints of the XC7Z020 device. However, the AXI DMA-based architecture exhibits higher hardware resource utilization, with average consumption approximately 54% greater than that of the BRAM-based design. Performance analysis reveals a pronounced latency difference: the AXI DMA design required an average of ~1.19 ms per operation. In comparison, the BRAM-based approach achieved a reduction of ~0.10 ms, resulting in a total execution time of 32,487 µs compared to 359,919 µs.These findings demonstrate a clear trade-off between scalability and latency. While AXI DMA provides flexibility and throughput for stream-oriented applications, BRAM-based integration delivers superior efficiency in small-scale, low-latency scenarios. The study offers practical insights for guiding the design of Field-Programmable Gate Array (FPGA)-based accelerators on heterogeneous computing platforms.
Video-based steganography has attracted increasing attention due to its high payload capacity and improved imperceptibility compared to image-based approaches. In this study, a deep learning–based steganographic framework is proposed to embed and recover textual information within video content using the U-Net architecture. Unlike traditional least significant bit (LSB)–based techniques, the proposed method utilizes region-of-interest (ROI) selection and patch-based embedding to enhance robustness and visual quality. Textual data are first encoded into image patches and embedded into selected regions of video frames via a trained hiding network. A corresponding revealing network is employed to recover the hidden information, followed by an optical character recognition (OCR) pipeline for text extraction. Experimental results demonstrate character recovery accuracies between 81% and 88% while preserving high visual fidelity in the stego videos. This ROI-guided U-Net framework provides an effective and scalable solution for secure and imperceptible text hiding in video streams.
This study investigates the influence of source parameter variability—specifically stress drop and anelastic attenuation—on site response analysis using the Random Vibration Theory (RVT) framework. Site amplifications were computed for a representative site in the Aegean region of Türkiye using four different source parameter sets based on regional literature. A single set of randomized Vs profiles was used to account for subsurface variability, while varying source models enabled the evaluation of epistemic uncertainty in site amplification. The epistemic uncertainty associated with source parameters, expressed as τ_"source" , peaks at approximately 0.04 s, reflecting the sensitivity of high-frequency site response to stress drop and attenuation. The study provides a regionally calibrated quantification of source-parameter-driven epistemic uncertainty for the Aegean region. The findings emphasize the need to account for source-related uncertainty in seismic hazard assessment. Unlike most previous RVT-based studies, this work explicitly isolates and quantifies the effect of stress drop and attenuation on site amplification, highlighting a novel contributor to epistemic uncertainty. Future work should incorporate Monte Carlo simulation of source parameters to support PSHA applications.
In this experimental study, the rheological properties of cement-sodium silicate (C-SS) mixtures prepared at different ratios were investigated. The experiments were carried out by adding 3% to 6% cement by mass of the total volume to 30%, 50%, and 70% sodium silicate/solution (SS/S) mixing at 25 °C. The samples prepared for the rheological property determination experiments were mixed at a constant speed of 100 rpm. In the gelation experiments, two different cements, ordinary Portland cement (OPC) and ultra-fine cement (UFC) were used as reactants. Sodium silicate, also known as water glass, was used in the experiments in liquid form, and its modulus was 3. The Blaine value of OPC was 350 m2/kg, whereas that of UFC was 900 m2/kg. UFC-sodium silicate mixtures have faster gelation times than OPC-sodium silicate mixtures due to their high Blaine values. For both cements, the gelation time was shortened as the cement ratio increased, and the gelation time was extended as the sodium silicate ratio increased. The syneresis percentages decreased as the cement ratio increased, whereas the sodium silicate ratio increased up to 50% and decreased after 50%. Viscosity values increased with increasing sodium silicate ratio and cement amount. In this experimental study, gelation reactions could be achieved as a result of sodium silicate-cement reactions using lower cement ratios than those reported in the literature. Another unique aspect of the study is that cement can also be used as a reactant to prepare silicate grouts. The rheological properties of these mixtures, which can be used as alternative grout materials were determined.
Because they have become an integral part of our daily lives, large quantities of hazardous materials are produced and transported each year. In most industrial societies, life without hazardous materials has become almost unimaginable. Hazardous materials are defined as substances that, during transportation, have the potential to pose adverse effects or risks to public health, safety, or property due to their quantity or form. In this context, hazardous materials or products include explosives, gases, flammable and oxidizing substances, toxic and infectious materials, as well as radioactive, corrosive substances and their associated hazardous wastes. The safe transportation of hazardous materials is considered a comprehensive and multidimensional issue, influenced by various legal and physical factors, as well as the numerous risks that vehicles may encounter during transit. Increasing environmental awareness of the potential impacts of hazardous material accidents on public health has significantly heightened both academic and institutional interest in this field. This study proposes a risk-averse solution to the hazardous material transportation problem through a model developed by integrating the Tabu Search algorithm with a game theory–based approach. Within the model, the dispatcher aims to minimize the expected loss under the worst possible conditions in the event of a disruption in any link of the distribution network. In this framework, the expected cost determined through Nash equilibrium is evaluated as an effective and practical analytical tool for strategic decision-making in the selection of safe routes for hazardous material transportation.
This study examines the seismic performance of masonry buildings in Turkey between 1992 and 2023, focusing on damage assessments after major earthquakes. Common failure mechanisms identified include out-of-plane wall collapse, inadequate connections between perpendicular walls, improper wall openings, and the absence of bond beams. Heavy compacted clay roofs, often used in rural masonry structures, contribute significantly to seismic vulnerability by increasing inertial forces. The irregular use of rubble stones and insufficient mortar adherence led to early structural degradation. The study emphasizes the need for improved construction practices, compliance with seismic codes, and effective retrofitting strategies to enhance the seismic resilience of masonry buildings. Lessons learned from past earthquakes can enable engineers and policymakers to develop more robust guidelines for future earthquake-resistant masonry construction.
In this study, the capture of CO2 released from an industrial production facility located between the provinces of Batman and Siirt using suitable solutions was modelled using the Aspen Plus simulation. In the study, a methyldiethanolamine (MDEA) solution was chosen as the CO2 capture agent. The process design utilized absorbers, strippers, heat exchangers, flash evaporators, and recirculation loops; different feed rates and tray counts were tested. In the first stage, only 47% of the CO2 could be captured with a feed solution of 400 t/h. However, by increasing the solution flow rate to 832 t/h, approximately 72.14 t/h (95%) of the 76 t/h of CO2 present in the feed stream was successfully captured. Furthermore, it was observed that the number of trays, initially set at 30 in both the absorber and stripper, could be reduced to 21 without any significant change in performance, thereby increasing the cost-effectiveness of the process.The results obtained demonstrate that the CO2 capture method using MDEA solutions in industrial facilities provides high efficiency. However, it is emphasized that CO2 should not only be captured but also utilized in chemical/fuel production. The widespread adoption of such technologies will significantly contribute to Turkey's achievement of future climate goals.
The automated colorization of manga presents unique challenges due to its distinctive artistic style and complex visuals. While deep learning has shown promise in image colorization, existing approaches often struggle with consistency and artistic integrity of manga artwork. This paper presents a comparative analysis of two deep learning architectures for manga colorization: a modified U-Net with progressive dropout and a ResNet-based autoencoder with adaptive skip connections. We introduce a novel composite loss function that specifically addresses manga-specific challenges by incorporating structural and perceptual components. Experiments on a diverse manga dataset show that the ResNet-based model achieves higher color consistency and better stability, producing fewer artifacts in uniform areas. However, U-Net preserves fine details more effectively. These results provide insights into trade-offs between architectures, guiding practical implementations of manga colorization systems.
This study numerically investigates the compressive behavior and energy absorption performance of bio-inspired honeycomb structures. A validated LS-DYNA finite element framework was first established for a regular hexagonal honeycomb and correlated with experimental results, showing good agreement in terms of force–strain response and deformation modes. Subsequently, seven bio-inspired configurations—spider, snail, wavy, bamboo, pomelo peel, grass stem, and hierarchical—were modeled under quasi-static compression. The results revealed that bio-inspired designs significantly influence deformation pathways and energy absorption capacity compared to the regular hexagon. Among the proposed designs, the pomelo peel, grass stem, and hierarchical honeycombs exhibited the highest specific energy absorption (7.88, 7.50, and 7.39 J/g, respectively), representing an improvement of up to 47% compared to the reference structure. This improvement is attributed to their multi-cell and hierarchical load-transfer mechanisms that delayed densification and ensured a prolonged plateau region. While spider and bamboo designs provided balanced performance with moderate specific energy absorption, the wavy and snail geometries demonstrated smoother plateau behavior with lower peak forces. Overall, the findings highlight that bio-inspired geometrical features can be effectively employed to enhance the crashworthiness of lightweight structures, offering valuable insights for future applications in transportation, packaging, and energy storage systems, and guiding the development of next-generation lightweight structural designs.
With the increasing use of Android devices, forensic investigations have become crucial in uncovering cybercrimes involving mobile malware. Android devices, as one of the mobile device types, can be easily exploited due to weaknesses in the Android operating system and security vulnerabilities in the application store. While existing studies primarily focus on malware detection using machine learning models, there is a gap in the literature regarding the effectiveness of examination tools in analyzing harmful applications. This study evaluates forensic methods used to extract and analyze digital evidence from compromised Android devices. We compare manual inspection, logical imaging, and physical imaging in retrieving nine key evidentiary features. Our findings indicate that while manual and logical imaging recovered 55.56% of these indicators, physical imaging offered broader access (66.67%), particularly facilitating the recovery of deleted data and data from unallocated space. Using the Magnet AXIOM tool and manual analysis methods, we conducted static and dynamic analyses of malicious softwares. The results demonstrate the utility of specialized analysis tools in both identifying malicious activity and recovering critical information, offering guidance to practitioners in choosing the most effective approach for Android-related casework.
In recent decades, several environmental parameters such as climate change, expanding urbanization, deforestation, and land use land cover change resulted in various environmental issues. To evaluate and observe these impacts, this study indicated the Land Surface Temperature (LST), and some spectral indices including Normalized Difference Vegetation Index (NDVI), Normalized Difference Bareness Index (NDBaI), and Normalized Difference Built-Up Index (NDBI) in the Burdur Basin for the years between 2014-2025. Geographical Information System (GIS) and Remote Sensing technologies were integrated by using the images obtained from multi-temporal Landsat satellites to calculate correlation analysis, temporal changes, and spatial distribution of indices used in this study over the Burdur Basin. Based on the results of this study, LST values increased especially in the northeastern and central parts of the basin in 2016 and 2019. Although no significant variation was seen for NDBI and NDBaI values during the studied period, slight upward changes were partly observed in 2018. NDVI values varied between -0.36 and 0.68 while those values were observed mostly stable over years. A very weak positive correlation analysis was indicated between LST and NDVI. Additionally, the relationship between LST and NDBI, and NDBaI produced a moderately positive correlation between 2014-2025. The results indicated that, expanding urban and bare soil areas, resulted in increased LST values. In contrast, presence of vegetation can help to reduce surface temperature. To conclude, this study proved that, the presence of vegetation, urbanization, and land use and land cover properties has an impact on LST values over the Burdur Basin. Furthermore, this study indicated that integrating GIS and Remote Sensing technologies are valuable technique to evaluate temporal and spatial environmental changes.
DC-DC converters are essential components in a wide range of applications, including motor drive systems, renewable energy sources, and computer power supplies. These converters regulate an unregulated input voltage to provide a stable DC output under varying load conditions. Depending on the desired voltage level, several converter topologies have been developed. Among them, the buck converter stands out as one of the most widely adopted solutions due to its ability to step down the input voltage efficiently. To ensure that the output voltage remains close to a predefined reference value, various closed-loop control strategies are commonly employed.This study presents the implementation of a model-based, auto-tuning Proportional–Integral–Derivative (PID) control algorithm for precise output voltage regulation in a DC-DC buck converter. Unlike classical PID controllers, which operate with fixed gain parameters and exhibit limited adaptability to parameter variations and external disturbances, the proposed controller dynamically adjusts its parameters in real time. This adaptive capability enhances the system's robustness under varying operating conditions. Simulation results validate the effectiveness of the proposed method, demonstrating fast transient response, minimal overshoot, and high steady-state stability. These findings suggest that the approach offers a practical and efficient control solution, especially for embedded systems requiring adaptive and high-precision regulation.
This study enhances the classical z-score-based pairs trading strategy by introducing dynamic signal delay mechanisms to develop a time-adaptive approach to statistical arbitrage. Using Apple Inc. (AAPL) as a benchmark, 29 stock pairs from the Dow Jones Industrial Average (DJIA) index were analyzed to assess the impact of execution delays ranging from t+1 to t+5 on trading performance. Positions were opened and closed based on z-score signals derived from daily closing prices, where delayed execution aimed to reduce short-term market noise and optimize trade timing.Empirical results demonstrate that the proposed strategy achieved favorable performance, with 93.8% of the pairs generating positive returns and 89.7% attaining a Sharpe Ratio greater than 1.0. On average, the t+3 delay window yielded the most effective balance between risk and return, achieving a Sharpe Ratio of 2.371 and a cumulative return of 192.98%. Only 24.1% of the pairs performed best under immediate execution (t+0), highlighting the advantages of adaptive timing in arbitrage.Overall, the findings confirm that optimizing trade timing significantly enhances the profitability and stability of arbitrage models. The results provide empirical evidence supporting the potential of time-adaptive execution as a valuable improvement to traditional pairs trading frameworks.
Increasing charging demand with the widespread use of electric vehicles leads to negative effects such as load imbalance, sudden load changes, harmonics and voltage fluctuations in the electricity distribution network. Furthermore, irregular charging demand negatively impacts electric vehicle user comfort and traffic management. This study presents a dynamic pricing-based energy management model developed for use in urban electric vehicle charging infrastructures to address these challenges. The proposed model considers price not only as an economic output but also as a control variable that manages grid load balance. There are four input parameters (traffic, station occupancy rate, location and state of charge) in the pricing model and these parameters are dynamically updated at each iteration. The model was developed in MATLAB environment and was employed real-time traffic data obtained through the Google Maps API. The model tested for ten iterations. The results show that the pricing model prioritizes low charge levels vehicles. But the model maintaining balanced grid load simultaneously. Furthermore, price output increases high occupancy rates charging stations in order to encourage users to choose stations with lower occupancy rates. Results of this study demonstrates that pricing mechanism can be used as a decision variable both economic reasons and system efficiency. In future works, the model might be extended with artificial intelligence and optimization-based methods. Pricing model serves as a potential solution to challenges in energy and transportation networks with the help of test systems.
This study presents the design, construction, and evaluation of a fixed-wing Vertical Take-Off and Landing (VTOL) unmanned aerial vehicle (UAV) equipped with an onboard real-time visual-intelligence system optimized for critical missions such as search and rescue, surveillance, and precision agriculture. The UAV was built using lightweight, cost-effective materials and 3D-printed components, which considerably reduced development costs and improved accessibility for both academic research and field applications. A central contribution of this work is the integration of VTOL functionality with real-time deep-learning inference on embedded hardware within a fully open-source architecture. Unlike most existing UAVs that depend on bulky or expensive hardware, the proposed system performs efficient object detection (YOLOv5s) directly on a Raspberry Pi 4B, enabling onboard processing without external computation. Three detection models—YOLOv5s, Tiny-YOLOv4, and MobileNet-SSD—were trained on a custom aerial dataset and evaluated for real-time performance. YOLOv5s achieved the highest accuracy, with a mean Average Precision (mAP@0.5) of 82.4 % at 4.2 FPS. Owing to its modular and scalable design, the proposed UAV platform offers a practical and affordable solution for implementing intelligent aerial systems in real-world critical-mission environments.
The safety of hydraulic structures has social, economic, and environmental significance. To eliminate the damage induced by local scouring, either the scouring should be prevented entirely or the scouring depth should be minimized. This study numerically investigated the local scour that develops in the plunge pool downstream of high-head, classical contracted rectangular weirs operating under free overfall conditions. In the numerical analysis, a three-dimensional Reynolds-averaged Navier–Stokes (RANS) model with a standard k–ε turbulence closure was implemented in Flow-3D®. The analyses were conducted for three unit flow rates (q=0.15, 0.30, and 0.60 m3/s m) and three drop heights (H=0.25, 0.50, and 1.00 m) to determine the maximum scouring depth. Furthermore, the water jet velocity and impingement angle were analyzed. The experimental results were utilized to test the accuracy of the numerical model. For all cases, about 90 % of the equilibrium scour depth was reached within the first 15–20 min, consistent with previous studies. According to the results, the equilibrium scour depth increased from approximately 0.16 m to 0.23 m when the unit discharge was raised. However, after a specific value of the head height, the effect on the depth of scouring is slightly reduced due to the increased air entrainment entering the downstream pool. Computed jet impact velocities ranged from about 2.0 to 4.9 m/s, with impact angles increasing up to ~82° as the drop height increased. The results of the numerical model are generally compatible with experimental studies. This study is original in that it provides a three-dimensional CFD-based assessment of local scour downstream of high-head weirs under free overfall conditions and proposes a practical approach to estimate long-term equilibrium scour depth and jet impact characteristics.
