Istanbul Technical University
Ulusal Üniversitelerarası Açık Erişim Sistemi - İstanbul Teknik ÜniversitesiNot a member yet
67356 research outputs found
Sort by
Optimization of a Power Divider and Patch Antenna Array for Ku-Band Satellite Communication
No full texthttps://doi.org/10.1109/isas66241.2025.1110179
Numerical and Experimental Performance Assessment of Helical Screw Inserts in a Horizontal Evaporator Tube
No full textThe influence of a helical screw insert on heat transfer and pressure drop in a circular tube under single-phase water flow conditions was investigated using both experimental techniques and CFD simulations with ANSYS Fluent. Experiments were conducted with water flowing through a horizontally oriented copper tube located in a pressurized tank, across seven different Reynolds numbers ranging from 24,000 to 100,000. A constant surface temperature was maintained via saturated pool boiling of R134a surrounding the tube. The internal flow characteristics were assessed based on measured pressure, temperature, and flow rate data, with convective heat transfer coefficients evaluated using the Wilson plot method. Simulations were conducted for both plain and insert-equipped tubes under identical conditions. The results for the helical insert case aligned well with experimental observations. Compared to literature correlations, heat transfer results were similar, though friction factor predictions were underestimated. A 19% improvement in heat transfer was observed with the insert, accompanied by a significant increase in pressure drop.https://doi.org/10.5541/ijot.170219
Maintenance Work Order Prioritization for Scheduling Using a Spherical Fuzzy Inference System
No full textExisting fuzzy inference systems are generally based on ordinary fuzzy sets, which do not let the second and third dimensions of the other fuzzy sets extensions to be employed. This paper suggests a decision-making approach by utilizing the fuzzy inference systems (FIS) based on spherical fuzzy sets (SFS). We prefer spherical fuzzy sets to consider the indecision degree together with membership and non-membership degrees in the proposed FIS. During the defuzzification of SF inference system, the indecision degree is distributed over membership and non-membership degree in balance regarding to indecision degree by using a special transformation function. By applying the proposed approach on FIS, it aims to cover hesitancies and uncertainties caused by insufficient assessments of the decision makers more effectively. The proposed decision-making approach is tested with a real-world application in the field of maintenance work order prioritization for scheduling. Finally, the result of the suggested approach based on SFS is compared with the risk assessment matrix technique (RAM) existing in the literature and Picture Fuzzy Inference Systems (PiFIS). It is observed that the proposed Spherical Fuzzy Inference System (SFIS) is more efficient than RAM and PiFIS methods.https://doi.org/10.15388/25-infor58
Implications for Non-Self Similar Energy and Moment Scaling of Small-to-Moderate Earthquakes Along the NAFZ: Source Displacement Spectra Derived from Coda Waves
No full textFor decades, the seismological community has debated the scaling relationships of earthquake sources. The debate centers around whether the scaled energy (ER/M0) remains uniform across all magnitudes, indicating self-similarity, or if there is an increase in scaled energy with seismic moment, M0. To contribute to this discussion, we analyzed coda derived source displacement spectra of 303 local earthquakes that occurred in and around the segments of the North Anatolian Fault Zone (NAFZ) within the Sea of Marmara. Our database includes digital waveform recordings of the events that were occurred between 2018 and 2020 (2.5≤ ML ≤5.7 within a radius of 200 km) and were recorded at 49 seismic stations operated by the Kandilli Observatory and Earthquake Research Institute (KOERI) in the study area. We employed a joint inversion technique to optimize source-, path-, and site-specific factors simultaneously. This was achieved by comparing the observed coda envelope with its physically derived representative synthetic coda envelope based on Radiative Transfer Theory. Our inversion process, conducted across various frequency bands, enabled us to make reliable coda-based seismic moment (M0) and moment magnitude estimates (Mw-coda) consistent with local catalogue magnitudes. The variation of the scaled energy (ER/M0) calculated from the total seismic radiated energy (ER) using coda-derived source displacement spectra for each event tends to increase with seismic moment across most magnitude ranges. This indicates that the crustal earthquakes with Mw-coda 2.5 and Mw-coda 5.7 in this laterally heterogeneous region are likely to follow non-self-similarity. Our findings imply different rupture dynamics working for large earthquakes than small ones and relatively more efficient seismic energy radiation for larger earthquakes along the northwestern part of the NAFZ.https://doi.org/10.5194/egusphere-egu24-74
Assessment of a Substandard Reinforced Concrete Frame’s Beam–Column Joint Using Shake Table Testing
No full textThis study investigates the seismic performance of substandard reinforced concrete (RC) frames, particularly emphasizing the behavior of beam–column joints through comprehensive shake table testing. Historical evidence indicates that RC structures erected prior to the 1970s frequently exhibit critical deficiencies in shear reinforcement, significantly undermining their seismic resilience. The primary objective of this research is to experimentally quantify the shear capacity, deformability, and failure mechanisms of beam–column joints within substandard RC frames. Shake table experiments were conducted on a meticulously scaled (one-third scale), single-story, single-bay RC frame, representative of construction practices typical of 1980s Türkiye. The input seismic loading was derived and scaled from the recorded ground motions of the 2023 Kahramanmaraş earthquake to ensure realistic seismic demand conditions. Experimental outcomes revealed a maximum lateral displacement of 53.1 mm, corresponding to a story drift ratio of approximately 0.055 radians and a computed damage index of 0.758, indicative of near-collapse performance levels. Notable damage observations included extensive shear cracking and concrete spalling at beam–column interfaces, accompanied by pronounced pinching effects during cyclic loading. The findings emphasize an urgent need for targeted seismic retrofit solutions, specifically addressing shear vulnerabilities in beam–column joints. Furthermore, the results carry substantial implications for revising current seismic design codes and enhancing the earthquake resilience of existing RC infrastructure in seismic-prone regions.https://doi.org/10.3390/app15084168https://doaj.org/article/98b4d2d73096431693bdbd53bf3fc74
Protein Katlanmasının Çözülmesi: Evrimsel Algoritmalar ve Yapay Zekanın Gücüyle Yenilikçi Yaklaşımlar
No full textAlphaFold, protein dizisinden üç boyutlu (3D) yapı tahmini yaparak yapısal biyoloji alanını değiştirmektedir. Bu olağanüstü başarı, protein katlanma probleminin "çözüldüğü" iddialarına bile yol açmıştır. Ancak protein katlanma problemi, diziden yapı tahmininden daha fazlasını içermektedir. Şu anda, AlphaFold devriminin protein katlanma ile ilgili diğer sorunları çözmeye yardımcı olup olmayacağı bilinmemektedir. Bu çalışmada, AlphaFold'un tekli mutasyonların protein stabilitesi (ΔΔG) ve fonksiyonuna etkisini tahmin etme yeteneğini değerlendiriyoruz. Bu soruyu incelemek için, AlphaFold tahminlerinden pLDDT ve metriklerini bir proteinde tekli mutasyondan önce ve sonra çıkarıyoruz ve tahmin edilen değişikliği deneysel olarak bilinen ΔΔG değerleriyle ilişkilendiriyoruz. Ayrıca, aynı AlphaFold pLDDT metriklerini yapı üzerindeki tekli mutasyonun etkisi ile büyük ölçekli bir GFP mutant veri setini kullanarak deneysel olarak test edilen floresans seviyeleri ile ilişkilendiriyoruz. AlphaFold çıktı metrikleri ile protein stabilitesi veya floresan değişimi arasında çok zayıf veya hiçbir korelasyon bulamıyoruz. Sonuçlarımız, AlphaFold'un diğer protein katlanma problemleri veya uygulamalarına hemen uygulanamayabileceğini göstermektedir.https://doi.org/10.46740/alku.152259
Cloud‐Based Control System with Sensing and Actuating Textile‐Based IoT Gloves for Telerehabilitation Applications
No full textRemote manipulation devices extend human capabilities over vast distances or in inaccessible environments, removing constraints between patients and treatment. The integration of therapeutic and assistive devices with the Internet of Things (IoT) has demonstrated high potential to develop and enhance intelligent rehabilitation systems in the e‐health domain. Within such devices, soft robotic products distinguish themselves through their lightweight and adaptable characteristics, facilitating secure collaboration between humans and robots. The objective of this research is to combine a textile‐based sensorized glove with an air‐driven soft robotic glove, operated wirelessly using the developed control system architecture. The sensing glove equipped with capacitive sensors on each finger captures the movements of the medical staff's hand. Meanwhile, the pneumatic rehabilitation glove designed to aid patients affected by impaired hand function due to stroke, brain injury, or spinal cord injury replicates the movements of the medical personnel. The proposed artificial intelligence‐based system detects finger gestures and actuates the pneumatic system, responding within an average response time of 48.4 ms. The evaluation of the system further in terms of accuracy and transmission quality metrics verifies the feasibility of the proposed system integrating textile gloves into IoT infrastructure, enabling remote motion sensing and actuation.https://doi.org/10.1002/aisy.202400894https://pubmed.ncbi.nlm.nih.gov/40852091/https://pmc.ncbi.nlm.nih.gov/articles/PMC12370170/http://dx.doi.org/10.1002/AISY.20240089
Derin pekiştirmeli öğrenme ile sürü savaş uçaklarının kontrolü
No full textThesis (M.Sc.) -- Istanbul Technical University, Graduate School, 2025This thesis explores the application of deep reinforcement learning (DRL) to achieve autonomous control of fixed-wing aircraft in single-agent and swarm configurations, addressing critical challenges in modern aviation. Autonomous unmanned aerial vehicles (UAVs) are pivotal for diverse applications, including civilian tasks like cargo delivery, agricultural surveillance, and disaster response, and defense operations such as reconnaissance and coordinated missions. However, nonlinear flight dynamics, dynamic mission requirements, environmental uncertainties, and communication constraints often limit traditional model-based control methods, which rely on complex mathematical models or predefined trajectory planners. DRL offers a data-driven alternative, learning optimal control policies through environmental interactions. This study leverages the Deep Deterministic Policy Gradient (DDPG) algorithm to develop robust control policies for F-16 aircraft, evaluated in a high-fidelity simulation environment integrating JSBSim and MATLAB/Simulink. The research investigates single-aircraft waypoint navigation and decentralized coordination in a three-aircraft swarm, demonstrating smooth, continuous control, high learning efficiency, generalizability, and robustness. Shared policy learning enables stable formation tracking and waypoint navigation in swarm scenarios without inter-agent communication, highlighting DRL's potential for scalable aerospace systems. Reinforcement learning (RL) frames sequential decision-making within Markov Decision Processes, comprising states, actions, transition probabilities, rewards, and a discount factor. DDPG, a model-free, off-policy actor-critic algorithm, is tailored for continuous action spaces. Its actor network maps states to actions, while the critic estimates state-action value functions. Enhanced by deep neural networks, experience replay buffers, and target networks, DDPG ensures stable learning by storing past transitions and mitigating training fluctuations. In this work, DDPG controls aircraft heading, roll, pitch, and thrust, adeptly managing complex, nonlinear dynamics. The simulation environment employs JSBSim's six-degree-of-freedom (6-DoF) model, which accurately simulates aerodynamic forces, propulsion, and environmental effects using Newton-Euler equations. JSBSim's modular design supports independent aircraft instances, enabling scalable swarm simulations. Integrated with MATLAB/Simulink, it facilitates real-time control algorithm development. MATLAB's Reinforcement Learning Toolbox supports DDPG training, offering tools for reward design, state/action space definition, and parallel computing, creating a robust RL framework. The DDPG agent operates with a 12-dimensional observation space, including normalized altitude, heading, roll, and pitch errors, angular rates, angle of attack, sideslip angle, and prior control signals, ensuring comprehensive state perception. The action space comprises four continuous commands—thrust, aileron, rudder, and elevator—normalized to actuator limits. The actor network, with two 800-neuron hidden layers and a tanh activation, produces bounded actions, while the critic's dual-path architecture estimates value functions. The reward function combines hyperbolic penalties for smooth error penalization, control effort penalties for energy efficiency, and temporal shaping for sustained stability, optimizing altitude and heading tracking. The closed-loop control architecture, implemented in MATLAB/Simulink, ensures real-time interaction with JSBSim dynamics. Simulations assess performance in single-agent and swarm scenarios. In the single-agent case, an F-16 navigates eight waypoints (27,000–34,000 ft altitude, 47.01°–47.12° latitude, 122.01°–122.14° longitude). Training shows rapid reward improvement, converging near zero by 6,000 episodes, indicating a stable policy. The aircraft achieves precise waypoint tracking, smooth transitions, and stable attitude, validated by metrics like waypoint accuracy and control smoothness. In the swarm scenario, three F-16s follow four shared waypoints in a triangular formation, using the same DDPG policy without communication. Initialized with ±0.005° offsets, the aircraft maintain altitude consistency, avoid collisions, and achieve synchronized arrivals, demonstrating formation coherence and decentralized stability. These results confirm DDPG's generalizability and scalability. The thesis establishes DDPG-based DRL as an effective approach for autonomous aircraft control. Contributions include a DDPG control framework, a randomized waypoint environment, decentralized swarm simulations, and a reward design promoting stability and efficiency. All hypotheses are validated: the agent tracks waypoints without planners, enables shared-policy swarm coordination, accelerates learning via reward shaping, and performs robustly under perturbations. Future work could explore larger swarms, obstacle avoidance, alternative DRL algorithms (e.g., SAC, TD3), and real-world policy transfer via hardware-in-the-loop testing. This study underscores DRL's transformative potential for autonomous aerospace systems, offering a foundation for scalable, learning-based control in dynamic flight environments.M.Sc
Variations of Environmental Niche Breadth, Range Sizes and Geographic Exclusion With Bat Species Richness
No full textABSTRACTAimMore species‐rich communities are often assumed to contain more specialist species with narrower niches and smaller ranges. Stronger interspecific competition in species‐rich communities is thought to be a key mechanism explaining these patterns. Yet, the relationship between richness and specialisation has so far only been studied for a few taxa, and characterising the effects of interspecific competition on species distributions is challenging. Here, we assess broad‐scale relationships between niche breadth, range sizes and geographic exclusion along richness gradients of bats.LocationEastern Mediterranean, Western Asia, and Central Asia.TaxonBats (Chiroptera).MethodsBased on a novel integrated species distribution modelling approach that combines occurrence information with expert range maps, we assessed how environmental niche breadth and range sizes varied with species richness. In addition, by contrasting species' potential and realised distributions in areas where species pairs overlap, we derived indicators of geographic exclusion to understand how potential interspecific competition is affecting range limits along richness gradients.Results and Main ConclusionsWe found a nonlinear association between environmental niche breadth and richness, with the most specialised species occurring in species‐poor regions and niche breadth peaking at intermediate richness. Despite a positive association of niche breadth and range sizes at the species level, range sizes in predicted bat communities declined continuously with species richness. In addition, patterns of geographic exclusion were linked to patterns of niche breadth, with species filling less of their potential range overlaps when overlapping species were more specialised. Our findings suggest that small range sizes in species‐rich bat communities are better explained by the number of interacting species than by environmental specialisation or stronger exclusion between individual species. More broadly, we show how integrated distribution modelling approaches can shed new light on the interplay of species richness, specialisation and community structure, and caution against generalising relationships between richness and specialisation across taxa and geographies.https://doi.org/10.1111/jbi.15125https://dx.doi.org/10.18452/34331http://edoc.hu-berlin.de/18452/34964https://doi.org/10.18452/3433
Streamflow Intervals Prediction Using Coupled Autoregressive Conditionally Heteroscedastic With Bootstrap Model
No full textABSTRACTStreamflow (Qflow) process is one of the complex stochastic processes in the hydrology cycle owing to its associated non‐linearity and non‐stationarity characteristics. It is an essential hydrological process to address the complex time series nonlinear phenomena. In this research, a novel approach was proposed by integrating an autoregressive conditionally heteroscedastic (ARCH) method with bootstrap model to predict future Qflow intervals. For this purpose, two Qflow series located at the Eastern Black Sea basin (Turkey) were subjected to the application of the proposed methodology. Among other regression and machine learning (ML) models, which are suitable for Qflow modeling, the autoregressive integrated moving average (ARIMA), seasonal autoregressive integrated moving average (SARIMA), and artificial neural network (ANN) were selected for modeling validation in this study. A group of three numerical metrics and graphical presentations were used for the modeling evaluation and assessment. The proposed ARCH approach performed a superior mathematical model to address the Qflow interval prediction. Remarkable prediction accuracy was shown against the benchmark models. Overall, the approach of coupling the bootstrap procedure with the ARCH model exhibited a robust modeling strategy for predicting Qflow intervals suggested as a new analysis tool.https://doi.org/10.1111/jfr3.70009https://doaj.org/article/68bd8c1e2f1c4c1a9d049a8f549862d