LEE- Mekatronik Mühendisliği Lisansüstü Programı

Bu topluluk için Kalıcı Uri

http://hdl.handle.net/11527/19385

Gözat

Quadcopter trajectory tracking control using reinforcement learning

(Institute of Science and Technology, 2019-06-11) Erdem, Mustafa ; Altuğ, Erdinç ; 518151036 ; Mechatronics Engineering

Unmanned aerial vehicles (UAVs) have gained enormous popularity since the last couple of decades. Quadcopters are the most popular subdivisions of UAVs. Their vertically taking-off, landing and hovering abilities make them ideal platforms for military, agriculture, surveillance and exploration missions. Their mechanical simplicity and agile maneuverability are other two reasons why the quadcopters are so popular. These mentioned reasons make the quadcopters excellent proving grounds for control theory applications. Even though designing a conventional controller for quadcopters is a relatively easy task, tuning those control parameters might easily become a time consuming challenge. Moreover, this requires a model of the system and any uncertainties in the system model or later modifications on the vehicle can quickly cause instabilities. Reinforcement learning is a subclass of artificial intelligence. The idea behind reinforcement learning is making an agent learn in an interactive environment by trial and error principle to achieve a specific task. Notwithstanding it has been discovered long ago, it has got its popularity back with the last advancements in the technology. In this thesis, at first, a conventional PD controller performance on a quadcopter model that is modeled on ETHZ Rotors framework in the Gazebo simulation environment was improved by implementing metaheuristic particle swarm optimization (PSO) algorithm. Thereafter using an actor-critic reinforcement learning algorithm called deep deterministic policy gradient (DDPG), quadcopter was trained to follow different trajectories. DDPG is an off-policy and model-free method, which has proven itself in different domains and tasks. DDPG has four neural network function approximators. These are actor, critic, target actor and target critic networks. The critic network approximates the current value of the agent state and the actor network generates actions with respect to state of the agent. During training, network values shift constantly. Using a constantly shifting set of values to adjust network parameters is not a reasonable thing to do. This makes the value estimations unmanageable. In order to avoid this, DDPG algorithm uses target networks that are used to make the training process more stable. These target networks are not updated at each step, contrary only periodically or slowly updated. Weight decay and batch normalization techniques that are normally not part of the original DDPG algorithm were also implemented to improve algorithm's performance. ADAM algorithm was used for optimization purpose. While training continues, the agent was presented a reward for each step in all episodes. Reward function is defined as negative weighted sum of quadcopter's position, velocity and acceleration errors. Tracking was assumed to be successful, if the tracking error is less than 10%. Tracking performances of both controllers were analyzed for different trajectories. PD controller outperforms reinforcement learning agent in most cases. However, it is needed to be stated that performance differences between two controllers are hard to notice and generalization, which is working on different quadcopter models under some assumptions, is the real advantage of reinforcement learning agent. Hyperparameters of the DDPG algorithm shape the learning behavior of the agent. It is highly possible for a reinforcement learning agent to perform equally or better compared to the conventional controllers. Therefore, as future work, with a given sufficient time, optimizing learning algorithm's hyperparameters and modifying network architectures are worth to investigate in order to have better performances.
İşbirlikçi robotların haptik arayüzlerle teleoperasyonu

(Lisansüstü Eğitim Enstitüsü, 2021-03-03) Argın, Ömer Faruk ; Bayraktaroğlu, Zeki Yağız ; 518142008 ; Mekatronik Mühendisliği ; Mechatronics Engineering

Robotlar günlük hayatımızda ve üretim alanında katlanarak artan bir şekilde kullanılmaktadır. Gelecekte daha da akıllanarak insanlar ile birlikte çalışacak ve diğer robotlarla işbirliği içinde çalışarak akıllı fabrikalar oluşturacaklardır. Robotlar bazı uygulamalarda, insan operatörünün kontrolünde ve uzaktan teleoperasyonla kontrol edilmektedir. Uzakta çalışan robot bilinmeyen ve dinamik bir çevreyle etkileşime girmektedir. Bazı uygulamalarda daha hassas ve kararlı teleoperasyon gerçekleştirebilmek için operatör uzak ortam hakkında görsel geri beslemenin yanında haptik geri beslemeyle beslenmektedir. Bu haptik geribeslemeye sahip teleoperasyon uygulamalarına haptik teleoperasyon sistemleri denir. Bu çalışmada, bir lokal manipülatör ve çok uzak manipülatörle işbirlikçi haptik teleoperasyon için kontrol şemaları önerilmektedir. İşbirilikçi teloperasyon için konum kontrollü, kuvvet/konum kontrollü paylaşımlı kontrollü kontrol şemalar sunulmaktadır. Uzak manipülatörler için önerilen kontrol yapılarının asimptotik kararlılıkları iletişim kanallarında sınırlı zaman gecikmelerinin olduğu durum için Lyapunov analizi ile gösterilmektedir. Önerilen kontrol şemaları endüstriye yönelik basit manipülasyon ve pim yerleştirme uygulamaları ile deneysel olarak gerçekleştirilmektedir. Deneysel uygulamalarda önerilen kontrolcüler, kararlı teleoperasyonlarda konum takip performansları, eğitimli ve eğitimsiz kullanıcılar ile uygulanabilirlik performansı, engelli ortamda manipülasyon ile sağlamlık analizleri ve sistemin şeffaflığı incelenmektedir. Önerilen kontrol şemalarının uygulanması ve deneysel doğrulaması, modüler bir sanal modelleme ortamı yardımıyla elde edilir. Çift yönlü haptik teleoperasyon için önerilen kontrol şeması, birden fazla uzaktan manipülatörle işbirliği görevlerine sanal modelleme ortamı yardımıyla kolayca genişletilebilmektedir. Manipülatörün geometrik parametrelerini içeren kinematik zincirin sanal ortamda tanımlanması ile önerilen iki yönlü ve işbirlikçi kontrolörlerin uygulanması için yeterlidir. Başka bir matematiksel modele ihtiyaç duyulmadığından, geometrik parametreleri mevcut olan herhangi bir rastgele robot kolu haptik teleoperasyon şemasına kolayca entegre edilebilir. Endüstriyel manipülatörlerin teleoperasyon ile kontrolünde, kullanıcı tarafından uygulanan yüksek frekanslı hareket referansları, sistemin kararsız davranışına neden olabilir. Ayrıca manipülatör kontrolünde aktüatör akımları genellikle güvenlik nedenleriyle sınırlandırılır bu da takip problemine neden olmaktadır. Herhangi bir haptik arayüzle kuvvet geri beslemesi olmadığında, insan operatör tarafındaki hareket referansları izin verilen giriş frekanslarını aşabilir. Bu durumlar uzak ortamda istenmeyen hareketlere ve hasar vb. neden olur ve ayrıca deneyimsiz kullanıcıların eğitim sürelerini uzatır. Bu çalışmada önerilen sanal haptik etkileşim kuvveti kullanıcıya teleoperasyon sırasında görünür bir atalet hissi sağlar. Yapay olarak oluşturulan bu ikinci dereceden dinamikler, kullanıcıdan gelen yüksek frekanslı hareket referanslarını filtreler ve bu nedenle deneyimsiz kullanıcıların kolayca teleoperasyon gerçekleştirmesine olanak tanır. Önerilen haptik etkileşim sistemi, manipüle edilen yük ile birlikte çalışan manipülatörler arasında sanal yay sönümleyicilerin kullanılmasına dayanmaktadır. Bu çalışmada sunulan deneysel sonuçlar, önerilen haptik etkileşim sisteminin tahmin edilen katkılarını doğrulamaktadır.
Image processing software tools development for shoulder arthroplasty

(Graduate School, 2021-05-21) Sadeghi, Majid Mohammad ; Ertuğrul, Şeniz ; 518132013 ; Mechatronics Engineering ; Mekatronik Mühendisliği

Reverse shoulder arthroplasty is an operation performed on shoulder joints with diseases such as osteoarthritis and rheumatoid arthritis, complex fractures of the proximal humerus, and osteonecrosis of the humeral head. This operation can face problems and create risky conditions for the patient, which might end in revision operations. In this work, methods are investigated to reduce one of the main reasons for the problems faced in reverse shoulder arthroplasty. This main reason is the wrong positioning of the K-wire which itself results in the wrong positioning of the implant baseplate. The wrong positioning can reduce the range of motion of the shoulder or can lead to complete malfunctioning of the joint. Using pre-operative planning of the surgery, and patient-specific instrumentation, are the solutions evaluated for improvement of the condition and reducing the risk of malpositioning. Preoperative planning which is deciding the correct choice of the procedure before the operation based on the patient, injury type, facilities available, and surgeon's skills, is an important method in improving implant positioning. Preoperative planning can be performed using two-dimensional images of the patient, but use of three dimensional images and computer preoperative planning software tools can improve planning. Patient-specific instrumentation which is a modern orthopaedics technique, uses Computed Tomography or Magnetic Resonance Imaging of a specific patient to create customized guides preoperatively. Prostheses or guides that are designed based on the specific anatomy or injury of a patient provide an opportunity to be implemented more precisely and hence can help improve implant positioning and reduce the risk of complications resulting from malpositioning. The PSI guide generation process is performed using software tools that perform preoperative planning on the three-dimensional models of the patient's data. A new open-source software tool, that provides preoperative planning capability for the surgeon, and also creates a patient-specific guide for K-wire positioning, is developed in this work to test the presented solutions. First, the development of the software, using only open-source platforms, is explained, then using the results of the software an experiment is designed and performed. The experiment evaluated the accuracy of the software results and also compared the results with other existing methods. The experiment contained five different shoulder anatomies and glenoid types. For each type ten different samples were manufactured. Two experienced surgeons experimented on the manufactured bone models and the results were evaluated to differentiate between different anatomies. The results were evaluated to control the version angle, inclination angle, and the entry point location of the K-wire after the experiment. The evaluation of the results presented that this proposed method has good accuracy for all three parameters. Also, the results showed better outcomes for specific types of anatomies when compared to the freehand method and the conventional guide method.
Development and control of an active torsional vibration damper for vehicle powertrains

(Graduate School, 2021-05-25) Yüceşan, Alişan ; Mugan, Ata ; 518152002 ; Mechatronics Engineering

The emission regulations on internal combustion engines (ICEs) have become more stringent and the importance of fuel efficiency has enhanced due to environmental pollution concerns. As a result, studies on optimization of ICEs put forward the consideration of downsizing, downspeeding and turbo supercharging concepts in designing modern ICEs and powertrains. Despite their numerous advantages, they result in boosted engine torsional vibrations which demands innovative vibration isolation solutions. Such a design solution should be uncomplicated and simple from an automotive manufacturer's point of view, meanwhile, be an extreme performer besides being a cost-effective solution. Passive and active dampers have been utilized to suppress torsional vibrations in the literature. At this point, the passive dampers appear more preferable at the first sight due to their cost advantage while active systems have the disadvantages of having higher costs due to the presence of actuators, sensors and peripherals, advanced complexity and potential of lower efficiency. But conventional passive torsional vibration damper systems like dual-mass flywheel (DMF) reached their limits and they are no longer able to sufficiently isolate the torsional vibrations of the state-of-the-art ICEs. Also, as seen in the literature survey section, there is no kind of active dampers for torsional vibrations which is equipped alone or proposed to be used in this context for vehicle powertrains and there is a huge demand for a new design solution to beat the performance constraints of passive isolation systems.
Energy-efficient velocity trajectory optimization using dynamic programming for electric vehicles

( 2021-10-22) Kızıl, Abdullah ; Sezer, Volkan ; 518161030 ; Mechatronics Engineering ; Mekatronik Mühendisliği

The electrification and autonomous systems developed in the automotive industry in the last decade bring different solutions. Many methods have been developed and still continue to be developed to reduce energy consumption in vehicles, especially with electrified, connected vehicle technologies and navigation systems. Speed trajectory optimization is part of these methods. The main motivation of speed trajectory optimization is to prevent excessive energy consumption due to driver driving style. In order to prevent this, information such as the slope and speed limit of the road to be traveled is used over the navigation system. When we consider only energy while optimizing the speed trajectory, the prolongation of the driving time will appear as a concern. Because if the vehicle goes faster, the energy consumed will increase quadratically. Therefore, optimization will always demand the vehicle to go slower in order to consume less energy and there must be a balance between energy and travel time. In this thesis, a study has been carried out that periodically updates the speed trajectory, which will ensure that the destination point and arrival time information are provided into the navigation system by the driver while consuming the least energy in the given time. The dynamic Programming (DP) method is used to solve this problem. Dynamic programming always presents the global optimum behavior under the given boundary conditions. The speed of the vehicle was used as the only state variable and its optimization was performed separately over the distance stages. The average speed required to reach the destination on time, based on the destination point and travel time information obtained from the navigation system, is given as an input to the optimization, and the DP state space is constantly updated. The main reason for this is to reduce the memory load required by DP. Thus, a fixed number of states are scanned. But the scanned range values are updated according to this speed input. A longitudinal vehicle model was used for optimization. The limits of the powertrain are also part of the optimization as a boundary condition. Before the optimization is run, a pre-calculation is also made to include the states where the transition between states is possible only in the optimization. Thus, it is aimed to shorten the calculation time by not including the unreachable situations in the optimization. Optimization takes place along a certain horizon. The speed trajectory calculated for this horizon is transmitted to the vehicle speed control unit as an input. The vehicle follows this speed profile. The optimization is updated again after a certain period of time and transmits the speed trajectory calculated for the next horizon to the vehicle. The purpose of this is if the vehicle cannot follow the given speed for any reason during real driving, the optimization is performed again based on the new conditions. This allows the vehicle to progress in real-time using the speed trajectory closest to the global optimum. In the study, simulation and analysis of the all-electric truck were carried out on two different slope routes. Tests were performed with different fixed velocity values and velocity profiles produced by velocity trajectory optimization in both routes. As a result of the simulations carried out, it has been observed that up to 4% of energy consumption and up to 2.5% of the targeted time are saved. Thanks to the proposed adaptive weight factor, it has been observed that the time-energy balance is maintained for different routes, arrival times, and vehicle parameters.
Predictive powertrain control

(Graduate School, 2022) Lalek, Ufuk ; Üstoğlu, İlker ; 755241 ; Mechatronics Engineering Programme

The main purpose of this study is to design predictive powertrain controller that finds optimal gear and optimal pedal position to minimize the fuel consumption based on the given road data. Model predictive control (MPC) method is used as the controller. Dynamic programming (DP) method is selected as the optimization solver of the algorithm due to the discrete structure of the gear states. Within the scope of this thesis, it is also aimed to develop a truck model with automated manual transmission (AMT) gearbox that can be controlled via MATLAB. The simulation environment is AVL Cruise and MATLAB/ Simulink. AVL Cruise is a powertrain simulation program that allows users to model a variety of powertrains using a specific library of modules. It is used to create the plant model and test it out with various drive cycles. The plant model is evaluated using MATLAB/ Simulink via interface block in the AVL Cruise program and transmission control algorithm which is developed in MATLAB based on the model predictive control method.
Retrieving and classifying emergency situations for smart home applications

(Graduate School, 2022) Uçar Mavi, Seher ; Doğan, Mustafa ; 738349 ; Mechatronics Engineering Programme

One of the uses of Bluetooth, which is a wireless communication technology, is to communicate with smart devices in order to increase the functionality of not very advanced electronic devices and to carry data that appeals to the user. Bluetooth has also paved the way for devices without internet access to access the internet at a much lower cost by communicating wirelessly with devices with internet access. Data transferred to the Internet can be stored and used for any purpose. However, while doing this, user data must be kept encrypted by law and should not be shared with third parties. In this study, it has been investigated how user data can be shared with authorized institutions for emergencies. A mesh network was created using the Cypress BT-Mesh kit. Each of the 4 development boards is programmed to simulate a smart home device. Cypress's Software Development Kit(SDK) was used during programming. In addition, an Arduino GPS system included in the system is used to receive location data. The RSA (Rivest-Shamir-Adleman) algorithm is used to send all user data encrypted to authorized institutions in case of emergency. Thus, both the confidentiality was not violated and the emergency situation was reported to the authorized institutions and the user. The RSA algorithm was executed on the Espressif processor and the ESP-IDF SDK was used. Assuming that the decryption process is carried out in the cloud environment of the institutions, a scenario in which institutions intervene by looking at the lookup table and comparing them is considered. In this study, the scenario of notifying the fire department in case of fire and informing the police in case of a thief is simulated. In the simulation, it is shown how user data can be sent in an encrypted way. When the functioning of the network and the transmission of data are examined, it is concluded that the RSA algorithm provides a high-security solution for smart home applications. Thus, it has also been proven that a recovery notification use-case can be added to a smart home system without violating privacy.
Design and control of ros based omnidirectional vehicle

(Graduate School, 2022) Nalbant, İbrahim Dinçer ; Temeltaş, Hakan ; 710624 ; Mechatronics Engineering Programme

Nowadays, with the fact that the hardware and software of mobile robots are cheaper and easier to find, studies on them have increased, and as a result of research and development activities carried out by academia and large companies, mobile robots have been widely used in the areas such as exploration, humanoids, drones, automation, transportation, space missions, patrolling, search and rescue and service robots. In addition, a wide variety of driving systems have been designed to increase the driving characteristics of mobile robots, as well as sensors and software architectures to make them autonomous. In this thesis, a mecanum wheeled, ROS-based autonomous, omnidirectional vehicle prototype with high load carrying capacity, called ITU omnidirectional vehicle, has been designed and produced to be used in future research and development activities at the ITU Robotics Laboratory. ITU omnidirectional vehicle is capable of moving to all directions on the ground plane without changing its heading angle, with the help of its specially designed mecanum wheels. Mecanum wheels are driven individually by four Maxon DC motors. For the autonomous navigation, there are two Hokuyo LIDAR sensors and a Xsens IMU sensor mounted in the vehicle. Vehicle inverse and forward kinematic analyses were carried out according to the mecanum wheel kinematics and chassis dimensions, then these equations were applied to the control software of the vehicle. DC motor selection of the vehicle was made according to the payload and acceleration values needed. A special drivetrain was designed between motors and wheels to reduce the lateral force, vibration and balancing effects of the mecanum wheels to the motors. Electrical design of the vehicle was made by using DC regulators and supplying appropriate power for the hardware. For the software of the vehicle, open source Robot Operating System(ROS), was integrated to the vehicle, so that various control, localization, path planning and mapping algorithms could be developed and tested. Both simulation and experimental results of different path following scenarios for the vehicle is presented in the last chapter of this thesis. Results of these case studies satisfied the requirements of the thesis.
Nominal capacity calculation for lithium-ion batteries with advanced algorithms

(Graduate School, 2022) Nalbant, Harun ; Gökaşan, Metin ; 735121 ; Mechatronics Engineering Programme

The speed of global warming and demand for energy increased rapidly after Industrial Revolution by high energy usage for the current high-level technologies. The automotive industry is one of main industries demanding high energy resources to grant quality transportation. However, due to this high usage of energy, the Carbon emissions rapidly increased due to fossil-fuel usage. Many countries adjusted their emission regulations to decrease the devastating effects of global warming. Therefore, many automotive producers started to search for technologies to comply with these strong regulations. Electric vehicles emerged from these regulations. In addition to that, the comfort and the decreased complexity of the design in the electric vehicles are also very appealing for the people which means lower effort on the service needs. Therefore, it is expected that the number of electric vehicles (EVs) will increase on the way. The main components for the electric vehicle are the propulsion and energy storage components. As a driver, there are 2 main parameters selecting a vehicle: Propulsion Power and Range. Propulsion power could be granted with a good electric motor and inverter design. On the other hand, EV batteries play a key role in increased range. The majority of the people generally focus on the range and usage instead of high-powers. Therefore, the point of focus in this study is the EV batteries. EV batteries are one of the main and most expensive components in the vehicle. The drivers mainly tend to use the vehicle as much as they can without any component change. However, as time goes on, the battery gets older and its performance would be decreased. Furthermore, an aged battery could harm the driver and the passengers. As battery developers tend to avoid any unwanted situation, the batteries are restricted to be used for a limited time. This limited-time is mainly dependent on battery State-of-Health (SOH). The SOH is the main indicator in a Battery Management System (BMS) to monitor the battery. The developers target to ensure a safe range for battery usage. On the other hand, the driver would like to challenge the limits to use the battery longer to avoid frequent battery-related service costs. State-of-Health of the battery determines the battery life cycle. If a battery is aged due to driving, temperature and cycle conditions, then, the usage of that battery is restricted by the Battery Management System (BMS). To avoid failure due to ageing, there are some mechanisms to warn the driver before the battery fails so that any kind of incident is prevented. This restriction relies on the accuracy of the SOH calculation. In the automotive industry, the battery manufacturers' warranty period mainly depends on the battery SOH. In other words, if the actual SOH of the battery is lower than the guaranteed value, then, the warranty period is over. This SOH value is defined under the worst-case scenario. In other words, the calculation accuracy for SOH is added as an offset to the target value of the manufacturer. The worse value of SOH calculation accuracy would result in an earlier finish of the warranty period. In other words, better accuracy would result in a longer battery usage period. Better accuracy results in a longer warranty period. For example, the state-of-art batteries are generally guaranteed for 5 years under the condition SOH is calculated with %5 accuracy. If this accuracy is increased by ~% 3, then the battery life would be guaranteed for 1 more year. Since the battery technologies are growing, the warranty period could be more than 5 years which results in a higher increase based on SOH accuracy improvement. There are 2 main indicators for SOH: Capacity and Internal Resistance.
Automatic landing with model predictive control

(Graduate School, 2022) Ulukır, Talha ; Üstoğlu, İlker ; 518191029 ; Mechatronic Engineering Programme

In flight control systems, the landing maneuver is one of the most critical time periods for the aircraft, and it is of great importance to both respond to the disruptors and ensure durability in this time period when the disruptor activity is high. Within the scope of this thesis, 4 different controller type automatic landing systems were designed for a twin-engine passenger aircraft, this landing system provides fully automatic landing in both longitudinal and lateral planes. Within the scope of the thesis, different control architectures in the literature for the automatic landing autopilot were examined. Within the scope of the thesis, the change of system inputs and system outputs as a result of linearization under different conditions has been examined. The consistency of the nonlinear model of the aircraft with the linear model was compared, and this comparison was made by examining the behavior of the system variables in response to the binary commands given to the control surfaces. Within the scope of the thesis, what the sub-phases of the automatic landing autopilot are and according to which criteria and conditions these sub-phases are separated from each other are examined. The classical control architectures in the flight control system (stability-enhancing system and control-enhancing system) are discussed, and for what purposes and with what standards these architectures are designed. In fixed-wing aircraft control systems, the longitudinal and lateral states of the system are separable. In the scope of the thesis, the automatic landing architectures in the literature for these separated states are examined. The controllers designed for the descent system in the thesis are: PID, Linear Quadratic Integral, Model Predictive Control and Algebraic Model Predictive Control architectures. One of these four different control architectures (PID) is in a single-input-single-output control structure, while the other three control architectures (LQI, MPC, AMPC) are in a multi-input-multi-output control structure.
Real time pedestrian tracking using adaptive kalman filter

(Graduate School, 2022) Vural, Coşkun Orkun ; Doğan, Mustafa ; 518181042 ; Mechatronics Engineering Programme

Image processing has been one of the hot topics since early 1950's and it has developed from simple researches on the images to real time video processing. This development brought more challenges to the researches as the environment around became a complex variable to process. Thus new techniques with low computing time and flexible algorithms that adapt to changes in the environment emerged. Human detection and estimation of their movements is one of the research areas described before. For security and emergency situations, applications of real time human detection is fundamental. In this work it was aimed to develop a robust system that can predict human movement on real time video. The motivation behind the work is to design a system that can detect pedestrians and their movement to warn the driver beforehand. The system needed to do the lowest amount of computation time as possible and it needed to adapt to changes of the environment in order to work in real time with unstable surroundings. Thus an Adaptive Kalman algorithm is developed for prediction of the next steps as only the previous steps information was needed. For the development of the project Matlab is used as the programming language. The reason behind choosing Matlab is due to the fact that Matlab IDE includes many libraries and toolboxes useful for this work. As the first step of the project the camera input is evaluated in real time by taking snapshots and further algorithms are applied on the frames taken. For human detection the built in Matlab library which gets data from the Caltech Dataset is used. The data gathered from the human detection algorithm is used to find the centroids in the region of interest. Centroids are fed to the Adaptive Kalman Filter as the input data with corresponding error parameters. The term Adaptive Kalman Filter comes from the responsive error tuning from the previous input. P, Q and R parameters are tuned with the error values of the previous frame. It was observed that all P, Q and R parameters and K value is converged with the increasing number of frames. Thus the learning process is also included in the work. In the 1000 test frames the system is tested with single and multiple pedestrians in the frame. Out of total 1678 samples 1418 of them are the right classifications which results in a %84.5 success rate. In the test process, it was observed that the error is mainly caused when two objects move towards each other. From that observation it can be concluded that the system works better in areas where the crowd is less dense.
Paletli araçlar için otomatik transmisyonun dinamik modeli, kavrama parametre adaptasyonu ve kontrolü

(Lisansüstü Eğitim Enstitüsü, 2022-02-17) Arı, Ali ; Yalçın, Yaprak ; 518171037 ; Mekatronik Mühendisliği

Bu tez çalışmasında; paletli askeri araçlarda kullanılan otomatik transmisyonların vites değiştirme fonksiyonlarını gerçekleştirecek elektronik kontrol ünitesi kontrol algoritmaları tasarlanmıştır. Bu algoritmaların, transmisyonun dinamik modeli ile desteklenmesi sayesinde planet dişli sistemindeki kavramalarda herhangi bir hız ve basınç sensörü olmadan kontrol için gerekli değişkenler hesaplanmıştır. Kavramaların aşınması durumunda, kavrama parametreleri, yalnızca transmisyon giriş ve çıkış hızı sensörlerinden gelen bilgileri kullanan bir adaptasyon algoritması aracılığıyla kestirilmiştir. Bir yenilik olarak, atalet fazında dinamometredeki kalibrasyon işlemlerini kısaltacak, dayanıklı ve hassas kontrole olanak sağlayacak, dinamikleri doğrusal olmayan transmisyon sisteminde kullanılabilen, genetik algoritma ile optimize edilmiş bir PID kontrol tasarımı yapılmıştır. Tasarlanan dinamik model, uyarlama algoritması ve kontrolör, bir MATLAB/Simulink-Simscape güç aktarma sistemi simülasyon modeli üzerinde test edilmiştir.
Image reconstruction with deep learning and applications in MR images

(Graduate School, 2022-04-22) Aghabiglou, Amir ; Ekşioğlu, Ender Mete ; 518172001 ; Mechatronics Engineering

In this thesis, in the first step, the novel application of the U-Net structure was considered for the important inverse problem of MRI reconstruction. Deep networks are particularly efficient for the speed-up of the MR image reconstruction process by decreasing the data acquisition time, and they can significantly reduce the aliasing artifacts caused by the undersampling in the k-space. On the first try, it is aimed to develop a novel and efficient unfolding U-Net framework for reconstructing MR images from undersampled k-space data. The new framework should have improved reconstruction performance when compared to competing methodologies. In this step, a novel unfolding framework utilizing the U-Net as a sub-block is being proposed. The introduced U-Net unfolding structure is applied to the magnetic resonance image reconstruction problem. The connection between the unfolding U-Nets is realized in the form of a recently developed projection-based updated data consistency layer. The novel structure is implemented in the PyTorch environment, which is one of the standards for deep learning implementations. The recently created fastMRI dataset which forms an important benchmark for MRI reconstruction is used for training and testing purposes. Despite the many challenges in training rather large networks, novel methodologies have enhanced the capability for having clinical-grade MR image reconstruction in real-time. In recent literature, novel developments have facilitated the utilization of deep networks in various image processing inverse problems. In particular, it has been reported multiple times that the performance of deep networks can be improved by using short connections between layers. In the next step of this thesis, a novel MRI reconstruction method is introduced that utilizes such short connections. The dense connections are used inside densely connected residual blocks. Inside these blocks, the feature maps are concatenated to the subsequent layers. In this way, the extracted information is propagated until the last stage of the block. The efficiency of these densely connected residual blocks was evaluated in MRI reconstruction settings, by augmenting different types of effective deep network models with these blocks in novel structures. The quantitative and qualitative results indicate that this original introduction of the densely connected blocks to the MR image reconstruction problem improves the reconstruction performance significantly. In addition, a novel densely connected residual generative adversarial network (DCR-GAN) is proposed for fast and high-quality reconstruction of MR images. DCR blocks enable the reconstruction network to go deeper by preventing feature loss in the sequential convolutional layers. DCR block concatenates feature maps from multiple steps and gives them as the input to subsequent convolutional layers in a feed-forward manner. In this new model, the DCR block's potential to train relatively deeper structures is utilized to improve quantitative and qualitative reconstruction results in comparison to the other conventional GAN-based models. It can be see from the reconstruction results that the novel DCR-GAN leads to improved reconstruction results without a significant increase in the parameter complexity or run times. The GAN-based structures generally suffer from some limitations. They are slow in convergence and they are unstable during the training step. In this work, these limitations of GAN also was addressed by proposing a new wavelet-based structure. To accomplish this, the wavelet transform packet was incorporated into the GAN structure. The wavelet transform is used in the encoding and decoding steps to create this model. In another word, the downsampling and upsampling layers were replaced with Discrete Wavelet Transform (DWT). DWT is used to replace each pooling process during the contraction phase. As DWT is a reversible package, this downsampling approach guarantees that all information can be retained. DWT can also record both the frequency and position information of feature maps, which will aid in the preservation of fine texture. The inverse wavelet transform is employed in the expansion step to upgrade the size of feature maps. Moreover, recent breakthroughs in this field have inspired us to propose another novel deep unfolding structure for MR image reconstruction. In the last step, the model was trained using not only an iteration of the image itself but also utilizing an updated noise level parameter. The noise level parameter is calculated at each iteration using the error between the network output and the initial zero filling estimate. This new parameter is given as an additional input to the network, and it acts as an evolving regularizer for the image manipulation strength of the network over the unrolling iterations. The introduction of this adaptivity over iterations in the training step also improves the deep models reconstructed image quality in the inference stage. Empirical results indicate that the recommended technique can convergence to better reconstruction results when compared to state-of-the-art unfolding structures devoid of such an adaptive parameter. The introduction of the additional adaptive parameter results in an incremental increase in the parameter complexity, and the required reconstruction times also stand very similar. In this thesis, both quantitative and qualitative results were provided and the proposed model's results were evaluated with cutting-edge techniques in the MR image reconstruction field. Three commonly used evaluation metrics of PSNR, SSIM, and NMSE were used to evaluate simulation results. The statistical differences between developed techniques are investigated using the one-way ANOVA method. Additionally, a t-test is used to specify the major difference between the means of the two proposed structures. Additionally, the robustness of the proposed densely connected residual models was verified by testing them with another dataset type without retraining them. The other dataset differs in size and body tissue type compared to the training dataset. The suggested novel structures in this thesis are improved MR image reconstruction performance compared to state-of-the-art techniques regarding all evaluation metrics. They proved their capacity for reconstructing high-quality images. More importantly, the thesis goal was satisfied regarding the acceleration of MR imaging. The proposed models in this thesis are generally considered to be fast enough to be used even in real-time medical imaging.
İleri sürücü destek sistemleri için bir fonksiyonel güvenlik uygulaması

(Lisansüstü Eğitim Enstitüsü, 2022-05-26) Çağlayan, Ebru ; Kurtulan, Salman ; 518201010 ; Mekatronik Mühendisliği

İleri Sürücü Destek Sistemleri üzerinde ISO 26262'nin halihazırda var olan versiyonunun yetersizliği sonucunda ortaya çıkan yeni bir fonksiyonel güvenlik metodolojisi gerekliliği sonucunda tezin ilk amacı ortaya çıkmıştır. Bu amaç, ISO 26262'nin bu yeni teknolojiye uyarlanabilir bir versiyonunu ortaya koymaktır. Bunun için, var olan ISO 26262 standardizasyonuna, İleri Sürücü Destek Sistemleri ve tarihçesine ve ISO 26262'nin uygulama adımlarına değinilmiştir. Akabinde, bütün bu bilgiler ışığında İleri Sürücü Destek Sistemlerinde yaygın bir sistem olan İleri Acil Frenleme Sistemi (AEBS) fonksiyonu üzerinde örnek bir çalışma gerçekleştirilerek uyarlanan metodoloji tanıtılmıştır. İleri Acil Frenleme Sistemi, genellikle radar ve kameranın birlikte kullanılarak otomobil, motosiklet, yaya ve bisiklet gibi hedeflerin birlikte algılandığı sensör füzyonunun aktif olarak kullanıldığı bir fonksiyondur. Radarlar otomobilleri seçme konusunda daha efektif iken, kamera yaya tipi hedefleri seçmede daha büyük başarı göstermektedir. Bütün bu algılamanın gerçekleştiği karmaşık sistemlerde fonksiyonel güvenliğin gerekliliği tartışılmazdır. Bununla birlikte yollarda meydana gelen kazaların çoğunluğunun sürücü dikkatsizliğinden kaynaklandığı düşünülürse İleri Acil Frenleme Sistemi gibi akıllı bir teknolojinin fonksiyonel güvenliğin gerçekleştirildiği koşullarda trafik kazalarını büyük ölçüde engelleyeceği gerçeği yadsınamazdır. Dolayısıyla tezde otomotiv sektörüne daha büyük bir fayda sağlaması bakımından örnek fonksiyon olarak İleri Acil Frenleme Sistemi fonksiyonu seçilmiştir. Tezde ikincil bir amaç olarak hem İleri Sürücü Destek Sistemlerinin hem de fonksiyonel güvenliğin sektörde önem kazanmasıyla birlikte trendleşen bu iki alanın terminolojisinin Türkçe bir tez vasıtasıyla literatüre kazandırılarak Türkçe bir kaynak elde etmek hedeflenmiştir. Bu bağlamda ayrıntılı görseller ve kısaltmalarla tez Türk otomotiv sektörü için bir kaynak oluşturmaktadır.
Çamaşır makinelerinde yapay sinir ağları ile yıkama performansı ve enerji tüketiminin modellenmesi

(Lisansüstü Eğitim Enstitüsü, 2022-06-23) Aktaş, Yakup ; Altınkaynak, Atakan ; Kalafat Acer, Merve ; 518181036 ; Mekatronik Mühendisliği

Dünyada gelişen teknoloji ve mühendislik yetkinlikleriyle beraber endüstride bir rekabet ortamı oluşmuştur. İlgili sektör üreticileri fark yaratan ürünler otaya koymak için gelişen teknolojik adımları yakalamak ve ürünlerine değer katacak gelişmeleri takip etmek durumundadır. Özellikle tüketicinin doğrudan etkileşim halinde olduğu beyaz eşya ürünlerinde fark yaratan teknolojiler ön plana çıkmaktadır. Ancak ürünlere eklenen birçok yeni özellik beraberinde maliyetleri de doğurmaktadır. Aynı zamanda üreticiler için kaynak ve zaman yönetimi anlamında da ekstra yük getirmektedir. Bu nedenle ürünlerin ar-ge, tasarım ve üretim süreçleri ne kadar iyileştirilebilirse sektörde rekabetçi ve yenilikçi ürünler ortaya koyabilmek o kadar mümkün hale gelecektir. Üretilecek olan ürünlerin ar-ge ve tasarım aşamalarındaki test süreçlerinin iyileştirilmesi maliyet, kaynak ve zaman açısından üreticiler için olumlu katkı sağlamaktadır. Çamaşır makinaları günümüzde yaygın olarak kullanılan dayanıklı tüketim aletleridir. Su ve elektrik enerjisi ile çalıştıkları için, test süreçlerinde her bir çevrimdeki bu tüketimler ek maliyetlere ve aynı zamanda dünya kaynaklarının da tüketilmesine yol açmaktadır. Bununla birlikte zaman açısından da yeni ürün proje süreleri uzamakta ve teknolojik gelişimi yavaşlatmaktadır. Yani test süreçlerinin kısalması, hem sürdürülebilirliğe katkı yapacak, hem maliyetleridüşürecek hem de zamanın verimli kullanılmasına yol açacaktır. Tez kapsamında çamaşır makinalarının ar-ge ve tasarım süreçlerinde gerçekleştirilen test metotlarının kurulacak model yapısında incelenmesi ile çevresel sürdürülebilirliğe katkı sağlanması, üretici maliyetlerinin düşürülmesi ve zaman tasarrufu elde edilmesi amaçlanmaktadır. Çamaşır makinalarının sahip olduğu özelliklerin yanında, standartlarca belirlenmiş çeşitli sınırları da sağlıyor olması gerekmektedir. Bunlardan biri yıkama performansıdır. Çamaşır makinalarının temel özelliği olan yıkama işlemi, standartlarda belirlenmiş yöntemlerile ölçülebilmektedir. Üretilen çamaşır makinalarının da belirlenen limit değerin altına düşmeyecek etkinlikte yıkama performansına sahip olması gerekmektedir. Üreticiler ise bu sınır koşulu sağlayıp sağlamadığını test etmek için standart yıkama performansı testlerini laboratuvar oertamında gerçekleştirmektedir. Ancak farklı sınır koşullarından dolayı yıkama performansını sağlayabilmek adına birçok parametrenin optimize edilmesi gerekmektedir. Birden fazla parametrenin etki ettiği yıkama performansı hedef değerini yakalayabilmek adına yapılan bu deneme testleri ise su ve enerji tüketimlerinden dolayı beraberinde ekstra bir yük getirmektedir. Bu sebeple kurulacak model yapısı ile bu test sonuçlarının tahmin edilebilmesi hedeflenmektedir. Diğer bir yandan, standart olarak sağlanması gereken yıkama performansının belirli enerji tüketimi sınırları içerisinde gerçekleşiyor olması gerekmektedir. Üreticiler, üretilen çamaşır makinasının enerji tüketiminin, standartlarda belirlenen enerji sınıf aralıklarından hangisine denk geldiğini deklare etmek durumundadır. Doğal olarak daha düşük tüketime sahip enerji sınıfındaki ürünler son kullanıcı tarafından daha çok tercih edileceğinden yıkama performansı değerine olabilecek en düşük enerji tüketimi ile ulaşmak ana hedeftir. Bu nedenle yapılan performans testleri yerine yıkama performansını tahmin edecek model ihtiyacının yanında, optimum tasarımın yapılabilmesi için enerji tüketiminin de tahmin edilmesi gerekmektedir. Kurulacak enerji tüketimi modeli ile de enerji tüketimi değerinin test yapmadan tahmin edilebilmesi amaçlanmaktadır. Tez kapsamında kurulacak yıkama performansı ve enerji tüketimi tahmin modellerini elde edebilmek için öncelikle deneysel veriye ihtiyaç vardır. Bu amaçla laboratuvar ortamında deney istasyonları hazırlanmış ve standart yıkama performansı test sonuçları tüm analog ve dijital verileriyle birlikte toplanmıştır. Tahmin edilmek istenen yıkama performansı ve enerji tüketimi değerlerinin yanında model yapılarını girdi sağlayabilecek parametrelerin de değişimleri kaydedilmiştir. Toplanan verilerin analizi yapılarak yıkama performansı ve enerji tüketimi tahmin modelleri için ayrı ayrı girdi parametreleri seçilmiş ve çeşitli model yapıları oluşturulmuştur. Oluşturulan yapılardan en iyi performans gösteren modeller seçilmiştir. Elde edilen modeller sayseinde yıkama performansı ve enerji tüketimi için seçilen girdi parametresi değerleri verildiğinde yüksek doğrulukta sonuçlar alınmaktadır. Tezin ilk bölümünde literatürde çamaşır makinalarında gerçekleştirilen yıkama prosesine etki eden temel parametrelerden bahsedilmiştir. Ayrıca tezin ilk bölümünde çamaşır makinlarında geliştirilmiş makine öğrenmesi, yapay sinir ağı ve bulanık mantık algoritma çalışmalarından örnekler sunulmuştur. Yapılan çalışmalarda tahmin edilmesi kritik parametrelere yer verilmiş ve farklı yöntemler kıyaslanmıştır. Tezin ikinci bölümünde yıkama performansı ve enerji tüketimi modellerine veri girişi sağlamak amacıyla kurulan deney sisteminden, kullanılan ekipmanlardan ve ölçüm yöntemlerinden bahsedilmiştir. Bu bölümde ek olarak toplanan deneysel veri kümesi incelenmiştir. Verilerin makina özellikleri açısından yıkama performansı ve enerji tüketimine göre dağılımları gösterilmiştir. Tezin üçüncü bölümüden yıkama performansı modeli için girdi parametreleri seçilmiştir. Girdi parametrelerinin çıktı değerine etkileri detaylıca açıklanmıştır. Parametrelerin istatistiksel özellikleri elde edilmiş, girdi-çıktı parametreleri arasındaki lineer korelasyon ilişkileri çıkarılmıştır. Tezin bu bölümünde lineer yöntemlerin problemi çözümlemeye yetmeyeceği ve makine öğrenmesi yöntemlerinin denenmesi gerektiği yapılan lineer regresyon analizleri ile vurgulanmıştır. Bu amaçla aynı bölümde modelleme için kullanılacak yapay sinir ağları ile Levenberg-Marquardt geri yayılım algoritması açıklanmıştır. Kurulacak modelin algoritma parametreleri detayları ile verildikten sonra farklı katman ve nöron sayılarındaki yapay sinir ağı sonuçları elde edilmiş ve en iyi performansı veren modeller belrtilmiştir. Yapay sinir ağı modeli Matlab programı kullanılarak Levenberg-Marquardt geri yayılım öğrenme algoritmasının model parametre detayları değiştirilerek oluşturulmuştur. Tezin dördüncü bölümünde de yıkama performansı yapay sinir ağı modeline benzer şekilde enerji tüketimi modeli için de girdi parametreleri belirlenip lineer korelasyon ilişkileri belirtilmiştir. Lineer regresyon analizi sonuçları paylaşılmış ve enerji tüketimi modeli için de yapay sinir ağı modeli kurulmuştur. Yıkama performansı yapay sinir ağı modeli ile aynı ağ yapısı özelliklerinde modeller karşılaştırılmış ve en yüksek performansı veren model seçilmiştir Tezin beşinci bölümünde elde edilen model yapıları, ortak model arayüzü oluşturmak adına Simulink ortamına aktarılmış ve tasarım süreçlerinde kullanıma hazır hale getirilmiştir. İlgili girdi parametrelerinin değerleri verildiğinde elde edilen en iyi modellerin tahmini sonucu yıkama performansı ve enerji tüketimi değerleri elde edilebilmektedir. Tezin beşinci ve son bölümünde ise yapılan tez çalışmasının sonucuna ve gelecek çalışmalar için önerilere yer verilmiştir.
Dynamic weighing method for checkweigher

(Graduate School, 2022-06-23) Gülbaş, Mustafa Can ; Yalçın, Müştak Erhan ; Ayhan, Tuba ; 518191025 ; Mechatronics Engineering

The main purpose of this study is to examine the checkweigher system. It is necessary to accurately estimate the weight of the product in motion with a checkweigher. This is an estimation process, as the measurement data obtained in motion will never be as clean as the static measurement data. The measurement error tolerance is limited by state regulations. Therefore, a dynamic weight measurement system should obey particular regulations in order to be used in industry. The automatic weight control system, checkweigher, which is used in many areas from production to shipment in the industry, consists of three conveyor belts, at least two photocells, load cell, processor, control screen for the user and rejector/router arms. In the system, the product carried by the in-feed conveyor belt is guided by the output conveyor belt after passing over the conveyor belt connected to the load cell. At this point, at the time of passing over the conveyor belt connected to the load cell, data is received by the load cell at a sampling frequency of 1600 Hz and sent to the processor for processing. The working principle of the load cell depends on the stretching and compression state of the resistors called strain gauges. When force is applied to a load cell, some resistors compress while others flex. When this change is converted to voltage, an inference can be made about the weight of the product. There are many noise factors that affect the measurement signal. The time-variant low-pass filter in the cascade form can effectively filter out the noise from the load cell signal, but quite a lot of filters are required to achieve high accuracy. In this study, a different approach was tried with a time-variant low-pass filter in order to accelerate measurements. The number of cascade form low-pass filter is optimized to shorten the response time while providing regulation-complaint measurement accuracy. By applying the filter, it is aimed to reach the mass of the product from the oscillations with minimum number of filters. The maximum speed obtained within the error limits was specified. As a result, by reducing the number of filters and increasing the damping, the weight data of product from the oscillations were reached faster within the error limits given in the regulation.
Design and evaluation of energy management systems for connected hybrid and electric vehicles

(Graduate School, 2022-07-04) Özdemir, Abdulehad ; Koç, İlker Murat ; 518132001 ; Mechatronics Engineering

Transportation is one of the most significant sources of emissions across various industries. With the effect of Paris Climate Agreement and the Green Deal, environmental concerns and technological progress push the development and market penetration of electric vehicles and hybrid electric vehicles. The number of electric and hybrid vehicles which can be considered as a stepping stone for electric vehicles are increasing day by day. On the other hand, transportation systems are becoming more efficient and safe by the improvement of the communication systems both on the vehicles and the infrastructure. There are significant improvements in connected and autonomous vehicles which has been started with the development of advanced driver assistance systems. The automotive industry, which plays a key role in the development of many accompanying technological ecosystems, is expected to be enhance more changes in the next 10 years than in the past 100 years. It is estimated that this transformation will dominance especially with the technologies progress in connected and autonomous vehicles. The main purpose of the study is to develop smart energy management strategies for connected, hybrid and electric vehicles and evaluate the benefits of developed smart energy management strategies. At the same time, the effects of the transition to electric vehicles in terms of energy consumption and environment is evaluated. For the optimization studies the Well-to-Wheels emission values are calculated and used in order to ensure apple-to-apple comparison. During the thesis study, three articles have been prepared and the preapared articles have been the substructure of the thesis. As of the date of submission of the thesis, one of the articles has been published and the requested revisions have been made for the other two articles and resubmitted. Prepared articles entitled as " Dynamic Programing Based Green Speed Advisory System Design for Mixed Platooning Vehicles", " Driving Cycle Based Energy Management Strategy Development for Range Extended Electric Vehicles " and " Comparative Study on Well-to-Wheels Emissions between Fully Electric and Conventional Automobiles in Istanbul". The article about the comparative study on Well-to-Wheels emissions has been published in the eighty-seventh issue of the "Transportation Research Part D" journal. Turkey's energy mix is analyzed and the emission factor of electricity production of Turkey is estimated in order to make appropriate comparisons during optimization studies. The Well-to-Wheels equivalent carbon dioxide emissions of the electricity is calculated. By considering energy sources, the Well-to-Wheels emission of Turkey is calculated as 520 g carbon dioxide equivalent per kWh. By using the carbon intensity of electricity, it is possible to compare the same variable for electric energy and fossil fuels for hybrid and electric vehicles. Vehicle models are created to use for model-based optimization studies. In order to develop an energy management system for serial hybrid vehicles, all critical subsystems are tested and a vehicle model which is validated by the test data is created. The model is developed by mathematical modelling of vehicle dynamics and testing the the electric motor, motor driver, battery cells and internal combustion engine. The developed models are validated by vehicle level testing on chassis dynamometer. A driving cycle based energy management strategy is developed for range extended electric vehicles to increase system efficiency and equivalent vehicle range. The results showed that; the optimized strategy can save CO2 emission by 6.21%, 1.77% and 0.58% for heavy, moderate and light traffic respectively. The usage of range extender in an efficient way by taking the traffic data into account extends the vehicle range, especially in heavy traffic conditions. For the hybrid vehicles which consumes both electric energy and fossil fuels, It will is important to compare the same value fort he objective function such as equivalent carbon dioxide emission. This study is a good example from this point of view. The developed energy management system will enable connected hybrid vehicles to be in more efficient way by using the route and traffic density information. In addition, vehicle emission maps are developed as a vehicle feature. The vehicles are tested on the chassis dynamometer and emission maps which are based on speed and wheel force are created. It is offered that vehicle emission maps can be used for optimization studies, especially in traffic with different types of vehicles. Considering that there are many ongoing studies on reducing tranportation based emissions, the use of the standardized emission maps are important for system level efficient use of connected vehicles. From this point of view, a multi-layer dynamic programing based optimizer is designed to minimize platooning Well-to-Wheels emissions of platooning vehicles where the platoon consists of an electric, a gasoline and a diesel vehicle. Vehicle emission maps and longitudinal dynamics are used for vehicle modelling. Tank-to-Wheels emission maps of internal combustion engine vehicles are produced by testing the vehicles on a chassis dynamometer. The optimization process has exploration and exploitation layers. The cost function is total Well-to-Wheels emission, design variable is speed trace, constraints are speed limits, traffic light states and vehicle accelerations limits. The test results show that the developed optimizer helps to achieve a 19.8% reduction in total Well-to-Wheels emissions for the defined use case. Thus, there is a significant emission saving potential in using speed advisory system for platooning vehicles through signalized intersections. On the other hand, driving cycles are used to examine the energy consumption and emission emissions of vehicles. In order to analyze the environmental effects of electric vehicles on a real driving cycle, a driving cycle has been developed for Istanbul by statistically analyzing the data collected on the determined routes. By using the developed driving cycle, the vehicle test are conducted. Acoording to the results electric vehicles emit 73.9 g carbon dioxide equivalent per kilometer on the same route, while gasoline vehicles emit 183.4 g equivalent carbon dioxide emissions. Therefore, the transition to electric vehicles should be strengthened by more widespread use of renewable energy in order to effectively reduce emissions associated with electric vehicles in general. At the same time, the results of this study can be a guide for policy makers. In summary, within the scope of the thesis electric carbon intensity of Turkey is calculated by considering Turkey's energy mix and Well-to-Wheels greenhouse gas emissions are analyzed both for conventional and electric vehicles are measured. A dynamic programing based optimizer is developed to decrease total Well-to-Wheels emissions of the mixed conventional and electric platooning vehicles through signalized intersections. Vehicle emission maps are generated both for electric and conventional vehicles for model-based optimization. A driving cycle based energy management strategy is developed for range extended electric vehicles to increase system efficiency and equivalent vehicle range. The vehicle model is developed by critical subsystem testing. An up to date driving cycle for Istanbul is developed (so called Istanbul Driving Cycle) by using collected traffic data across various sections of the city. An internal combustion engine vehicle and an electric vehicle are tested on a chassis dynamometer under the same conditions to determine specific energy consumption and specific emissions.
Human factor based advanced driver-assistance system (ADAS) design for electric vehicle

(Graduate School, 2022-07-06) Doğan, Dağhan ; Estrada, Ovsanna Seta ; 518122006 ; Mechatronics Engineering

Every year, thousands of traffic accidents occur and thousands of people die or are injured in these accidents. Considering the causes of accidents, it can be said that most of them are human errors. For this reason, studies focus on advanced driver assistance systems, increasing vehicle autonomy levels and driver behavior in traffic, and aim to prevent possible accidents. For a similar purpose, in this study, I aimed to collect data and analyze some human factor technologies that will support advanced driver assistance systems (ADAS) and to produce suggestions on how researchers and manufacturers producing ADAS can use these technologies. Our study focuses on the data of the galvanic skin response (GSR) sensor, which is a wearable sensor and aims to contribute to human factor studies by analyzing the GSR sensor and other sensor data collected from the drivers and the prototype electric vehicle. The study is experimental and requires a realistic vehicle and realistic driver data. Thus, first of all, we aim to design a novel, low-cost and open to development embedded data collection system for the research and education in human factor technologies and ADASs. Equipment used in this simultaneous data acquisition system: an electric vehicle with the power of 750W, Arduino Mega 2560 electronic card, a 10-turn Vishay 860 potentiometer used for steering angle data, the Tamura 300 A AC/DC hall-effect current sensor used for current (torque) data, Pololu force-sensing resistor (FSR) to detect force on the steering wheel and brake pedal, Seeedstudio GSR sensor to detect stress, MinIMU-9 v3 inertial measurement unit (IMU) to detect gyro, accelerometer, and compass, GY-NEO6MV2 global positioning system (GPS) to detect chassis velocity and position, Scancon 2RM 200 encoder to detect wheel velocity and Techsmart dashcam to record the environment and driver behavior. After designing the data collection system and implementing it for the prototype electric vehicle, we are looking for an answer to the question of whether the driver's stress in traffic can be detected with the GSR and FSR sensor data. We collect the GSR and FSR sensor data for 38 drivers using the designed data collection system in the Istanbul Technical University campus and analyze the GSR and FSR sensor data. In addition, a post-driving stress survey is used to improve the reliability and consistency of the stress level analysis and to validate the results. According to analysis results, the GSR sensor detects stress level-gender, stress level-driving experience, stress level-driving frequency and stress level-representative of normal driving behavior relationship, and the FSR sensor determines only gender stress level. Here, stress level-gender results for the GSR and FSR sensor, stress level-driving experience results for the GSR sensor and stress level-driving frequency results for the GSR sensor are consistent with the results of the survey with an accuracy of 100 %. Stress level-representative of normal driving behavior results for GSR sensor are consistent with the results of the survey with an accuracy of 50 %. As a result, the GSR sensor stress results are consistent with the results of the survey with a total accuracy of 87.5 %. The FSR sensor gender stress results are consistent with the results of the survey with an accuracy of 100 %. After the stress level detection study, we collect the IMU, FSR, GSR, current sensor, potentiometer, encoder and GPS data from 38 drivers along a route. Drivers are divided into 2 (risky and normal) classes according to their Euclidean distance from expert driver data for each sensor. The best classification methods are determined along this way for each sensor. Accordingly, all data are classified with the highest accuracy of 92.1% using the Medium Gaussian Support Vector Machine (SVM) method. IMU data is classified with the highest accuracy of 89.5% using the Artificial Neural Network (ANN) method. FSR data is classified with the highest accuracy with 94.7% accuracy using the Medium Gaussian SVM method. GSR data is classified with the highest accuracy with 97.4% accuracy using the Fine K-nearest Neighbors (KNN) method. Current data is classified with the highest accuracy with 100% accuracy using the ANN method. Potentiometer data are classified with the highest accuracy of 97.3% using the ANN method. Encoder data is classified with the highest accuracy with 92.1% accuracy using the Medium Gaussian SVM method. GPS chassis velocity data is classified with the highest accuracy with 94.7% accuracy using the Medium Gaussian SVM method. Thus, we can say that driver behavior is highly predictable for the batch data along a road. Secondly, it is tried to reveal whether the driver behavior we obtained along the above road can be detected instantly. GSR data of the drivers is analyzed individually because the GSR sensor gives the driver instant excitement and stress information. The driving videos of the drivers are shown to the expert driver. The faults and fault moments of the drivers are labeled by the expert driver. On the other hand, the data obtained by the GSR sensor are used to determine when the drivers are excited (stressed) and the reasons for stress are identified by the expert driver. In the analysis, driver-4 (male) and driver-7 (female) data are examined for individual classification. Stress moments are considered class-2 as dangerous situations. Others are considered class-1. In this way, classification methods are applied. As a result, it is found that the fault moments of the drivers are a subset of the stressful moments of the drivers for all drivers. For driver-4, all sensor data which is tagged by stress moments are classified with the highest accuracy with 97.4% accuracy using the ANN method. For driver-7, all sensor data is classified with the highest accuracy with 98.6% accuracy using the Bagged Tree method. Thirdly, we validate the driver status/behavior analysis above by detecting anomalies using Local Outlier Factor (LOF) values for GSR sensor data as a different method. This analysis provides the detection of a driver status with LOF anomaly values of GSR sensor data and other sensor support (camera and GPS) without the need for machine learning. Lastly in the chapter, we analyze the driving confidence in turns. The excitement increases obtained from the GSR sensor on turns have defined the unconfidence of the driver. The velocity and current data of the drivers determined by the GSR sensor in turns are examined and thus drivers are analyzed individually. When the first junction maneuvering data of drivers are analyzed based on the GSR sensor data, drivers numbered 7, 9, 20, 23, 27 and 34 fail at authorizing driving confidence on the first turn. When the second turn data are analyzed, drivers with an ID 9, 20, 23 and 38 could not show a confident drive on the second turn. Skin conductivity information, including abnormal, risky, and unconfident driving information, can be used for torque control of an autonomous electric vehicle. We transform the semi-autonomous electric vehicle into an electric vehicle with longitudinal autonomy. To improve the study, the distance sensor is also used simultaneously with the GSR sensor to detect collisions and intervene. It means that the GSR data is used to control a vehicle with closed-loop and longitudinal autonomy, depending on the driver's condition. Stress data of 38 drivers along a road are obtained above and averaged. These averages are used as input to the system via radio frequency identification (RFID) cards. Thus, an autonomous vehicle with GSR sensor-based torque control is designed. The purpose of this transformation in this study is to integrate our work into autonomous vehicles as well as semi-autonomous vehicles. This shows that biosensors can also be used as input for autonomous vehicles. In the takeover request (TOR) study, different TOR times are tested on five different driving cases with 18 participating drivers. Three of these cases are used to detect the TOR time of drivers, while the other two scenarios are used sequentially to increase the effects of other participating vehicles such as a vehicle approaching the intersection and then is stopped on a lane along the test route causing a possible hazard situation. According to the analysis, drivers do not prefer the authority transition that is very close to the critical situation (TOR 6 s). Because in TOR 6 s, due to the approaching the critical situation, the (g) (pulse deviation) and (fxa) (current deviation multiplied by the average of five consecutive acceleration or deceleration values during manual driving) values are higher and a smooth transition does not occur. It has been observed that most of the drivers make the comfortable and smooth authority transition for TOR 4 s and TOR 2 s. The experienced drivers prefer TOR 4 for authority transition. Since the TOR 0 s authority transition is also a sudden transition, the driver is unready and causes a higher (g) and (fxa). In other words, even if the drivers are far from the critical situation, they do not prefer a sudden authority transition. Take-over request time is evaluated for each driver and three driver categories such as experienced, semi-experienced and inexperienced, and validated by a questionnaire. The TOR time is extracted and personalized for each driver, which may improve the current conditional automated driving technologies' penetration and acceptance. As the experience of the driver increases, more stable results are obtained. The TOR time for inexperienced drivers varies for each case. As a result of data analysis, the wearable biosensor GSR sensor data can be used in different human factor technologies to support ADAS. Because, as seen in our study, we detected the stress and status of the driver using the GSR sensor, detected the driver's fault cluster in traffic and trained it with machine learning methods, transformed a semi-autonomous vehicle into a GSR-based torque-controlled vehicle with longitudinal autonomy, and finally, evaluated takeover request performance using the GSR sensor.
Uzun kısa süreli bellek ile altın fiyatı tahmini

(Lisansüstü Eğitim Enstitüsü, 2022-07-06) Birecik, Sina ; Günel Öke, Gülay ; 518181032 ; Mekatronik Mühendisliği

Thanks to its many physical and chemical properties, gold has been a mineral that has attracted the attention of people since ancient times. Although it has various usage areas such as the defense industry, electrical electronics industry, and jewelry, its economic value can be given as the most important feature from the first ages to the present day. With the use of gold in coins by the Lydians, gold became a tool of exchange and investment. It still maintains this feature today. Thanks to these features, it has become a symbol of power in societies. With the Bretton-Woods system, which came into effect towards the end of World War II, the gold price was indexed to the US dollar. With the increasing tension in the world markets, this system collapsed, and gold started to be priced dynamically as of 1971. The gold price in Turkey is calculated using the worldwide accepted gold price in US dollars and the Turkish lira / US dollar parity. Today, investors want to make a profit in the long or short term with minimum risk factors. The most risk-free investment instruments preferred by investors can be given as precious commodities such as gold, foreign currencies, stocks, real estate, cryptocurrencies, bonds, mutual funds, and government bonds. It is preferable to make estimates of the corresponding returns or losses on such investments. For this reason, forecasting is of great importance when investing and the reaction of investment instruments affected by events in the world should be analyzed. Our main purpose in this research is to predict the future behavior of gold based on past data. The recurrent neural network (RNN), which is a type of deep learning, was chosen as the forecasting method. The problem to be studied in this research is considered a regression problem that requires a nonlinear solution to the time series. In this study, feature selection and gold price prediction in multivariate financial time series were examined. There are many dynamic factors for pricing gold used for investment. Since not every factor has the same effect, the most effective factors on gold should be determined. After the data engineering part, the main factors affecting gold price have been identified. Features were determined using the factors examined, and the most important of these features were selected and formed the basis of the study. The datasets were created using various features such as the US dollar index, cryptocurrencies, commodities, stock market indices, volatility index, inflation, and interest rates. Although gold prices in the real-world act according to basic theory and criteria, gold is a commodity type that is affected by many technical and fundamental parameters. Before the forecasting section, feature selection was made using Random Forest Regression and Linear Regression. In this section, it has been determined that the parameters that affect the gold price the most are the USD/JPY parity, 10-year expected inflation (USA), 10-year real interest rate (USA), US gross national product (GDP), and the amount of USD in circulation. No improvement was observed in the forecasting performance criterion even if more variables were added. In the principal component analysis, the most important variables representing the main dataset were determined as oil, US real interest rate, Bitcoin, silver price, LME index, 10-year inflation (USA), TXBM index, USD money supply M1, and volatility index. A basic recurrent artificial neural network (RNN) and long short-term memory deep learning network (LSTM) were used in the forecasting study. The dataset combinations were created by using 5 variant variables on 3 main datasets so that there are 15 combinations in total. These variant variables are economic indicators LMACD and MACD. Test combinations were created using dataset combinations also batch size and window size values determined for RNN and LSTM networks. It has been tried to give an idea about the reaction of the gold price against these inputs. The window size is the hyperparameter that determines how many days the historical data will be retrieved when creating the observation unit. RNN and LSTM hyperparameters were also derived on each dataset combination and forecasting was made. After the training process, the predictive model performances of the applications were calculated. Parameters with appropriate estimation results were determined in the tests performed. While RNN performs at par with LSTM in one main dataset, LSTM has a higher predictive success than RNN in the other two main datasets. In tests where the datasets created by adding the MACD indicator were trained with lower window sizes, these models gave superior results than other combinations. In addition, it was observed that there were deviation errors in the training of the models due to the Covid-19 crisis, which started in March 2020. In the forecasting study on the training set, it was determined that the network could not perform as well as before this date in the part of the training data after the onset of Covid-19. To improve the model, the existing parameters were selected more precisely, and optimization was made. Although it is not possible to use it professionally yet, it has given promising results for the first study. In this context, the aim and scope of the study have been met. It will also be a starting point for future work.
Functional safety mechanism development of creep monitoring in automatic transmission

(Graduate School, 2022-12-28) Ardıç, Burak ; Üstoğlu, İlker ; 518191007 ; Mechatronics Engineering

Functional safety one of the most of important feature of new development lifecycle of the vehicle systems. ISO 26262 is known as "Road vehicles – Functional safety" which is an international standard for functional safety of electrical and/or electronic systems of road vehicles. This definition comes from International Organization for Standardization (ISO) in 2011 and revised in 2018. In today's powertrains, mostly modern automatic transmissions are used for road vehicles. Those transmissions have electronic systems that supports driver activities in better way. I.e., in manual transmission, driver has to control 3 pedal (clutch, acceleration and brake) and gear lever while driving but with help of automatic transmission, driver only controls 2 pedal (acceleration and brake) and usually gear lever always stay in D (Drive) or R (Reverse) based on which direction driver wants to move. In automatic transmissions, clutch pedal is controlled by electronic control units such as transmission control unit. One of the functional safety responsibilities is controlling these electronic control unit activities via different safety mechanism whether they work in proper and safety way. Because in case of wrong detections, wrong calculations in electronic control units or wrong requests of drivers might cause very dangerous severities. In this thesis, it is aimed to develop a functional safety mechanism that monitors the creep/Creep function of the automatic transmission and takes the necessary measures before the accidents caused by this function. Before starting of modelling this safety mechanism monitoring these functions in MATLAB/Simulink, firstly some functional safety concept development has to be done to define procedures. In this study, functional safety development is done based on V-model. Firstly, Item definition is done to define specification of item which is investigated. Since transmission control unit was our main item, all specifications that includes gear ratios, transmission maximum torque, clutch engagement information to transmit torque, communications with other electronic control units and also since transmission control unit is related to vehicle also operational driving and vehicle movement states are given. Then hazard analysis and risk assessment (HARA) is done to define potential hazards and operational situation which can be seen during creep function is investigated and safety goals are determined derived from ASIL. After safety goal determination, functional safety concept that includes safety mechanisms is done by defining functional safety requirements to fulfill safety goals. Before start on development of safety mechanism monitoring, all technical safety requirements are set with hardware and software with including architecture of system. To monitor creep function, in a first-place automatic transmission plant model which includes engine, transmission, vehicle, gear shift mechanism, and CAN/HW state model is implemented in MATLAB/Simulink platform. This plant model also includes a creep function to be monitored. In the plant model development phase, the transmission gear ratio is selected from the item and all other vehicle parameters as engine inertia, and engine and torque converter characteristic values are taken from the vehicle that is thought of as a concept. After functional safety concept development and plant model development, the safety mechanism of creep function monitoring is implemented based on defined safety requirements. The safety mechanism of creep monitoring is responsible for detecting high creep torque errors mainly for driver torque demand, engine torque from plant model, engine speed, and vehicle velocity. During creep, the transmission control unit can request increased engine idle speed/torque if needed or unintentionally close the lockup clutch. Both cases might cause unintended acceleration. The safety mechanism receives the engine torque from the plant model and calculates the consumed by the engine based on engine inertia and engine speed. The safety mechanism of creep monitoring checks the difference between engine torque from the plant model and consumed engine torque. This difference is accepted as creep torque which is the torque transmitted to wheels during creeping. If torque transfer is higher than the defined safety torque threshold for the allowed fault reaction time interval, then safe state which leads to force to bring the vehicle to a standstill via setting gearbox torque to zero is triggered. Therefore, the safety mechanism of the creep function is implemented by considering these conditions. After all these development processes, testing of specific driving test scenarios is simulated to check that if the plant model works as intended then specific functional safety fault injection test cases are simulated to see if the safe state which is defined based on safety goals works as intended to prevent severe accidents.

Gözat

Çıkarma tarihi ile LEE- Mekatronik Mühendisliği Lisansüstü Programı'a göz atma

Sayfa başına sonuç

Sıralama Seçenekleri