LEE- Uçak ve Uzay Mühendisliği-Doktora
Bu koleksiyon için kalıcı URI
1 - 5 / 8
ÖgeNumerical and experimental study of fluid structure interaction in a reciprocating piston compressor(Graduate School, 2022-01-14)Consisting of household refrigerators, cold storages, cold chain logistics, industrial freezers, air conditioners, cryogenics and heat pumps, refrigeration industry are a vital part of many sectors such as food, health care, air conditioning, sports, leisure, production of plastics and chemicals along with electronic data processing centers and scientific research facilities, which can not operate without refrigeration. There are roughly 5 billion in operation refrigeration systems which consumes 20% of the electricity used worldwide, responsible of 7.8% of GHG emission of the world, 500 billion USD cost of annual equipment sale, 15 million of employed people. Around 37% of global warming impact caused by refrigeration is direct emission of fluorinated refrigerants (CFCs, HCFCs and HFCs), 63% is due to indirect emission caused by electricity generation required for refrigeration. Both economic goals of making refrigeration units cheaper, more durable, and environment concerns of making these units more efficient and less hazardous for the world, require meticulous research and study on these refrigeration units. Approximately 40% of refrigeration units consist of domestic refrigeration systems alone where mostly hermetic, reciprocating type compressors are used. Design and improvement of such compressors is a multidisciplinary subject and requires deep understanding of heat and momentum transfer between refrigerant and solid component of compressor which can only be done through scientific investigation, using experimental and numerical techniques. In this thesis study, concerning the advantages of numerical studies, a multi-physics numerical model of flow through the gas line of a household, hermetically sealed, reciprocating piston compressor and the fluid structure interaction around the valve reeds including the contact between deformable parts was developed. Concerning the complexity of the model, the problem divided into several steps and at each step, numerical results are validated with experiments. In the first chapter of this thesis, the motivation behind the thesis study is discussed along with a theoretical background about refrigeration, compressors, fluid-structure interaction and a comprehensive literature survey are summarized to express the position of the thesis study among academic literature and it's novelty. In the second chapter, experimental studies conducted throughout the thesis are presented. Experimental studies divided into two sections. In the first section, the valve reed dynamics are investigated experimentally outside the compressor in multiple test conditions. A test rig is built for this reason, and the displacement of valve reed under constant point load, free oscillation and the impact of valve reed to valve plate from a pre-deformed form are measured, in order to validate the numerical work. In the second section, the compressor specifications such as cooling capacity, compression work, average refrigerant mass flow rate, along with surface temperature and instantaneous pressure variation from several locations inside the compressor are measured inside a calorimeter setup, to provide boundary conditions and validation for numerical analyses. Numerical work of the thesis study is explained in the third chapter. Modelling the whole compressor gas line between compressor inlet and outlet, including the strong coupled interaction between the refrigerant and deformable solid parts such as valve reeds is too complex of an attempt to do in a single step. Therefore, the numerical problem divided into seven smaller numerical problems and investigated consecutively. At each consecutive steps, problems are isolated, identified, solved and results are validated. The similarity of each step to the final model is increased along with it's complexity as a natural consequence at each consecutive steps. The numerical studies also briefly cover the advantages and disadvantages of using an open source or a commercial multi-physics solver, where OpenFOAM and Ansys Workbench software are utilized for this purpose, respectively. After the simplified steps of the numerical model are completed, the whole gas line of a compressor produced by Arçelik is modelled. The numerical results compared against experimentally obtained data and a good agreement is achieved between them. The developed method is further used for parametric investigation on compressor design to show the capabilities and the benefits of the numerical model. Finally, results of whole thesis study, the experience gained throughout the thesis work and the planned future work are discussed in the final chapter.
ÖgeDevelopment of single-frame methods aided kalman-type filtering algorithms for attitude estimation of nano-satellites(Graduate School, 2021-08-20)There is a growing demand for the development of highly accurate attitude estimation algorithms even for small satellite e.g. nanosatellites with attitude sensors that are typically cheap, simple, and light because, in order to control the orientation of a satellite or its instrument, it is important to estimate the attitude accurately. Here, the estimation is especially important in nanosatellites, whose sensors are usually low-cost and have higher noise levels than high-end sensors. The algorithms should also be able to run on systems with very restricted computer power. One of the aims of the thesis is to develop attitude estimation filters that improve the estimation accuracy while not increasing the computational burden too much. For this purpose, Kalman filter extensions are examined for attitude estimation with a 3-axis magnetometer and sun sensor measurements. In the first part of this research, the performance of the developed extensions for the state of art attitude estimation filters is evaluated by taking into consideration both accuracy and computational complexity. Here, single-frame method-aided attitude estimation algorithms are introduced. As the single-frame method, singular value decomposition (SVD) is used that aided extended Kalman filter (EKF) and unscented Kalman filter (UKF) for nanosatellite's attitude estimation. The development of the system model of the filter, and the measurement models of the sun sensors and the magnetometers, which are used to generate vector observations is presented. Vector observations are used in SVD for satellite attitude determination purposes. In the presented method, filtering stage inputs are coming from SVD as the linear measurements of attitude and their error covariance relations. In this step, UD is also introduced for EKF that factorizes the attitude angles error covariance with forming the measurements in order to obtain the appropriate inputs for the filtering stage. The necessity of the sub-step, called UD factorization on the measurement covariance is discussed. The accuracy of the estimation results of the SVD-aided EKF with and without UD factorization is compared for the estimation performance. Then, a case including an eclipse period is considered and possible switching rules are discussed especially for the eclipse period, when the sun sensor measurements are not available. There are also other attitude estimation algorithms that have strengths in coping well with nonlinear problems or working well with heavy-tailed noise. Therefore, different types of filters are also tested to see what kind of filter provides the largest improvements in the estimation accuracy. Kalman-type filter extensions correspond to different ways of approximating the models. In that sense, a filter takes the non-Gaussianity into account and updates the measurement noise covariance whereas another one minimizes the nonlinearity. Various other algorithms can be used for adapting the Kalman filter by scaling or updating the covariance of the filter. The filtering extensions are developed so that each of them is designed to mitigate different types of error sources for the Kalman filter that is used as the baseline. The distribution of the magnetometer noises for a better model is also investigated using sensor flight data. The filters are tested for the measurement noise with the best fitting distribution. The responses of the filters are performed under different operation modes such as nominal mode, recovery from incorrect initial state, short and long-term sensor faults. Another aspect of the thesis is to investigate two major environmental disturbances on the spacecraft close enough to a planet: the external magnetic field and the planet's albedo. As magnetometers and sun sensors are widely used attitude sensors, external magnetic field and albedo models have an important role in the accuracy of the attitude estimation. The magnetometers implemented on a spacecraft measure the internal geomagnetic field sources caused by the planet's dynamo and crust as well as the external sources such as solar wind and interplanetary magnetic field. However, the models that include only the internal field are frequently used, which might remain incapable when geomagnetic activities occur causing an error in the magnetic field model in comparison with the sensor measurements. Here, the external field variations caused by the solar wind, magnetic storms, and magnetospheric substorms are generally treated as bias on the measurements and removed from the measurements by estimating them in the augmented states. The measurement, in this case, diverges from the real case after the elimination. Another approach can be proposed to consider the external field in the model and not treat it as an error source. In this way, the model can represent the magnetic field closer to reality. If a magnetic field model used for the spacecraft attitude control does not consider the external fields, it can misevaluate that there is more noise on the sensor, while the variations are caused by a physical phenomenon (e.g. a magnetospheric substorm event), and not the sensor itself. Different geomagnetic field models are compared to study the errors resulting from the representation of magnetic fields that affect the satellite attitude determination system. For this purpose, we used magnetometer data from low Earth-orbiting spacecraft and the geomagnetic models, IGRF and T89 to study the differences between the magnetic field components, strength, and the angle between the predicted and observed vector magnetic fields. The comparisons are made during geomagnetically active and quiet days to see the effects of the geomagnetic storms and sub-storms on the predicted and observed magnetic fields and angles. The angles, in turn, are used to estimate the spacecraft attitude, and hence, the differences between model and observations as well as between two models become important to determine and reduce the errors associated with the models under different space environment conditions. It is shown that the models differ from the observations even during the geomagnetically quiet times but the associated errors during the geomagnetically active times increase more. It is found that the T89 model gives closer predictions to the observations, especially during active times and the errors are smaller compared to the IGRF model. The magnitude of the error in the angle under both environmental conditions is found to be less than 1 degree. The effects of magnetic disturbances resulting from geospace storms on the satellite attitudes estimated by EKF are also examined. The increasing levels of geomagnetic activity affect geomagnetic field vectors predicted by IGRF and T89 models. Various sensor combinations including magnetometer, gyroscope, and sun sensor are evaluated for magnetically quiet and active times. Errors are calculated for estimated attitude angles and differences are discussed. This portion of the study emphasizes the importance of environmental factors on the satellite attitude determination systems. Since the sun sensors are frequently used in both planet-orbiting satellites and interplanetary spacecraft missions in the solar system, a spacecraft close enough to the sun and a planet is also considered. The spacecraft receives electromagnetic radiation of direct solar flux, reflected radiation namely albedo, and emitted radiation of that planet. The albedo is the fraction of sunlight incident and reflected light from the planet. Spacecraft can be exposed to albedo when it sees the sunlit part of the planet. The albedo values vary depending on the seasonal, geographical, diurnal changes as well as the cloud coverage. The sun sensor not only measures the light from the sun but also the albedo of the planet. So, a planet's albedo interference can cause anomalous sun sensor readings. This can be eliminated by filtering the sun sensors to be insensitive to albedo. However, in most of the nanosatellites, coarse sun sensors are used and they are sensitive to albedo. Besides, some critical components and spacecraft systems e.g. optical sensors, thermal and power subsystems have to take the light reflectance into account. This makes the albedo estimations a significant factor in their analysis as well. Therefore, in this research, the purpose is to estimate the planet's albedo using a simple model with less parameter dependency than any albedo models and to estimate the attitude by comprising the corrected sun sensor measurements. A three-axis attitude estimation scheme is presented using a set of Earth's albedo interfered coarse sun sensors (CSSs), which are inexpensive, small in size, and light in power consumption. For modeling the interference, a two-stage albedo estimation algorithm based on an autoregressive (AR) model is proposed. The algorithm does not require any data such as albedo coefficients, spacecraft position, sky condition, or ground coverage, other than albedo measurements. The results are compared with different albedo models based on the reference conditions. The models are obtained using either a data-driven or estimated approach. The proposed estimated albedo is fed to the CSS measurements for correction. The corrected CSS measurements are processed under various estimation techniques with different sensor configurations. The relative performance of the attitude estimation schemes when using different albedo models is examined. In summary, the effects of two main space environment disturbances on the satellite's attitude estimation are studied with a comprehensive analysis with different types of spacecraft trajectories under various environmental conditions. The performance analyses are expected to be of interest to the aerospace community as they can be reproducible for the applications of spacecraft systems or aerial vehicles.
ÖgeA high-order finite-volume solver for supersonic flows(Lisansüstü Eğitim Enstitüsü, 2022)Nowadays, Computational Fluid Dynamics (CFD) is a powerful tool in engineering used in various industries such as automotive, aerospace and nuclear power. More than ever the growing computational power of modern computer systems allows for realistic modelization of physics. Most of the open-source codes, however, offer a second-order approximation of the physical model in both space and time. The goal of this thesis is to extend this order of approximation to what is defined as high-order discretization in both space and time by developing a two-dimensional finite-volume solver. This is especially challenging when modeling supersonic flows, which shall be addressed in this study. To tackle this task, we employed the numerical methods described in the following. Curvilinear meshes are utilized since an accurate representation of the domain and its boundaries, i.e. the object under investigation, are required. High-order approximation in space is guaranteed by a Central Essentially Non-Oscillatory (CENO) scheme, which combines a piece-wise linear reconstruction and a k-exact reconstruction in region with and without discontinuities, respectively. The usage of multi-step methods such as Runge-Kutta methods allow for a high-order approximation in time. The algorithm to evaluate convective fluxes is based on the family of Advection Upstream Splitting (AUSM) schemes, which use an upwind reconstruction. A central stencil is used to evaluate viscous fluxes instead. When using high-order schemes, discontinuities induce numerical problems, such as oscillations in the solution. To avoid the oscillations, the CENO scheme reverts to a piece-wise linear reconstruction in regions with discontinuities. However, this introduces a loss of accuracy. The CENO algorithm is capable of confining this loss of accuracy to the cells closest to the discontinuity. In order to reduce this accuracy loss Adaptive Mesh Refinement (AMR) is used. This algorithm refines the mesh near the discontinuity, confining the loss of accuracy to a smaller portion of the domain. In this study, a combination of the CENO scheme and the AUSM schemes is used to model several problems in different compressibility regimes, with a focus on supersonic flows. The scope of this thesis is to analyze the capabilities and the limitations of the proposed combination. In comparison to traditional implementations, which can be found in literature, our implementation does not impose a limit on the refinement ratio of neighboring cells while utilizing AMR. Due to the high computational expenses of a high-order scheme in conjunction with AMR, our solver benefits from a shared memory parallelization. Another advantage over traditional implementations is that our solver requires one layer of ghost cells less for the transfer of information between adjacent blocks. The validation of the solver is performed in different steps. We assess the order of accuracy of the CENO scheme by interpolating a smooth function, in this case the spherical cosine function. Then we validate the algorithm to compute the inviscid fluxes by modeling a Sod shock tube. Finally, the Boundary Conditions (BCs) for the inviscid solver and its order of accuracy are validated by modeling a convected vortex in a supersonic uniform flow. The curvilinear mesh is validated by modeling the flow around a NACA0012 airfoil. The computation of the viscous fluxes is validated by modeling a viscous boundary layer developing on a flat plate. The BCs for viscous flows and the curvilinear implementation are validated by modeling the flow around a cylinder and a NACA0012 airfoil. The AUSM schemes are tested for shock robustness by modeling an inviscid hypersonic cylinder at a Mach number of 20 and a viscous hypersonic cylinder at a Mach number of 8.03. Then, we validate our AMR implementation by modeling a two-dimensional Riemann problem. All the validation results agree well with either numerical or experimental results available in literature. The performance of the code, in terms of computational time required by the different orders of approximation and the parallel efficiency, is assessed. For the former a supersonic vortex convection served as an example, while the latter used a two-dimensional Riemann problem. We obtained a linear speed-up until 12 cores. The highest speedup value obtained is 20 with 32 cores. Furthermore, the solver is used to model three different supersonic applications: the interaction between a vortex and a normal shock, the double Mach reflection and the diffraction of a shock on a wedge. The first application resembles a strong interaction between a vortex and a steady shock wave for two different vortex strengths. In both cases our results perfectly match the ones obtained by a Weighted Essentially Non-Oscillatory (WENO) scheme documented in literature. Both schemes are approximating the solution with the same order of accuracy in both, time and space. The second application, the double Mach reflection, is a challenging problem for high-order solvers because the shock and its reflections interact strongly. For this application, all AUSM-schemes under investigation fail to obtain a stable result. The main form of instability encountered is the Carbuncle phenomenon. Our implementation overcomes this problem by combining the AUSM+M scheme with the formulation of the speed of sound of the AUSM+up scheme. This combination is capable of modeling this problem without instabilities. Our results are in agreement with those obtained with a WENO scheme. Both, the reference solutions and our results, use the same order of accuracy in both, time and space. Finally, the third example is the diffraction of a shock past a delta wedge. In this configuration the shock is diffracted and forms three different main structures: two triple points, a vortex at the trailing edge of the wedge and a reflected shock traveling upwards. Our results agree well with both, numerical and experimental results available in literature. Here, a formation of a vortex-let is observed along the vortex slip-line. This vorticity generation under inviscid flow condition is studied and we conclude that the stretching of vorticity due to compressibility is the reason. The same formation is observed when the angle of attack of the wedge is increased in the range of 0-30. In general, the AUSM+up2 scheme performed best in terms of accuracy for all problems tested here. However, for configurations, in which the Carbuncle phenomenon may appear, the combination of the AUSM+M scheme and the computation of the speed of sound formula of the AUSM+up scheme is preferable for stability reasons. During our computations, we observe a small undershooting right behind shocks on curved boundaries. This is imputable to the curvilinear approximation of the boundaries, which is only second-order accurate. Our experience shows that the smoothness indicator formula in its original version, fails to label uniform flow regions as smooth. We solve the issue by introducing a threshold for the numerator of the formula. When the numerator is lower than the threshold, the cell is labeled as smooth. A value higher than 10^-7 for the threshold might force the solver to apply high-order reconstruction across shocks, and therefore will not apply the piece-wise linear reconstruction which prevents oscillations. We observe that the CENO scheme might cause unphysical states in both inviscid and viscous regime. By reconstructing the conservative variables instead of the primitive ones, we are able to prevent unphysical states for inviscid flows. For the viscous flows, temporarily reverting to first-order reconstruction in the cells where the temperature is computed as negative, prevents unphysical states. This technique is solely required during the first iterations of the solver, when the flow is started impulsively. In this study the CENO, the AUSM and the AMR methods are combined and applied successfully to supersonic problems. When modeling supersonic flow with high-order accuracy in space, one should prefer the combination of the AUSM schemes and the CENO scheme. While the CENO scheme is simpler than the WENO scheme used in comparison, we show that it yields results of comparable accuracy. Although it was beyond the scope of this study, the AUSM can be extended to real gas modeling which constitutes another advantage of this approach.
ÖgeDynamic and aeroelastic analysis of advanced aircraft wings carrying external stores(Lisansüstü Eğitim Enstitüsü, 2021)Bu çalışma gelişmiş uçak kanatlarında harici yük ve takip edici kuvvet altında kanadın dinamik ve aeroleastik davranışlarını incelemektedir. Harici yüklerin ağırlığı, pozisyonu, birbirine göre yerleşimi, kompozit katmanların yönelimi ile itki kuvveti etkileri incelenmiş ve hepsinin kanadın doğal frekansı ve kritik çırpınma hızına olan etkileri tespit edilmiştir.
ÖgeA modified anfis system for aerial vehicles control(Lisansüstü Eğitim Enstitüsü, 2022)This thesis presents fuzzy logic systems (FLS) and their control applications in aerial vehicles. In this context, firstly, type-1 fuzzy logic systems and secondly type-2 fuzzy logic systems are examined. Adaptive Neuro-Fuzzy Inference System (ANFIS) training models are examined and new type-1 and type-2 models are developed and tested. The new approaches are used for control problems as quadrotor control. Fuzzy logic system is a humanly structure that does not define any case precisely as 1 or 0. The Fuzzy logic systems define the case with membership functions. In literature, there are very much fuzzy logic applications as data processing, estimation, control, modeling, etc. Different Fuzzy Inference Systems (FIS) are proposed as Sugeno, Mamdani, Tsukamoto, and ¸Sen. The Sugeno and Mamdani FIS are the most widely used fuzzy logic systems. Mamdani antecedent and consequent parameters are composed of membership functions. Because of that, Mamdani FIS needs a defuzzification step to have a crisp output. Sugeno antecedent parameters are membership functions but consequent parameters are linear or constant and so, the Sugeno FIS does not need a defuzzification step. The Sugeno FIS needs less computational load and it is simpler than Mamdani FIS and so, it is more widely used than Mamdani FIS. Training of Mamdani parameters is more complicated and needs more calculation than Sugeno FIS. The Mamdani ANFIS approaches in the literature are examined and a new Mamdani ANFIS model (MANFIS) is proposed. Training performance of the proposed MANFIS model is tested for a nonlinear function and control performance is tested on a DC motor dynamic. Besides, ¸Sen FIS that was used for estimation of sunshine duration in 1998, is examined. This ¸SEN FIS antecedent and consequent parameters are membership functions as Mamdani FIS and needs to defuzzification step. However, because of the structure of the ¸Sen defuzzification structure, the ¸Sen FIS can be calculated with less computational load, and therefore ¸Sen ANFIS training model has been created. These three approaches are trained on a nonlinear function and used for online control. In this study, the neuro-fuzzy controller is used as online controller. Neuro-fuzzy controllers consist of simultaneous operation of two functions named fuzzy logic and ANFIS. The fuzzy logic function is the one that generates the control signal. It generates a control signal according to the controller inputs. The other function is the ANFIS function that trains the parameters of the fuzzy logic function. Neuro-fuzzy controllers are intelligent controllers, independent of the model, and constantly adapting their parameters. For this reason, these controllers' parameters values are constantly changing according to the changes in the system. There are studies on different neuro-fuzzy control systems in the literature. Each approach is tested on a DC motor model that is a single-input and single-output system, and the neuro-fuzzy controllers' advantages and performances are examined. In this way, the approaches in the literature and the approaches added within the scope of the thesis are compared to each other. Selected neuro-fuzzy controllers are used in quadrotor control. Quadrotors have a two-stage controller structure. In the first stage, position control is performed and the position control results are defined as angles. In the second stage, attitude control is performed over the calculated angle values. In this thesis, the neuro-fuzzy controller is shown to work perfectly well in single layer control structures, i.e., there was not any overshooting, and settling time was very short. But it is seen from quadrotor control results that the neuro-fuzzy controller can not give the desired performance in the two-layered control structure. Therefore, the feedback error learning control system, in which the fuzzy controller works together with conventional controllers, is examined. Fundamentally, there is an inverse dynamic model parallel to a classical controller in the feedback error learning structure. The inverse dynamic model aims to increase the performance by influencing the classical controller signal. In the literature, there are a lot of papers about the structure of feedback error learning control and there are different proposed approaches. In the structure used in this work, fuzzy logic parameters are trained using ANFIS with error input.The fuzzy logic control signal is obtained as a result of training. The fuzzy logic control signal is added to the conventional controller signal. This study has been tested on models such as DC motor and quadrotor. It is seen that the feedback error learning control with the ANFIS increases the control performances. Antecedent and consequent parameters of type-1 fuzzy logic systems consist of certain membership functions. A type-2 FLS is proposed to better define the uncertainties, because of that, type-2 fuzzy inference membership functions are proposed to include uncertainties. The type-2 FLS is operationally difficult because of uncertainties. In order to simplify type-2 FLS operations, interval type-2 FLS is proposed as a special case of generalized type-2 FLS in the literature. Interval type-2 membership functions are designed as a two-dimensional projection of general type-2 membership functions and represent the area between two type-1 membership functions. The area between these two type-1 membership functions is called Footprint of Uncertainty (FOU). This uncertainty also occurs in the weight values obtained from the antecedent membership functions. Consequent membership functions are also type-2 and it is not possible to perform the defuzzification step directly because of uncertainty. Therefore, type reduction methods have been developed to reduce the type-2 FLS to the type-1 FLS. Type reduction methods try to find the highest and lowest values of the fuzzy logic model. Therefore, a switch point should be determined between the weights obtained from the antecedent membership functions. Type reduction methods find these switch points by iterations and this process causes too much computation, so many different methods have been proposed to minimize this computational load. In 2018, an iterative-free method called Direct Approach (DA) was proposed. This method performs the type reduction process faster than other iterative methods. In the literature, studies such as neural networks and genetic algorithms on the training for parameters of the type-2 FLS still continue. These studies are also used in the interval type-2 fuzzy logic control systems. There are proposed interval type-2 ANFIS structures in literature, but they are not effective because of uncertainties of interval type-2 membership functions. FLS parameters for ANFIS training should not contain uncertainties. However, the type-2 FLS should inherently contain uncertainty. For this reason, Karnik-Mendel algorithm is modified, which is one of the type-reduction methods, to apply the ANFIS on interval type-2 FLS. The modified Karnik-Mendel algorithm gives the same results as the Karnik-Mendel algorithm. The modified Karnik-Mendel algorithm also gives exact parameter values for use in ANFIS. One can notice that the ANFIS training of the interval type-2 FLS has been developed successfully and has been used for system control.