975
Archived Theses
0
DOIs Assigned
0%
DOI Rate
Archived Theses
Proteinlerin ikincil yapılarının tahmini amaçlı iki aşamalı matematiksel programlama algoritması
ABSTRACT Recognition of function of protein is one of the major steps in drug discovery process and it is known that functions of polypeptides are dictated by their structures. This study aims to predict secondary structures of proteins utilizing a two-stage probabilistic algorithmic approach. First stage is determination of structural class (all-a, all-P, a/p or a+P) of unknown protein by a Mixed Integer Linear Programming formulation. Next phase is searching 3 to 7-residues-long segments of unknown protein's residue sequence in database of structures of experimentally identified proteins, belonging to the same structural class determined for the unknown one. Source for data on structures of known proteins is Protein Data Bank; http://www.pdb.org. Three states (a-helix, P-sheet or loop) per residue predictions are obtained through a probabilistic approach utilizing outcomes of database search process. We achieved 100% accuracy in folding type determination phase. Weights to put on probabilities obtained from 3 to 7-residues-long segments for each structural class are determined optimally for each structural class via Non-Linear Programming Formulations again utilizing structural information on experimentally identified proteins. The prediction method is tested on 419, 579, 707 and 601 known proteins from all-a, all-P, a/p and a+p classes, respectively, as if they were not known. 3-states-per-residue accuracy levels obtained for all-a, all-p, a/p and a+p classes are 80.5%, 72.4%, 71.9% and 75.5%, respectively. m
Merkezi ve gayri merkezi tedarik zinciri sistemlerinin en iyilemesinde model öngömeli kontrol metodunun uygulanması
ABSTRACT This thesis studies Model Predictive Control (MPC) as an optimization tool for supply chain systems and compares its results on multi-product centralized and decentralized systems. The main challenge in supply chain optimization studies is to reduce the negative effects of rapid changes in demand. Previous work on application of MPC techniques onto supply chain systems aimed to keep inventory or customer satisfaction levels at a certain value, rather than optimizing the supply chain as a whole. In this study overall profit of the supply chain systems are optimized through MPC and compared with open-loop optimization results. Different MPC optimization horizon lengths are studied and MPC performance on centralized supply chains with different production size, initial inventory values and unpredictable demand change are observed. Semi- decentralized and decentralized supply chain systems are optimized with different MPC horizon lengths and their performance are studied. Performances of the decentralized configurations are compared with that of the centralized configuration while each configuration observes the same demand pattern and implements MPC with the same horizon length. It is observed that for both centralized and decentralized systems as the MPC horizon length increases, higher profits are obtained, yet a trade off must be made between high profits and low solution times. In the decentralized cases, the lack of information causes high bullwhip effect in the supply chains, increasing the costs and decreasing the profits. IV
Çoklu dikey iletkenli düzlemsel geometriler için verimli elektromagnetik simülasyon algoritmasının geliştirilmesi
ABSTRACT An efficient electromagnetic simulation algorithm, which is based on the solution of mixed-potential integral equation (MPIE) via the method of moments (MoM), is presented for the analysis of general planar geometries with multiple vertical metallizations. Although the algorithm itself is the main contribution of this thesis, there have been two additional contributions in the steps towards this goal. These contributions are in the form of thor ough investigation of the two subtopics leading to the main algorithm, namely, closed-form Green's functions and MoM matrix entries. Since the closed-form Green's functions are employed in and the vital part of the al gorithm presented in this thesis, the steps of getting these Green's functions are critically reviewed, not only for the sake of completeness, but also for clarifying some misunderstand ings in the derivation, and the limitations of the method. As it is well-known, the derivation of the closed-form Green's functions begins with the approximation of the spectral-domain Green's functions by complex exponentials, and continues with applying the Sommerfeld identity to cast these approximated spectral-domain Green's functions into the space do main in closed forms. During the last decade, several questions have been raised on the mathematical validity of some steps in the derivation, and remedies have been proposed. These questions can be categorized under the topics of branch point contributions, the ef fect of quasi-static terms, and the contributions of the surface wave poles to the closed-form Green's functions. The potential problems of the derivation of the closed-form Green's func tions, with a priority given to the topics addressed in the literature, have been discussed in detail, and demonstrated that some are due to misunderstandings, some due to numerical errors interpreted as physical errors, and the others are the mere limitations of the method. The numerical technique employed in this work for the analysis of planar geometries is based on the solution of MPIE using the spatial-domain MoM in conjunction with the closed-form Green's functions. The introduction of closed-form Green's functions into the MoM formulation has already resulted in a significant computational improvement for the ivanalysis of planar geometries. However, in cases of vertical metallization, the computational efficiency of the method cannot be improved by simply casting the Green's functions in closed forms, which is due to some difficulties encountered in incorporating the closed-form Green's functions into the MoM formulation for such geometries. In this work, a solution to circumvent these difficulties is proposed, and it is demonstrated that multiple vertical conductors can be efficiently modelled and incorporated into the simulation algorithm, pro vided all vertical metallization have the same dependence in the vertical direction. The results obtained from the algorithm proposed in this thesis are compared to those obtained from commercially available software and experiments. Keywords: Green's functions, DCIM, Pull- wave EM analysis, Method of Moments, 3-D Planarly Layered Media, Vertical Conductors
Serbest form yüzeylerin işlenmesi
ABSTRACT The ball end milling process is used extensively in machining of sculpture surfaces in automobile, die/mold and aerospace industries. In planning machine operations, the process planner has to be conservative in selecting machining conditions with respect to metal removal rate in order to avoid unwanted results such as chipping, cutter breakage or overcut due to excessive cutter deflection. These problems are particularly important for machining of sculptured surfaces where axial and radial depths of cut are abruptly changing. For this reason it will be very critical to provide the process planner with guidance in selecting machining conditions. One solution to this problem is to consider the cutting force in the stage of selecting machining conditions. If the process planner can predict the cutting force for a certain machining condition, it would be very helpful in judging whether the selected condition is appropriate. As a result, a mathematical model is developed in this work for the prediction of cutting force system in ball end milling of sculpture surfaces. The model has the ability to calculate the workpiece/cutter intersection domain automatically for a given cutter path, cutter and workpiece geometries. In addition to predicting the cutting forces, the model also determines the surface topography and scallop height variations along the workpiece surface which can be visualized in solid form. Extensive experiments are performed to validate the theoretical model with measured forces and surface topography outputs. Even for complex part geometries, the mathematical model predictions agree well with experimental measurements. m
Entegre araç rotalama ve depo yerleşimi problemi
ABSTRACT This thesis proposes an exact algorithm for the integrated vehicle routing and warehouse location problem, known as WLRP. WLRP helps the supply chain managers to make strategic level decisions, and may be used in a supply chain environment at different industrial sectors. Previous work on WLRP emphasized heuristic and metaheuristic solution approaches that do not guarantee optimality. In search of optimality, other related and well-studied problems are investigated and it is concluded that there is a close relationship between the vehicle routing problem with time windows (VRPTW) and WLRP. Therefore, first the VRPTW is studied extensively as a foundation to WLRP. Recent studies on VRPTW call attention to column generation methods together with the branch and bound algorithm. The column generation algorithm with dynamic programming is used and in this thesis a new dynamic programming formulation is presented. Furthermore, for heterogeneous fleet problems, a new variant of vehicle routing problem (VRP) is formulated which is named as the vehicle routing problem with time windows and discrete capacities (VRPTWDC); and the dynamic program is modified for this problem. The WLRP is formulated as a set partitioning problem and column generation technique is applied. The dynamic program is extensively modified to handle the problem more effectively, the bounds are tightened using 2-path cuts and the subtours are eliminated using a separation algorithm. Finally, branch and bound method is applied to solve WLRP optimally. All benchmark problems in WLRP literature are solved and it is illustrated that the proposed algorithm yields better solutions compared to algorithms reported in the literature. IV
Çok şirketli işbirlikçi tedarik zinciri yönetiminin karma tamsayılı doğrusal programlama ile analizi
ABSTRACT Production systems must interact with other systems for a better performance. The interaction among production systems is usually established when these systems exchange materials. Integrated analysis of different production systems can provide valuable insight and identify improvements in the financial and environmental performance of industrial supply chain systems. The optimal integration of production systems at an industrial zone for multi-period operation has been addressed in this thesis. A systematic approach to identify the synergy among different companies at the strategic level has been developed. The proposed approach uses three steps; the generation of standardized models for process equipment in the production systems, integration of these models that make up the supply chain system, and solution of the model and analysis of the results. The approach includes detailed MILP models for each production system to realize collaboration in a multi-period setting. The approach is illustrated with an example that is a simplified version of a real problem and tested on an industrial problem. It is shown that important improvements in the cost and release of environmentally harmful emissions can be accomplished by collaboration in supply chain systems. T.C. YOKSEK^RCTiM KURULU iv B9KUMANTm*yQN MERKEZİ
Çoklu-ortam ses-görüntü işleme ile biometrik konuşmacı tanıma
ABSTRACT In this thesis we present a multimodal text-dependent speaker identification system. The objective is to improve the recognition performance over conventional unimodal or bimodal schemes. The proposed system decomposes the information existing in a video stream into three modalities: voice, face texture and lip motion. Lip motion between successive frames is first computed in terms of eigenlip coefficients and then encoded as a feature vector. The feature vectors obtained along the whole stream are linearly interpolated to match the rate of the speech signal and then fused with mel frequency cepstral coefficients (MFCC) of the corresponding speech signal. The resulting joint feature vectors are used to train and test a Hidden Markov Model (HMM) based identification system. Face texture images are treated separately in eigenface domain and integrated to the system through decision-fusion. Experimental results are also included for demonstration of the system performance. IV
Çoklu ortam servislerini sağlamak için spektrumu verimli telsiz paket veri sistem tasarımı
ABSTRACT Efficient wireless packet data access is possible by exploiting the multi-user diversity using an opportunistic multiple access system that allocates system resources to one user at a time while employing adaptive coding and modulation. Based on this outcome, recently the IS-856 system was developed. A scheduling algorithm provides resource allocation in the IS-856 system, and its proper design is perhaps one of the most crucial aspects for ensuring good system performance. The thesis presents a number of such algorithms, five of which are novel. The proposed algorithms appear to have the best overall performance of achieving high system throughput without diverging much from the optimal latency performance. The thesis shows that the IS-856 system can easily be adjusted to provide a multitude of services, each with different QoS requirements. Extensive performance evaluations show that good system performance can be maintained in the multi-service scenario. The thesis also presents the means of providing multicast service provisioning in the IS-856 system. The selection of the scheduling algorithm influences not only the system throughput but also the average delay exposed by users in between successive accesses to the system. The thesis proposes new access method, Two Users At A Time System, and its scheduling algorithms which decrease latency figures approximately 50%. Present 3G carrier technologies cannot separately cover all the demands of the end- user in terms of coverage, bandwidth, quality of service (QoS) and cost. The 4G networks are heterogeneous networks that eliminate previous technologies drawbacks and contain large number of different access methods. Therefore, the thesis demonstrates new access methods are needed to combine some of the existing network topologies. These access methods provide wireless data connectivity for nomadic users when away from their offices and homes. The envisioned corporation increases the IS-856 network throughput more than 50%. IV
Çok tedarikli ve rassal getirisi olan tek dönemlik bir envanter modeli
ABSTRACT Inventory management requires handling uncertainties arising at procurement and pro duction stages, transportation channels and demand attributes effectively and efficiently in order to have a robust operational control. In this thesis, a single period inventory model with unreliable suppliers and uncertain demand is analyzed. Supplier unreliability is mod eled by stochastic proportions and/or random supplier capacity in related literature. We combine these two approaches where a stochastic proportion is applied to the minimum of the order amount and random supplier capacity. There are studies in the literature on ran dom yield that primarily consider models with only a single vendor. The main contribution of this work is the extension to the case where there are multiple vendors. We consider cases involving distinct and identical servers to discuss a number of issues including order diver sification. We show that the optimal ordering policy does not have a simple order-up-to or base stock structure. A number of numerical illustrations are given to discuss the structure of optimal policies. IV
Altbayır tabanlı algoritmalarla hızlı kör eşitleme
ABSTRACT In this thesis, several novel blind equalization methods based on subgradient search over a convex cost surface are presented. These subgradient-based methods are alternatives to the existing iterative blind equalization approaches (such as the Constant Modulus Algorithm (CMA)) which mostly suffer from the convergence problems caused by their nonconvex cost functions. The proposed methods are variations of on an iterative algorithm (called SubGradient based Blind Algorithm (SGBA) ) for both real and complex constellations. SGBA is based on the minimization of the loo norm of the equalizer output under a linear constraint on the equalizer coefficients using subgradient iterations. The algorithm has a nice convergence behavior attributed to the convex loo cost surface as well as the step size selection rules connected with the subgradient search. We study four different variations of the SGBA algorithm: Fixed Window SGBA, Moving Window SGBA, Weighted SGBA and the Fractionally-Spaced SGBA. The performances of these algorithms are illustrated using examples in both complex and real constellations, where it is shown that the convergence behaviors of the proposed algorithms are in general less sensitive to initial point selection and fast convergence speeds can be achieved with a wise selection of step sizes. Furthermore, the amount of data required for the training of these equalization algorithms and their complexities are significantly low. IV