The reseach on neural networks has been done for six decades. In this period, many neural models and learning rules have been proposed. Futhermore, they were popularly and successfully applied to many applications. They successfully solved many problems that traditional algorithms could not solve efficiently . However, applying multilayer neural networks to applications, users are confronted with the problem of determining the number of hidden layers and the number of hidden neurons in each hidden layer. It is too difficult for users to determine proper neural network architectures. However, it is very significant, because neural network architectures always influence critically their performance. We may solve problems efficiently, only when we has proper neural network architectures. To overcome this difficulty, several approaches have been proposed to generate the architecture of neural networks recently. However, they still have some drawbacks. The goal of our research is to discover better approachs to automatically determine proper neural network architectures. We propose a series of approaches in this thesis. First, we propose an approach based on decision trees. It successfully determines neural network architectures and greatly decreases learning time. However, it can deal only with two-class problems and it generates bigger neural network architectures. Next, we propose an information entropy based approach to overcome the above drawbacks. It can generate easily multi-class neural networks for standard domain problems. Finally, we expand the above method for sequential domain and structured domain problems. Therefore, our approaches can be applied to many applications. Currently, we are trying to work on quantum neural networks. We are also interested in ART neural networks. They are also incremental neural models. We apply them to digital signal processing. We propose a character recognition application, a spoken word recognition application, and an image compression application. All of them have good performances.

Assembly line balancing problem is one of the most studied NP-Hard problems. NP-Hardness leads us to search for a good solution instead of the optimal solution especially for the big-size problems. Meta-heuristic algorithms are the search methods which are developed to find good solutions to the big-size and combinatorial problems. In this study, it is aimed at solving the multi-objective multi-model assembly line balancing problem of a company. A meta-heuristic algorithm is developed to solve the deterministic assembly line balancing problems. The algorithm developed is tested using the test problems in the literature and the the real life problem of the company as well. The results are analyzed and found to be promising and a solution is proposed for the firm.

In this thesis, we study both dynamic and distributed facility layout problems, where the demand for product mix changes over time. We propose a new simulated annealing algorithm, SALAB, for the dynamic facility layout problem. Four variants of SALAB find the best known solution for 20 of the 48 benchmark problems from the literature, improving upon the best known solutions of 18 problems. We modify SALAB to obtain DSALAB, solving the dynamic distributed facility layout problem with the objective of minimizing relocation cost and total (full and empty) travel cost of the material handling system. We simulate DSALAB solutions of randomly generated problems to study the tradeoff between total cost and congestion in the system. Our experimental results indicate that distributing the department duplicates throughout the facility reduces the total cost with diminishing returns and causes increasing congestion. Therefore, distribution beyond a certain level is not justified.

The Multiple Cooling Multi-Objective Simulated Annealing Algorithm is used for the conceptual design optimization of a supersonic Unmanned Combat Aerial Vehicle (UCAV). Single and multiobjective optimization problems are addressed while limiting performance requirements between desired bounds to obtain viable aircraft configurations. A conceptual aircraft design code was prepared for planned but flexible combat missions. The results demonstrate that the optimization technique employed is an effective tool for the conceptual design of aircrafts.

As the complexity of a spatial phenomenon increases, traditional modeling becomes impractical. Alternatively, data-driven modeling, which is based on the analysis of data characterizing the phenomena, can be used. In this thesis, the generation of understandable and reliable spatial models using observational data is addressed. An interpretability oriented data-driven fuzzy modeling approach is proposed. The methodology is based on construction of fuzzy models from data, tuning and fuzzy model simplification. Mamdani type fuzzy models with triangular membership functions are considered. Fuzzy models are constructed using fuzzy clustering algorithms and simulated annealing metaheuristic is adapted for the tuning step. To obtain compact and interpretable fuzzy models a simplification methodology is proposed. Simplification methodology reduced the number of fuzzy sets for each variable and simplified the rule base. Prototype software is developed and mean annual precipitation data of Turkey is examined as case study to assess the results of the approach in terms of both precision and interpretability. In the first step of the approach, in which fuzzy models are constructed from data, &quot / Fuzzy Clustering and Data Analysis Toolbox&quot / , which is developed for use with MATLAB, is used. For the other steps, the optimization of obtained fuzzy models from data using adapted simulated annealing algorithm step and the generation of compact and interpretable fuzzy models by simplification algorithm step, developed prototype software is used. If the accuracy is the primary objective then the proposed approach can produce more accurate solutions for training data than geographically weighted regression method. The minimum training error value produced by the proposed approach is 74.82 mm while the error obtained by geographically weighted regression method is 106.78 mm. The minimum error value on test data is 202.93 mm. An understandable fuzzy model for annual precipitation is generated only with 12 membership functions and 8 fuzzy rules. Furthermore, more interpretable fuzzy models are obtained when Gath-Geva fuzzy clustering algorithms are used during fuzzy model construction.

In this study, the capacitated vehicle routing problem with shift based time constraints is taken into consideration. The study stemmed from an application in a cross dock. The considered cross dock is assumed to feed directly the production lines of its customer. The customer has a just-in-time production system that requires producing only in necessary quantities at the necessary times. This necessitates the arrival of the parts/products collected from different suppliers at the customer at the beginning of each shift of production. The shift times constitute deadlines for the products to be collected from the suppliers and used in each shift. The collection problem then can be seen as the capacitated vehicle routing problem with shift based time constraints. The objective of the collection problem is to minimize the routing costs. For the accomplishment of this objective it is required to decide on products of which shift(s) should be taken from a supplier when a vehicle arrives at that supplier. For the solution of the problem a mathematical model is formulated. Since the dealt problem is NP-Hard, meta-heuristic solution approaches based on variable neighborhood search and simulated annealing are proposed. Computational experimentation is conducted on the test problems which are tailored from the capacitated vehicle routing instances from the literature.

<p>Diffusion of liquids into and through polymers is an important factor that negatively may influence the durability or lifetime of a polymer structure used in industrial applications. In this work two types of polymers, a liquid crystalline polymer (LCP, Vectra A950) and various thermoset resins as used in fibre reinforced plastics (FRP) process equipment were studied with regard to barrier properties, chemical resistance and long-term performance. LCP are known for their outstanding chemical resistance and barrier properties. FRP used in process equipment may be a cost-efficient solution in chemically aggressive environments where standard carbon or stainless steel cannot be used due to its limited corrosion resistance. Transport properties of typical industrial environments were determined for the LCP and the influence of annealing and orientation was investigated to study whether the barrier properties can be improved. The possibility to use LCP as cost-effective lining for FRP was explored. Special focus was put on the diffusion of water and its effect on long-term transport properties and stability of thermoset resins as the performance of FRP is strongly related to the diffusion of water. </p><p>The results showed that Vectra A950 was suitable for organic solvent and non-oxidising acid environments. Its transport properties were gravimetrically determined and found to be 10 to 102 times lower than that of a high barrier fluoropolymer of type FEP. The degree of molecular packing increased with annealing time both below and above the melting point. Below the melting point this was – at least – partly due to crystal formation whereas above the melting point other mechanisms were involved. The effects of annealing and orientation on the transport properties in LCP were, however, very small or not significant and probably significantly longer annealing times are required. LCP has potential to be used as lining material for FRP as the use of an LCP-lining substantially reduced the permeability of and the solute sorption in a bisphenol A epoxy-based vinyl ester resin. The bonding strength was improved significantly by a combined abrasive and oxygen plasma treatment.</p><p>The long-term sorption of water in thermoset resins including bisphenol A epoxy-based vinyl ester, novolac-based vinyl ester, urethane modified vinyl ester and bisphenol A polyester resins was found to increase with exposure time whereas the diffusion coefficient was not significantly affected. It was shown that the presence of water induced relaxation processes that were considered to be the primarily reason for the increase in sorption coefficient in comparison to degradation processes, such as hydrolysis, causing osmotic processes. A general relationship for the estimation of the sorption coefficient at 80ºC in dependence of the water activity and the sorption coefficient at unit activity independent of the resin type was established.</p>

Thesis (Ph.D.(Computer Science))--University of Pretoria, 2009. / Abstract in English. Includes bibliographical references.

Thin films of GeTe₂ were deposited on glass substrates using RF magnetron sputtering with various hydrogen flow rates in the growth chamber. Transmission data of deposited films were taken and used to determine optical constants: refractive index (n), extinction coefficient (κ), and absorption coefficient (α)) and the energies: E₀₄, E₀₃, Tauc band gap E[subscript]Tauc and Urbach energy E[subscript]U. An increase of these energies was observed with increasing hydrogen flow rate. This increase is interpreted on the basis of the density of state model proposed by Mott and Davis. An increase of network disorder due to the inclusion of hydrogen into the GeTe₂ thin films was determined from the B[superscript]1/2 parameter, Urbach energy and full width at half maximum of Raman vibrational modes. The crystallization process induced by thermal annealing on GeTe₂ was studied. X-ray diffraction measurements were performed and the results suggest that crystallization of GeTe₂ occurs via a phase separation into Te and GeTe crystalline phases. This observation is in agreement with a previous report. The crystallization temperature increases with the addition of hydrogen. This increase is explained in terms of dangling bonds. A large change (approximately 60 percent decrease) of the optical transmission occurs after the phase change from amorphous to crystalline. This decrease is interpreted as a result of the observed phase separation.

Šiame darbe nagrinėjama negiljotininio, dvimačio, stačiakampių pjaustymo uždavinio atliekų minimizavimo problema ir jos sprendimo metodai. Dėl uždavinio kombinatorinio sudėtingumo neįmanoma tiksliai ir visais atvejais pateikti optimalų jo sprendinį, todėl pasirinkti apytiksliai sprendimo metodai. Uždavinys sprendžiamas metaeuristiniais hibridiniais genetiniu ir modeliuojamo atkaitinimo algoritmais apjungtais su euristiniais „Žemiausio kairėn užpildymo“ ir „Žemiausio tarpo“, kuris yra originali „Geriausiai tinkamo“ metodo modifikacija. Taip pat realizuojami minėti euristiniai algoritmai atskirai nuo hibridinių. Atliekama šių metodų lyginamoji analizė bei jų parametrų ir pradinių sąlygų parinkimo įtakos tyrimas sprendinio kokybei. Suformuojama ir pateikiama metodika pjaustymo uždavinių sprendimui. / A non – guillotinable, two – dimensional, rectangular cutting stock problem is being introduced in this paper and its solving methods either. Due to the combinatorial complexity of a problem, it is impossible to solve it optimally for every instance. Consequently an aproximate methods have been chosen. The problem is solved by metaheuristic genetic and simulated annealing methods hybridised with heuristic „Bottom Left Fill“ and „Lowest Gap“, which is an originally modified version of „Best Fit“ algorithm. The same heuristic algorithms are implemented separately from hybridised ones. A comparation analysis of these methods is done and the influence on solution quality depending on the selection of algorithms parameters and its initial conditions is considered. The methodology of solving cutting stock problems is being formulated and presented.