Submitted by Renata Lopes (renatasil82@gmail.com) on 2017-02-24T14:20:45Z No. of bitstreams: 1 rafaeldepaulagarcia.pdf: 1337243 bytes, checksum: b838edc08b3d115cfea5624cd3881538 (MD5) / Approved for entry into archive by Adriana Oliveira (adriana.oliveira@ufjf.edu.br) on 2017-02-24T15:39:22Z (GMT) No. of bitstreams: 1 rafaeldepaulagarcia.pdf: 1337243 bytes, checksum: b838edc08b3d115cfea5624cd3881538 (MD5) / Made available in DSpace on 2017-02-24T15:39:22Z (GMT). No. of bitstreams: 1 rafaeldepaulagarcia.pdf: 1337243 bytes, checksum: b838edc08b3d115cfea5624cd3881538 (MD5) Previous issue date: 2014-02-14 / CAPES - Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / A aplicação de metaheurísticas em problemas reais com restrições não é possível sem ajustes. Esta impossibilidade é devida ao fato delas serem desenvolvidas, em sua essência, para resolver problemas de otimização irrestritos. Esses ajustes são feitos por meio de técnicas que abordam as restrições apresentadas no problema. Técnicas de Penalização são comuns, transformando um problema de otimização restrito em um problema de otimização irrestrito, adicionando uma penalidade para a função aptidão das soluções infactíveis. Esta dissertação considera uma técnica que adapta o valor do coeficiente de penalização de cada restrição usando informações da população, tais como a média da função de objetivo e o nível de violação em cada restrição. Esta técnica é conhecida como Método de Penalizaçao Adaptativa (ou simplesmente APM). Existem na literatura várias variantes para o APM que podem ser sintetizadas como: APM Esporádico que mantém os coeficientes de penalização fixados em um número fixo de gerações, uma segunda abordagem semelhante à primeira, mas que acumula valores das violações; variante chamada APM Monotônico, que é semelhante ao APM original, mas que não permite que os coeficientes de penalização sejam reduzidos ao longo do processo evolutivo e variante APM Amortecida, que usa uma média ponderada dos valores atuais e anteriores dos coeficientes de penalização. Novas variantes para o APM são propostas nesta dissertação com a finalidade de buscar melhorias para o APM original. O desempenho destas novas variantes é examinado usando funções teste e problemas de engenharia mecânica e estrutural. Comparações são realizadas utilizando perfis de desempenho, que permitem identificar mais claramente a robustez dessas variantes apontando as melhores. / The application of metaheuristics on real problems with constraints is not possible without adjustments. This impossibility is due to the fact that they are developed, in their essence, to solve unconstrained optimization problems. These adjustments are made by techniques that address the constraints present in the problem. Penalty Techniques are common, transforming a constrained optimization problem into an unconstrained optimization problem, adding a penalty to the fitness function of infeasible solutions. This thesis considers a technique that adapts the value of the penalty coefficient of each constraint using the information of the population, such as the average of the objective function and the level of violation of each constraint. This technique is known as Adaptive Penalty Method (or simply APM). There are in the literature, several variants for the APM and they can be synthesized as: Sporadic APM which holds the fixed penalty coefficients for a fixed number of generations, a second approach similar to the first, but accumulating values of the violations; the variant entitled Monotonic APM, which is similar to the original APM but not allowing the penalty coefficients be reduced along the evolutionary process and the variant damped APM, which uses a weighted average of the current and previous values of the penalty coefficients. New variants for the APM are proposed in this thesis in order reach improvements in the original APM. The performance of these new variants is examined using test-functions and problems of mechanical and structural engineering. Comparisons are conducted using performance profiles, which allow to identify more clearly the robustness of these variants pointing out the best ones.

Diversos setores da economia tem alto grau de dependência de sistemas computacionais: telecomunicação, financeiro, infraestrutura, industrial dentre outros. Desta forma, a qualidade do software contido nestes sistemas é um ítem importante para o bom desempenho destes setores. A arquitetura de software é considerada fator determinante para a qualidade de software. Este trabalho estuda a maneira pela qual a arquitetura de software determina a qualidade do software produzido e as possibilidades de se obter os atributos de qualidade desejados através da especificação de uma arquitetura de software apropriada. O método de pesquisa se fundamenta na revisão da literatura e quatro abordagens para a especificação da arquitetura de software são consideradas: clássica, orientada a objetos, orientada a atributos e orientada a busca. A abordagem orientada a busca é um campo de estudo relativamente recente e os avanços realizados são reportados dentro da área de conhecimento denominada de Search Based Software Engineering. Esta área de conhecimento utiliza técnicas metaheurísticas para achar boas soluções para os problemas encontrados na Engenharia de Software. Uma das técnicas meta-heurísticas mais utilizadas, o algorítmo genético, é usada em uma aplicação cujo processo de design segue a abordagem orientada a busca. / Many sectors of economy depend highly on computing systems: telecommunication, finance, infrastructure, industrial, and others. Thus, the quality of software in these systems is an important item to achieve good performance in these sectors. The software architecture is considered one of the main factors that shape the software quality. This work studies the way software architecture determines the software quality and the possibilities to obtain the desired software quality attributes through specifying appropriate software architecture. The research method is based upon literature review and four approaches to software architecture design process are considered: classic, object oriented, attribute oriented and search oriented. The search oriented approach to software architecture design process is a relatively new field of study and advances are reported in a knowledge area called Search Based Software Engineering. This knowledge area uses metaheuristics techniques to find good solutions to problems found in software engineering. One of the metaheuristic technique most frequently used, the genetic algorithm, is used in an application that follows the search based approach.

Orientador: Auteliano Antunes dos Santos Júnior / Dissertação (mestrado) - Universidade Estadual de Campinas, Faculdade de Engenharia Mecânica / Made available in DSpace on 2018-08-24T20:50:58Z (GMT). No. of bitstreams: 1 Almeida_AndreBatistade_M.pdf: 6664818 bytes, checksum: 4641147a0beb2b3a4e2f768dda791fae (MD5) Previous issue date: 2014 / Resumo: Estudou-se, neste trabalho, o emprego de uma ferramenta de otimização metaheurística, o método dos Algoritmos Genéticos, para a otimização de forma de componentes mecânicos. O trabalho envolveu o estudo das ferramentas usuais e de diversas novas ferramentas para otimização, com foco em AG. Para tanto, foram desenvolvidas rotinas computacionais específicas. Os resultados obtidos com este método foram comparados com os obtidos com o emprego de um programa comercial (Ansys®), que utiliza ferramentas tradicionais de otimização (método de aproximação por subproblema). O trabalho buscou ainda comparar os seus resultados com os obtidos em literatura. Os resultados mostraram que é possível utilizar AG para otimização de forma, com desempenho adequado em relação ao método de aproximação por subproblema e aos resultados de literatura. Mostraram ainda que, utilizando parâmetros otimizados, é possível obter resultados mais adequados com o programa desenvolvido nesta dissertação / Abstract: This study consisted of using a metaheuristic optimization tool, Genetic Algorithm, for the mechanical components shape optmization. It involved the study of usual and several new tools for optimization, focusing GA. For this, speci?c computing routines were developed. The obtained results were compared to the ones generated by using a commercial program (Ansys), that uses traditional optimization tools (subproblem approximation method). The study also compared its results with the ones obtained in the literature. The results showed that it is possible to use AG for shape optimization with adequate performance related to the subproblem approximated method and the literature results. It was also showed that it is possible to obtain more adequate results, using optimized parameters, with the program developed in this study / Mestrado / Mecanica dos Sólidos e Projeto Mecanico / Mestre em Engenharia Mecânica

Orientador: Lubienska Cristina Lucas Jaquiê Ribeiro / Dissertação (mestrado) - Universidade Estadual de Campinas, Faculdade de Tecnologia / Made available in DSpace on 2018-08-20T06:43:15Z (GMT). No. of bitstreams: 1 Aleixo_AlineSerpeloni_M.pdf: 2178208 bytes, checksum: aaef09b93f33ada98dd56e8ff5a84ba0 (MD5) Previous issue date: 2012 / Resumo: Uma das principais tarefas de controle numa Usina Hidrelétrica é a regulação das máquinas hidráulicas. Seu monitoramento faz-se necessário para garantir e estabelecer regras operacionais seguras para a instalação. A turbina é um dos elementos básicos de uma Usina Hidrelétrica e o regulador tipo PID é um dos tipos usados para sua regulação. Os parâmetros do regulador da turbina hidráulica são muito estudados, inclusive verificando seu comportamento através da simulação computacional. As simulações computacionais são ferramentas de análise muito úteis e que permitem encontrar a melhor solução que proporcione o melhor desempenho do sistema estudado. Este trabalho apresenta uma metodologia para otimização das constantes do regulador tipo PID de turbina hidráulica utilizando um modelo de simulação aliado a técnicas atuais de otimização evolutiva baseada nos Algoritmos Genéticos. Os exemplos avaliados mostraram significativa melhora nos valores dessas constantes, mostrando a eficácia do uso dessa ferramenta na área da Hidráulica e possibilitando estudos futuros / Abstract: One of the main tasks of control in a Hydroelectric Power Plant is the hydraulic machines regulation. Its monitoring becomes necessary to guarantee and to establish safe operational rules for the installation. The turbine is one of the basic elements of the Hydroelectric Power Plant and the PID governor is one of the types used to its regulation. The governor parameters of the hydaulic turbine are studied, also verifying its behavior through the computational simulation. The computational simulations are very useful analysis tools to find the best solution to provide the best performance of the studied system. This work presents a methodology to optimize the PID governor parameters of the hydraulic turbine using a simulation model ally the current techniques of evolutionary optimization based in the Genetic Algorithms. The evaluated examples had shown significant improvement in the constants values, showing the effectiveness use of this tool in the hydraulic area and making possible future studies / Mestrado / Tecnologia e Inovação / Mestre em Tecnologia

This thesis provides an analysis of cost of ownership issues and techniques, and provides the supporting data to enable future system designers to make rational decisions on design options. It represents the experience gained whilst collecting cost and cost relationship data in the Rolls-Royce group over a period or more than four years. This, in a time of continuous change, in both the company and the wider IT industry. The thesis is arranged in chapters, each representing a milestone conference or journal paper. The exception to this is chapter Il- the conclusion and summary of the work in the thesis. The Chapter topics cover firstly the background of whole life cost and the aims and objectives of the research. A relationship between whole life cost and quality is considered and why whole life cost is a useful measure of quality. This is examined in practical terms of tools and methods. Case studies are used to illustrate the measurement and use of whole life cost. The impact of obsolescence risk is next considered, identifying the causes and implications of obsolescence. Case studies are used to show how the IT help desk can be used to identify and reduce whole life costs both in a deterministic and a probabilistic approach. This is followed by an examination of the costs of database systems at Rolls-Royce and Associates. Case studies of database systems are also used to show the need to collect in service data, and genetic algorithms are shown to be a useful tool for analysing the data. Whole life costing techniques applied to engineering systems at Rolls-Royce is examined. It is shown that a reliability centred maintenance database is a cost effective tool in collecting data. Network monitoring software is shown to be an effective tool for reducing the cost of ownership of IT systems. The overall conclusion is that whole life cost techniques have been shown to work for computer based systems, further work in this area is still needed to enable costs to be fully understood and optimised.

Due to the inherent problems with their structural elucidation in the laboratory, the computational prediction of membrane protein structure is an essential step toward understanding the function of these leading targets for drug discovery. In this work, the development of a genetic algorithm technique is described that is able to generate predictive 3D structures of membrane proteins in an ab initio fashion that possess high stability and similarity to the native structure. This is accomplished through optimisation of the distances between TM regions and the end-on rotation of each TM helix. The starting point for the genetic algorithm is from the model of general TM region arrangement predicted using the TMRelate program. From these approximate starting coordinates, the TMBuilder program is used to generate the helical backbone 3D coordinates. The amino acid side chains are constructed using the MaxSprout algorithm. The genetic algorithm is designed to represent a TM protein structure by encoding each alpha carbon atom starting position, the starting atom of the initial residue of each helix, and operates by manipulating these starting positions. To evaluate each predicted structure, the SwissPDBViewer software (incorporating the GROMOS force field software) is employed to calculate the free potential energy. For the first time, a GA has been successfully applied to the problem of predicting membrane protein structure. Comparison between newly predicted structures (tests) and the native structure (control) indicate that the developed GA approach represents an efficient and fast method for refinement of predicted TM protein structures. Further enhancement of the performance of the GA allows the TMGA system to generate predictive structures with comparable energetic stability and reasonable structural similarity to the native structure.

Context: The Vehicle Routing Problem is an NP-hard problem with a combination of varieties oftopics like logistics, optimization research and data mining. There is a vast need of vehicle routingsolutions in day to day like with different constraints. According to the requirements, this problem hasbeen a field of interest to a lot of researchers who incorporate scientific methods to combine andinnovate new solutions to optimize the routing. Being an np-hard problem, it is almost impossible tocompute the solutions to optimality but years of research on this area has paid off quite significantlyand the solutions are optimized little by little and better than before. Some applications may or maynot find slight difference in the performance as a considerable affect but some applications orscenarios heavily depend on the performance of the solution where it is very vital that the solution isoptimized to the fullest. As a data mining technique clustering has been used very prominently in caseof portioning scenarios and similarly it has also began to surface in implementing VRP solutions.Although it has recently emerged into the Vehicle Routing era and shown some significant results, ithas not yet come into an open state or awareness. The awareness regarding clustering matters in ahuge extent to be considered by most of the recent researchers who formulate new algorithms to solveVRP and help them further optimize their solution. Objectives: In this study the significance of clustering has been considered to find out how the usageof clustering techniques can alter the performance of VRP based solutions favorably. Then to test theresults of two recently proposed cluster based algorithms, a comparison has been made to other typesof algorithms which prove how the algorithms stand with various methods. Methods: A literature review is performed using various articles that have been gathered from GoogleScholar and then an empirical experiment was conducted on the results available in the papers. Thisexperiment was done by performing a comparative analysis. Results: For the literature review the results were gathered from all the articles based on theirresearch, experience, use of clustering and how their result was improved by using clustering methodsin their formulations. Considering the experiment, the results of both the algorithm were comparedwith the results of five other papers who aim to solve the VRP using exactly the same instances thatwere used in the two algorithms in order to compare valid results on the same variables. Then theresults were analyzed for the purpose of comparison and conclusions were drawn accordingly. Conclusions: From the research performed in this paper we can conclude the vast significance ofclustering techniques that were drawn based on practical test results of various authors. From theexperiment performed it is clear that the Hybrid algorithm has a much higher performance than anyother algorithm it has been compared to. This algorithm has also been proven to enhance itsperformance due to the implementation of clustering techniques in their formulation. Since the resultswere only based on performance that is, in this case the total distance of the final route, future studyindicates the implementation of algorithms to compare them on basis of time complexity and spacecomplexity as well.

Context: The workload at seaport container terminal is increasing gradually. We need to improve the performance of terminal to fulfill the demand. The key section of the container terminal is container stacking yard which is an integral part of the seaside and the landside. So its performance has the effects on both sides. The main problem in this area is unproductive moves of containers. However, we need a well-planned stacking area in order to increase the performance of terminal and maximum utilization of existing resources. Objectives: In this work, we have analyzed the existing container stacking system at Helsingborg seaport container terminal, Sweden, investigated the already provided solutions of the problem and find the best optimization technique to get the best possible solution. After this, suggest the solution, test the proposed solution and analyzed the simulation based results with respect to the desired solution. Methods: To identify the problem, methods and proposed solutions of the given problem in the domain of container stacking yard management, a literature review has been conducted by using some e-resources/databases. A GA with best parametric values is used to get the best optimize solution. A discrete event simulation model for container stacking in the yard has been build and integrated with genetic algorithm. A proposed mathematical model to show the dependency of cost minimization on the number of containers’ moves. Results: The GA has been achieved the high fitness value versus generations for 150 containers to storage at best location in a block with 3 tier levels and to minimize the unproductive moves in the yard. A comparison between Genetic Algorithm and Tabu Search has been made to verify that the GA has performed better than other algorithm or not. A simulation model with GA has been used to get the simulation based results and to show the container handling by using resources like AGVs, yard crane and delivery trucks and container stacking and retrieval system in the yard. The container stacking cost is directly proportional to the number of moves has been shown by the mathematical model. Conclusions: We have identified the key factor (unproductive moves) that is the base of other key factors (time & cost) and has an effect on the performance of the stacking yard and overall the whole seaport terminal. We have focused on this drawback of stacking system and proposed a solution that makes this system more efficient. Through this, we can save time and cost both. A Genetic Algorithm is a best approach to solve the unproductive moves problem in container stacking system.

In recent years, the field of evolutionary dynamic optimisation has seen significant increase in scientific developments and contributions. This is as a result of its relevance in solving academic and real-world problems. Several techniques such as hyper-mutation, hyper-learning, hyper-selection, change detection and many more have been developed specifically for solving dynamic optimisation problems. However, the complex structure of algorithms employing these techniques make them unsuitable for real-world, real-time dynamic optimisation problem using embedded systems with limited memory. The work presented in this thesis focuses on a compact approach as an alternative to population based optimisation algorithm, suitable for solving real-time dynamic optimisation problems. Specifically, a novel compact dynamic optimisation algorithm suitable for embedded systems with limited memory is presented. Three novel dynamic approaches that augment and enhance the evolving properties of the compact genetic algorithm in dynamic environments are introduced. These are 1.) change detection scheme that measures the degree of dynamic change 2.) mutation schemes whereby the mutation rates is directly linked to the detected degree of change and 3.) change trend scheme the monitors change pattern exhibited by the system. The novel compact dynamic optimization algorithm outlined was applied to two differing dynamic optimization problems. This work evaluates the algorithm in the context of tuning a controller for a physical target system in a dynamic environment and solving a dynamic optimization problem using an artificial dynamic environment generator. The novel compact dynamic optimisation algorithm was compared to some existing dynamic optimisation techniques. Through a series of experiments, it was shown that maintaining diversity at a population level is more efficient than diversity at an individual level. Among the five variants of the novel compact dynamic optimization algorithm, the third variant showed the best performance in terms of response to dynamic changes and solution quality. Furthermore, it was demonstrated that information transfer based on dynamic change patterns can effectively minimize the exploration/exploitation dilemma in a dynamic environment.

[pt] Esta dissertação investiga o uso de Algoritmos Genéticos e de Co-Evolução Cooperativa na otimização de problemas de planejamento com restrições de precedência. Neste tipo de problema algumas ou todas as tarefas têm restrições que implicam na necessidade de planejá-las ou executá-las antes ou depois de outras. Por esta razão, o uso de modelos evolucionários convencionais como, por exemplo, os baseados em ordem pode gerar soluções inválidas, não penalizáveis, que precisam ser descartadas, comprometendo assim o desempenho do algoritmo. O objetivo do trabalho foi, portanto, estudar formas de representação de soluções para este tipo de problema capazes de gerar somente soluções válidas, bem como avaliar o desempenho dos modelos propostos. O trabalho consistiu de 3 etapas principais: um estudo sobre problemas de otimização de planejamento com algoritmos genéticos; a definição de novos modelos usando algoritmos genéticos e co-evolução cooperativa para otimização de problemas de planejamento com restrições de precedência e a implementação de uma ferramenta para estudo de caso. O estudo sobre os problemas de otimização de planejamentos com algoritmos genéticos envolveu o levantamento de representações, dificuldades e características deste tipo de problema e, mais especificamente, de representações baseadas em ordem. A modelagem do algoritmo genético consistiu fundamentalmente na definição de uma representação dos cromossomas e da função da avaliação que levasse em conta a existência de restrições de precedência (tarefas que devem ser planejadas/executadas antes de outras). A construção do modelo co-evolucionário por sua vez consistiu em definir uma nova população, com uma outra representação, que se responsabilizasse pela distribuição dos recursos para execução das tarefas, responsabilidade esta que, no modelo com algoritmos genéticos convencionais, era tratada de forma simples por um conjunto de heurísticas. Finalmente, desenvolveu-se uma ferramenta para implementar estes modelos e tratar de um estudo de caso complexo que oferecesse as características necessárias para testar a qualidade das representações e avaliar os resultados. O estudo de caso escolhido foi a otimização do planejamento da descarga, armazenamento e embarque de minério de ferro de modo a minimizar o tempo de estadia dos navios em um porto fictício. Foram realizados vários testes que demonstraram a capacidade dos modelos desenvolvidos em gerar soluções viáveis, sem a necessidade de heurísticas de correção, e os resultados obtidos foram comparados com os de um processo de busca aleatória. Em todos os casos, os resultados obtidos pelos modelos foram sempre superiores aos obtidos pela busca aleatória. No caso do modelo de representação com uma única população obteve-se resultados até 41% melhores do que com os obtidos por uma busca aleatória. No caso do modelo de representação com co-evolução o resultado ficou 33% melhor que a busca aleatória com tratamento de solução idêntico ao da solução co-evolucionária. Os resultados da co-evolução comparados com o algoritmo genético com uma única espécie foram 29% melhores. / [en] This work investigates the use of Genetic Algorithms and Cooperative Co-Evolution in optimization of scheduling problems with precedence constraints. In this kind of problem some or all tasks have constraints that imply planning or executing them before or after others. For this reason, the use of order-based conventional evolutionary models may generate invalid solutions, which cannot be penalized, needing to be discarded and therefore compromising the algorithm performance. The main goal was therefore to study models for this kind of problem that are capable of generating only valid solutions. The work was divided in 3 main steps: a survey on scheduling optimization problems using genetic algorithms; definition of two models based on genetic algorithms and cooperative co-evolution for optimizing scheduling problems with precedence constraints; and the implementation of a tool for a case study. The study on scheduling optimization problems with genetic algorithms consisted in gathering information about representations and characteristics of this kind of problem and, more specifically, about order-based representations. The genetic algorithm modeling consisted basically in defining a chromosome representation and an evaluation function that took into account the existence of precedence constraints (tasks that must be scheduled or executed before others). The co-evolutionary model consisted in defining a new population, with another representation scheme, which was responsible for distributing resources for tasks execution. On the conventional genetic algorithm model, this role was played by a simple set of heuristics. Finally, a tool was developed for implementing those models and treating a complex case study which offered the needed characteristics for testing representation performance and evaluating results. The chosen case study was the optimization of iron ore dumping, stocking and ship loading on a fictitious harbor, targeting minimization of ships waiting time. Tests were done in order to demonstrate the ability of the developed models in generating viable solutions without the need of corrective heuristics and the results were compared to the results obtained through exhaustive search. In all cases, the models` results were better than the exhaustive search ones. In the case where the representation used a single population the results obtained were up to 41% better than the ones with the exhaustive search. The co- evolutionary results outperformed the co-evolutionary search with the same solution representation by 33%. Compared to the single specie genetic algorithm, the co- evolutionary model outperformed it by 29%.