Most real-world optimization problems behave stochastically. Evolutionary optimization algorithms have to cope with the uncertainty in order to not loose a substantial part of their performance. There are different types of uncertainty and this thesis studies the type that is commonly known as noise and the use of resampling techniques as countermeasure in multi-objective evolutionary optimization. Several different types of resampling techniques have been proposed in the literature. The available techniques vary in adaptiveness, type of information they base their budget decisions on and in complexity. The results of this thesis show that their performance is not necessarily increasing as soon as they are more complex and that their performance is dependent on optimization problem and environment parameters. As the sampling budget or the noise level increases the optimal resampling technique varies. One result of this thesis is that at low computing budgets or low noise strength simple techniques perform better than complex techniques but as soon as more budget is available or as soon as the algorithm faces more noise complex techniques can show their strengths. This thesis evaluates the resampling techniques on standard benchmark functions. Based on these experiences insights have been gained for the use of resampling techniques in evolutionary simulation optimization of real-world problems.

No description available.

The initial question addressed in this thesis is how to take into account the multi-objective aspect of decision problems in inverse optimization. The most straightforward extension consists of finding a minimal adjustment of the objective functions coefficients such that a given feasible solution becomes efficient. However, there is not only a single question raised by inverse multi-objective optimization, because there is usually not a single efficient solution. The way we define inverse multi-objective<p>optimization takes into account this important aspect. This gives rise to many questions which are identified by a precise notation that highlights a large collection of inverse problems that could be investigated. In this thesis, a selection of inverse problems are presented and solved. This selection is motivated by their possible applications and the interesting theoretical questions they can rise in practice. / Doctorat en Sciences de l'ingénieur / info:eu-repo/semantics/nonPublished

Supply chains are in general complex networks composed of autonomous entities whereby multiple performance measures in different levels, which in most cases are in conflict with each other, have to be taken into account. Hence, due to the multiple performance measures, supply chain decision making is much more complex than treating it as a single objective optimization problem. Thus, the aim of the doctoral thesis is to address the supply chain optimization problem within a truly Pareto-based multi-objective context and utilize knowledge extraction techniques to extract valuable and useful information from the Pareto optimal solutions. By knowledge extraction, it means to detect hidden interrelationships between the Pareto solutions, identify common properties and characteristics of the Pareto solutions as well as discover concealed structures in the Pareto optimal data set in order to support managers in their decision making. This aim is addressed through the SBO-framework where the simulation methodology is based on system dynamics (SD) and the optimization utilizes multi-objective optimization (MOO). In order to connect the SD and MOO software, this doctoral thesis introduced a novel SD and MOO interface application which allow the modeling and optimization applications to interact. Additionally, this thesis work also presents a novel SD-MOO methodology that addresses the issue of curse off dimensionality in MOO for higher dimensional problems and with the aim to execute supply chain SD-MOO in a computationally cost efficient way, in terms of convergence, solution intensification and accuracy of obtaining the Pareto-optimal front for complex supply chain problems. In order to detect evident and hidden structures, characteristics and properties of the Pareto-optimal solutions, this work utilizes Parallel Coordinates, Clustering and Innovization, which are three different types of tools for post-optimal analysis and facilitators of discovering and retrieving knowledge from the Pareto-optimal set. The developed SD-MOO interface and methodology are then verified and validated through two academic case studies and a real-world industrial application case study. While not all the insights generated in these application studies can be generalized for other supply-chain systems, the analysis results provide strong indications that the methodology and techniques introduced in this thesis are capable to generate knowledge to support academic SCM research and real-world SCM decision making, which to our knowledge cannot be performed by other methods.

Bus stops are key links in the journeys of transit patrons with disabilities. Inaccessible bus stops prevent people with disabilities from using fixed-route bus services, thus limiting their mobility. The Americans with Disabilities Act (ADA) of 1990 prescribes the minimum requirements for bus stop accessibility by riders with disabilities. Due to limited budgets, transit agencies can only select a limited number of bus stop locations for ADA improvements annually. These locations should preferably be selected such that they maximize the overall benefits to patrons with disabilities. In addition, transit agencies may also choose to implement the universal design paradigm, which involves higher design standards than current ADA requirements and can provide amenities that are useful for all riders, like shelters and lighting. Many factors can affect the decision to improve a bus stop, including rider-based aspects like the number of riders with disabilities, total ridership, customer complaints, accidents, deployment costs, as well as locational aspects like the location of employment centers, schools, shopping areas, and so on. These interlacing factors make it difficult to identify optimum improvement locations without the aid of an optimization model. This dissertation proposes two integer programming models to help identify a priority list of bus stops for accessibility improvements. The first is a binary integer programming model designed to identify bus stops that need improvements to meet the minimum ADA requirements. The second involves a multi-objective nonlinear mixed integer programming model that attempts to achieve an optimal compromise among the two accessibility design standards. Geographic Information System (GIS) techniques were used extensively to both prepare the model input and examine the model output. An analytic hierarchy process (AHP) was applied to combine all of the factors affecting the benefits to patrons with disabilities. An extensive sensitivity analysis was performed to assess the reasonableness of the model outputs in response to changes in model constraints. Based on a case study using data from Broward County Transit (BCT) in Florida, the models were found to produce a list of bus stops that upon close examination were determined to be highly logical. Compared to traditional approaches using staff experience, requests from elected officials, customer complaints, etc., these optimization models offer a more objective and efficient platform on which to make bus stop improvement suggestions.

Liquid biofuels produced from biomass have the potential to partly replace gasoline. One of the most promising biofuels is butanol which is produced in acetone-butanol-ethanol (ABE) fermentation. The ABE fermentation is characterized by its low butanol concentration in the final fermentation broth. In this research, the simulation of three in situ recovery methods, namely, vacuum fermentation, gas stripping and pervaporation, were performed in order to increase the efficiency of the continuous ABE fermentation by decreasing the effect of butanol toxicity. The non-integrated and integrated butanol production systems were simulated and optimized based on a number of objectives such as maximizing the butanol productivity, butanol concentration, and butanol yield. In the optimization of complex industrial processes, where objectives are often conflicting, there exist numerous potentially-optimal solutions which are best obtained using multi-objective optimization (MOO). In this investigation, MOO was used to generate a set of alternative solutions, known as the Pareto domain. The Pareto domain allows to view very clearly the trade-offs existing between the various objective functions. In general, an increase in the butanol productivity resulted in a decrease of butanol yield and sugar conversion. To find the best solution within the Pareto domain, a ranking algorithm (Net Flow Method) was used to rank the solutions based on a set of relative weights and three preference thresholds. Comparing the best optimal solutions in each case study, it was clearly shown that integrating a recovery method with the ABE fermentation significantly increases the overall butanol concentration, butanol productivity, and sugar conversion, whereas butanol yield being microorganism-dependent, remains relatively constant.

Optimizing supply chain is the most successful key for manufacturing systems to be competitive. Supply chain (SC) has gotten intensive research works at all levels: strategic, tactical, and operational levels. These levels, in some researches, have integrated with each other or integrated with other planning issues such as inventory. Optimizing inventory location and level of safety stock at all supply chain partners is essential in high competitive markets to manage uncertain demand and service level. Many works have been developed to optimize the location of safety stock along supply chain, which is important for fast response to fluctuation in demand. However, most of these studies focus on the design stage of a supply chain. Because demand at different horizon times may vary according to different reasons such as the entry of different competitors on market or seasonal demand, safety stock should be optimized accordingly. At the planning (tactical) level, safety stock can be controlled according to each planning horizon to satisfy customer demand at lower cost instead of being fixed by a decision taken at the strategic level. On the other hand, most studies that consider safety stock optimization are tied to a specific system structure such as serial, assembly, or distribution structure. This research focuses on formulating two different models. First, a multi- echelon safety stock optimization (MESSO) model for general supply chain topology is formulated. Then, it is converted into a robust form (RMESSO) which considers all possible fluctuation in demand and gives a solution that is valid under any circumstances. Second, the safety stock optimization model is integrated with tactical supply chain planning (SCP) for manufacturing systems. The integrated model is a multi-objective mixed integer non-linear programming (MINLP) model. This model aims to minimize the total cost and total time. A case study for each model is provided and the numerical results are analyzed.

Production scheduling is one of the most important issues in the planning and operation of manufacturing systems. Customers increasingly expect to receive the right product at the right price at the right time. Various problems experienced in manufacturing, for example low machine utilization and excessive work-in-process, can be attributed directly to inadequate scheduling. In this dissertation a production scheduling algorithm is developed for Optimatix, a South African-based company specializing in supply chain optimization. To address the complex requirements of the customer, the problem was modeled as a flexible job shop scheduling problem with sequence-dependent set-up times, auxiliary resources and production down time. The algorithm development process focused on investigating the application of both particle swarm optimization (PSO) and differential evolution (DE) to production scheduling environments characterized by multiple machines and multiple objectives. Alternative problem representations, algorithm variations and multi-objective optimization strategies were evaluated to obtain an algorithm which performs well against both existing rule-based algorithms and an existing complex flexible job shop scheduling solution strategy. Finally, the generality of the priority-based algorithm was evaluated by applying it to the scheduling of production and maintenance activities at Centurion Ice Cream and Sweets. The production environment was modeled as a multi-objective uniform parallel machine shop problem with sequence-dependent set-up times and unavailability intervals. A self-adaptive modified vector evaluated DE algorithm was developed and compared to classical PSO and DE vector evaluated algorithms. Promising results were obtained with respect to the suitability of the algorithms for solving a range of multi-objective multiple machine scheduling problems. Copyright / Dissertation (MEng)--University of Pretoria, 2009. / Industrial and Systems Engineering / unrestricted

Orientadores: José Roberto de França Arruda, Pierre Lamary / Tese (doutorado) ¿ Universidade Estadual de Campinas, Faculdade de Engenharia Mecânica / Made available in DSpace on 2018-08-26T13:47:00Z (GMT). No. of bitstreams: 1 Pegoretti_ThaisdosSantos_D.pdf: 2527596 bytes, checksum: 4a887632523490eee648b59c0de7e4a2 (MD5) Previous issue date: 2014 / Resumo: Este trabalho tem como objetivo desenvolver uma metodologia capaz de adicionar critérios ambientais à fase de pré-projeto de um tratamento acústico veicular. Essa integração foi realizada através de uma otimização multiobjetivo baseada em um algoritmo genético. Um caso real foi analisado com a metodologia proposta. Ele consiste em um painel acústico multicamadas aplicado em um automóvel de passeio. O método da matriz de transferência é usado para o cálculo do comportamento acústico do painel. Neste método é feita a hipótese simplificadora de painel de área infinita, o que permite um custo computacional muito menor do que modelos de elementos finitos. Para a modelagem de materiais poroelásticos, utiliza-se o modelo de Johnson-Champoux-Allard, que inclui os fenômenos de dispersão de energia resultante da interação térmica e viscosa entre as fases sólida e fluida. O custo computacional menor do modelo é essencial para a otimização. Foram estabelecidos como objetivos da otimização a curva de perda de transmissão desejada e os resultados da análise do ciclo de vida do painel. Uma curva de perda de transmissão em função de bandas de oitava foi estabelecida como um critério de desempenho acústico mínimo. Para os critérios ambientais, o impacto de um painel existente foi estabelecido como máximo. A análise do ciclo de vida quantifica o impacto do produto em relação a diversos aspectos. Na metodologia proposta três critérios foram selecionados inicialmente: aquecimento global, destruição de recursos abióticos e toxicidade da água doce. Finalmente, apenas um deles foi utilizado na otimização, o aquecimento global, pois os critérios máximos estabelecidos para os demais eram facilmente atingidos ao longo da otimização. A otimização multiobjetivos gera como resultado uma frente de Pareto com um conjunto de soluções, e cabe ao projetista escolher a melhor opção, analisando-a em relação ao impacto ambiental e a outros aspectos, tais como disponibilidade e custo / Abstract: This work aims at developing a methodology capable of adding environmental criteria to the pre-design of a vehicular acoustic treatment. This integration was accomplished through a multi-objective optimization based on a genetic algorithm. A real case study was analyzed with the proposed methodology. It consists of a multilayered acoustic panel applied in passenger vehicles. The transfer matrix method is used to calculate the acoustic behavior of the panel. In this method the panel area is infinite. It provides a lower computational cost than finite element models, which can take into account the real dimensions of the panel. The Johnson-Champoux-Allard model was used for poroelastic material modeling. It includes the energy loss generated by the viscous and the thermal interactions between the solid and the fluid media. The lower computational cost of the model is essential for the optimization. The desired acoustic transmission and results of the life cycle analysis of the panel were established as the optimization objectives. A transmission loss curve in octave bands was defined as a minimum noise performance criterion. For the environmental criteria, an existing panel behavior was established as the maximum. The life cycle assessment quantifies the product impact with respect to many aspects. In the proposed methodology, three criteria were initially selected: global warming, abiotic depletion, and fresh water aquatic ecotoxicity. Finally, only one of them was used in the optimization, the global warming, because the maximum values established for the other criteria were easily achieved during the optimization. The multi-objective optimization provides a Pareto front solutions set, and it is up to the designer to choose the best option, analyzing the solution set with relation to environmental impact and other aspects, such as availability and cost / Doutorado / Mecanica dos Sólidos e Projeto Mecanico / Doutora em Engenharia Mecânica

O problema reconstrução filogenética têm como objetivo determinar as relações evolutivas das espécies, usualmente representadas em estruturas de árvores. No entanto, esse problema tem se mostrado muito difícil uma vez que o espaço de busca das possíveis árvores é muito grande. Diversos métodos de reconstrução filogenética têm sido propostos. Vários desses métodos definem um critério de otimalidade para avaliar as possíveis soluções do problema. Porém, a aplicação de diferentes critérios resulta em árvores diferentes, inconsistentes entre sim. Nesse contexto, uma abordagem multi-objetivo para a reconstrução filogenética pode ser útil produzindo um conjunto de árvores consideradas adequadas por mais de um critério. Nesta tese é proposto um algoritmo evolutivo multi-objetivo, denominado PhyloMOEA, para o problema de reconstrução filogenética. O PhyloMOEA emprega os critérios de parcimônia e verossimilhança que são dois dos métodos de reconstru ção filogenética mais empregados. Nos experimentos, o PhyloMOEA foi testado utilizando quatro bancos de seqüências freqüentemente empregados na literatura. Para cada banco de teste, o PhyloMOEA encontrou as soluções da fronteira de Pareto que representam um compromisso entre os critérios considerados. As árvores da fronteira de Pareto foram validadas estatisticamente utilizando o teste SH. Os resultados mostraram que o PhyloMOEA encontrou um número de soluções intermediárias que são consistentes com as soluções obtidas por análises de máxima parcimônia e máxima verossimilhança realizados separadamente. Além disso, os graus de suporte dos clados pertencentes às árvores encontradas pelo PhyloMOEA foram comparadas com a probabilidade posterior dos clados calculados pelo programa Mr.Bayes aplicados aos quatro bancos de teste. Os resultados indicaram que há uma relação entre ambos os valores para vários grupos de clados. Em resumo, o PhyloMOEA é capaz de encontrar uma diversidade de soluções intermediárias que são estatisticamente tão boas quanto as melhores soluções de máxima parcimônia e máxima verossimilhança. Tais soluções apresentam um compromisso entre os dois objetivos / The phylogeny reconstruction problem consists of determining the evolutionary relationships (usually represented as a tree) among species. This is a very complex problem since the tree search space is huge. Several phylogenetic reconstruction methods have been proposed. Many of them defines an optimality criterion for evaluation of possible solutions. However, different criteria may lead to distinct phylogenies, which often conflict with each other. In this context, a multi-objective approach for phylogeny reconstruction can be useful since it could produce a set of optimal trees according to mdifficultultiple criteria. In this thesis, a multi-objective evolutionary algorithm for phylogenetic reconstruction, called PhyloMOEA, is proposed. PhyloMOEA uses the parsimony and likelihood criteria, which are two of the most used phylogenetic reconstruction methods. PhyloMOEA was tested using four datasets of nucleotide sequences found in the literature. For each dataset, the proposed algorithm found a Pareto front representing a trade-off between the used criteria. Trees in the Pareto front were statistically validated using the SH-test, which has shown that a number of intermediate solutions from PhyloMOEA are consistent with solutions found by phylogenetic methods using one criterion. Moreover, clade support values from trees found by PhyloMOEA was compared to clade posterior probabilities obtained by Mr.Bayes. Results indicate a correlation between these probabilities for several clades. In summary, PhyloMOEA is able to find diverse intermediate solutions, which are not statistically worse than the best solutions for the maximum parsimony and maximum likelihood criteria. Moreover, intermediate solutions represent a trade-off between these criteria