Diversas pesquisas e estudos científicos mostram que uma grande porcentagem das pessoas com deficiência é excluída do mercado de trabalho, sobretudo em países em desenvolvimento. Com o intuito de alterar essa realidade, destacam-se, entre outras medidas, a criação de Centros de Trabalho para Deficientes (CTDs). Tais organizações empregam trabalhadores com deficiência em vários setores empresariais, dando-lhes oportunidades iniciais e preparando-os para que possam, mais tarde, ser inseridos no mercado de trabalho convencional. Vários destes centros operam linhas de produção, principal objeto de estudo desta tese. Nosso estudo é situado em uma etapa idealmente posterior aos CTDs, referente à inserção de trabalhadores com deficiência em linhas de produção convencionais. A demanda por estudos neste contexto tem crescido nos últimos anos, devido sobretudo a políticas corporativas de responsabilidade social e exigências legislativas, como a \"Lei das Cotas\", presentes em diversos países. O planejamento da operação de linhas de produção na presença de trabalhadores com deficiência envolve uma série de desafios, devido à heterogeneidade entre trabalhadores, que faz com que o tempo de execução das tarefas seja dependente de cada indivíduo. Nos deparamos, assim, com um problema de dupla alocação, em que as variáveis de decisão determinam as tarefas a serem inseridas em estações e a alocação de trabalhadores para as mesmas, de modo a otimizar alguma medida de eficiência. O balanceamento de linhas de produção convencionais com uma parcela de trabalhadores com deficiência é denominado problema de balanceamento de linhas de produção e integração de trabalhadores (ALWIBP, do inglês: assembly line worker integration and balancing problem), sendo um caso particular do problema de balanceamento de linhas de produção e designação de trabalhadores (ALWABP, do inglês: assembly line worker assignment and balancing problem), cuja ocorrência é mais comum em linhas de CTDs. Nosso objetivo consiste em estudar formas eficientes de proporcionar a integração de trabalhadores com deficiência em linhas convencionais. Para tanto, abordamos variações do ALWIBP que consideram: (i) minimização de diferentes funções objetivo (número de estações ou tempo de ciclo); (ii) linha de produção com leiautes distintos (simples ou em U); (iii) incertezas quanto ao tempo de execução de cada tarefa (abordagem robusta); (iv) estratégias de rotação de tarefas ou alocação de trabalhadores com deficiência na linha com espaçamento regular. Para cada uma destas extensões, foram desenvolvidos formulações matemáticas, métodos de resolução e novos conjuntos de instâncias teste. Experimentos computacionais indicam possibilidades de adaptação de linhas de produção convencionais à inserção de trabalhadores com deficiência, a custos adicionais baixos ou quase nulos. Portanto, este trabalho oferece alternativas para uma maior flexibilidade na integração de pessoas com deficiência, tornando-os tão eficientes quanto qualquer outro trabalhador denominado \"convencional\". / A number of studies show that a large percentage of disabled people are excluded from the labor market, in particular in developing countries. In order to deal with this problem, one can highlight the importance of Sheltered Work Centers for Disabled (SWDs). These organizations employ disabled workers in various corporate sectors, giving them initial opportunities and preparing them so that they can be later integrated into the conventional labor market. Many of these centers operate assembly lines, the main object of study of this thesis. Our study considers an ideally later stage of SWDs, related with the insertion of disabled workers in conventional assembly lines. The demand for studies in this field has grown over the years, due to corporate social responsibility policies and legal requirements such as \"quotas legislations\", present in many countries. Planning the operation of assembly lines with disabled workers involves a series of challenges due to the heterogeneity among workers, which are reflected in task times being worker dependent. This results in a double allocation problem, where decisions must determine both the tasks and the workers to be assigned to the stations, in order to optimize some efficiency measure. The conventional assembly line balancing with a parcel of disabled workers is known as the assembly line worker integration and balancing problem (ALWIBP), being a particular case of the assembly line worker assignment and balancing problem (ALWABP), which occurance is more common in SWDs. Our goal consists in studying efficient ways to promote the integration of people with disabilities in conventional assembly lines. For that, we address ALWIBP variants that consider: (i) minimization of different objective functions (number of stations or cycle time); (ii) different assembly line layouts (simple or U-shaped); (iii) uncertainties on task execution times (robust approach); (iv) job rotation strategies or allocation of disabled workers in the line with regular spacing. For each of these extensions, we develop mathematical formulations, solution methods and new sets of benchmark instances. Computational experiments indicate possibilities for adapting conventional assembly lines to the insertion of disabled workers, at low or close to null additional costs. Therefore, this study offers alternatives ways of increasing exibility in the integration of people with disabilities, making them as efficient as any other conventional worker.

Nesta tese é realizado um estudo de caso em uma indústria química brasileira, no qual busca-se representar características da tomada de decisões para a programação da produção em plantas de bateladas. Para isso, foi proposto um modelo matemático do tipo MIP (Mixed Integer Programming) que considerou a disponibilidade de matérias-primas, múltiplas tarefas produtivas para um mesmo produto, tanques de armazenamento multiproduto, envase de produtos e demanda de produtos a granel e envasados. O objetivo principal desse estudo era permitir a obtenção de soluções compatíveis com a prática da empresa em tempo de processamento viável. A partir desse estudo de caso, foi efetuado um segundo estudo com objetivo de avaliar o desempenho de formulações matemáticas para a resolução de um problema de programação da produção. Foram considerados modelos clássicos das comunidades científicas de pesquisa operacional e de engenharia de sistemas de processo, além de um terceiro modelo desenvolvido a partir de conceitos dessas duas comunidades. Algumas características do estudo de caso não foram retratadas, como o consumo de matérias-primas e o envase dos produtos, porém, foram consideradas duas características comumente observadas em problemas da indústria de processos: bateladas com quantidade produzida flexível e tarefas que produzem mais de um produto. Por fim, um terceiro estudo foi realizado com base no estudo de caso da indústria química brasileira, porém, com um foco decisões mais próximas ao nível tático. Sendo assim, foi considerado apenas o dimensionamento de lotes, sem o sequenciamento da produção. Por outro lado, foram acrescentadas características pertinentes à aquisição de matérias-primas, como custos das matérias-primas e descontos por quantidade adquirida. O objetivo deste último trabalho era avaliar a influência da integração das decisões de dimensionamento de lotes e de aquisição das matérias-primas nos custos da cadeia produtiva durante todo o horizonte de planejamento. / In this thesis we developed a study case in a Brazilian chemical industry, in which the aim was to represent the characteristics of decision-making for production scheduling in batch plants. For this, a mixed integer programming model was proposed to consider the availability of raw materials, multiple productive tasks for the same product, multi-product storage tanks, product packaging and demand for products in bulk and packaged. The main objective of this study was develop a model that is able to obtain solutions that clould be used in practice for this chemical industry in viable processing time. From this study case, a second work was carried out to evaluate the performance of mathematical formulations to solve a problem of production scheduling. Classic models of operational research and process system engineering communities were considered, and a third model was developed from concepts of these two communities. Some features of the case study were not modelled, such as the consumption of raw materials and the product packaging, however, two characteristics usually present in process industries were considered: flexible batch production quantity and multi-product task production. Finally, a third study developed based on the study case of the Brazilian chemical industry, but with focus on decisions more familiar to the tactical level. Thus, only lot sizing was modelled, without production scheduling. On the other hand, features relevant raw material purchasing were included, such as raw material costs and discounts for quantity purchased. The objective of this last work was to evaluate the influence of integrating lot sizing decisions and raw material purchasing decisions in the overall costs of the production chain during the entire planning horizon.

Este texto trata de um dos temas fundamentais no planejamento de produção, o problema de dimensionamento de lotes de um único item. Uma descrição sucinta e informal do problema segue abaixo. Considere um intervalo de tempo dividido em períodos e que a cada período de tempo está associada a demanda de um item. Dados os custos e as eventuais restrições na produção e no armazenamento, determine os períodos em que se produzirá e em que quantidade para que as demandas sejam atendidas com o menor custo possível, respeitando as restrições impostas. Apresentamos aqui resultados sobre a estrutura ótima do problema, sobre complexidade e algoritmos para os casos básicos do problema / This text studies one of the core subjects in production planning, the single-item lot-sizing problem. A brief and informal description of this problem follows below. Considering a time interval split into time periods and that there is a demand of an item associated with each time period. Given production and holding costs and possibly production and holding restrictions, determine in which periods the production must occur and in which quantity, in order to attend the demands with a minimum cost, without violate any restriction. Here, it will be shown some results about the optimal structure of the problem, about the complexity and algorithms for the simpler cases

Optimization plays an important role in many methods routinely used in statistics, machine learning and data science. Often, implementations of these methods rely on highly specialized optimization algorithms, designed to be only applicable within a specific application. However, in many instances recent advances, in particular in the field of convex optimization, make it possible to conveniently and straightforwardly use modern solvers instead with the advantage of enabling broader usage scenarios and thus promoting reusability. This paper introduces the R Optimization Infrastructure which provides an extensible infrastructure to model linear, quadratic, conic and general nonlinear optimization problems in a consistent way. Furthermore, the infrastructure administers many different solvers, reformulations, problem collections and functions to read and write optimization problems in various formats. / Series: Research Report Series / Department of Statistics and Mathematics

O ambiente produtivo denominado job shop representa empresas manufatureiras com características como: alta variedade de produtos, volume baixo de produção e uma fábrica dividida em áreas funcionais. O problema abordado neste trabalho trata da determinação do programa de produção (scheduling) de cada lote de produtos no ambiente job shop, com a premissa de que cada produto a ser elaborado surge através de um pedido realizado pelo cliente com especificações e particularidades próprias. O objetivo do trabalho é apresentar e examinar de forma detalhada as formulações matemáticas do tipo linear inteira mista (PLIM), encontradas na literatura para o ambiente que consideram a função objetivo do makespan. Além disso, se estabelece uma nova formulação matemática que auxilia a simulação do ambiente. Todas as formulações foram comparadas através de suas dimensões e testes computacionais. Adicionalmente são apresentadas três diferentes estratégias de resolução que permitem a exploração de soluções obtidas através de diferentes metodologias. A primeira estratégia estabelece para cada instância uma solução inicial que promove uma redução do número de combinações a serem avaliadas pelo software, a segunda estratégia combina duas formulações tornando uma formulação unificada, e a terceira estratégia, estabelece um processo que utiliza duas formulações de forma consecutiva compondo um procedimento sistemático. Experimentos computacionais indicam que a formulação com melhor desempenho para o problema de job shop é a formulação de Manne (1960) por obter o melhor limitante superior (upper bound). A formulação proposta apresenta o melhor limitante inferior (lower bound). Todas as formulações melhoram seus resultados através do uso das estratégias propostas. / The operational job shop environment, represents manufacturing companies with high product variety, low volume production and an organization divided into functional areas. The problem addressed in this work determines the production schedule of each batch production, with the premise that each product results from a request made by the client with specifications and its own particularities. The main objective here is to present and to examine in detail the mathematical integer - linear program formulations (MILP) from the literature for the job shop classic environment, which considers the makespan objective. Furthermore, a new mathematical formulation is provided to help with the simulation of the environment. All the formulations were compared by mathematical dimensions and computational tests. In addition, three different strategies are presented to promote the exploration of solutions obtained from new methodologies. The first strategy defines an initial solution for each problem and promotes a reduction of the combination number to be evaluated by the software. The second strategy considers the combination of two mathematical formulations under one objective function. The third strategy establishes a procedure in which two mathematical formulations are used consecutively, creating a systematic procedure. Computational experiments demonstrate that the best formulation for the job shop problem is the Manne (1960) formulation, since it obtains the best upper bound. The proposal formulation obtains the best lower bound. All of the formulations improve their results through the use of the proposed strategies.

Este texto trata de um dos temas fundamentais no planejamento de produção, o problema de dimensionamento de lotes de um único item. Uma descrição sucinta e informal do problema segue abaixo. Considere um intervalo de tempo dividido em períodos e que a cada período de tempo está associada a demanda de um item. Dados os custos e as eventuais restrições na produção e no armazenamento, determine os períodos em que se produzirá e em que quantidade para que as demandas sejam atendidas com o menor custo possível, respeitando as restrições impostas. Apresentamos aqui resultados sobre a estrutura ótima do problema, sobre complexidade e algoritmos para os casos básicos do problema / This text studies one of the core subjects in production planning, the single-item lot-sizing problem. A brief and informal description of this problem follows below. Considering a time interval split into time periods and that there is a demand of an item associated with each time period. Given production and holding costs and possibly production and holding restrictions, determine in which periods the production must occur and in which quantity, in order to attend the demands with a minimum cost, without violate any restriction. Here, it will be shown some results about the optimal structure of the problem, about the complexity and algorithms for the simpler cases

abstract: Fraud is defined as the utilization of deception for illegal gain by hiding the true nature of the activity. While organizations lose around $3.7 trillion in revenue due to financial crimes and fraud worldwide, they can affect all levels of society significantly. In this dissertation, I focus on credit card fraud in online transactions. Every online transaction comes with a fraud risk and it is the merchant's liability to detect and stop fraudulent transactions. Merchants utilize various mechanisms to prevent and manage fraud such as automated fraud detection systems and manual transaction reviews by expert fraud analysts. Many proposed solutions mostly focus on fraud detection accuracy and ignore financial considerations. Also, the highly effective manual review process is overlooked. First, I propose Profit Optimizing Neural Risk Manager (PONRM), a selective classifier that (a) constitutes optimal collaboration between machine learning models and human expertise under industrial constraints, (b) is cost and profit sensitive. I suggest directions on how to characterize fraudulent behavior and assess the risk of a transaction. I show that my framework outperforms cost-sensitive and cost-insensitive baselines on three real-world merchant datasets. While PONRM is able to work with many supervised learners and obtain convincing results, utilizing probability outputs directly from the trained model itself can pose problems, especially in deep learning as softmax output is not a true uncertainty measure. This phenomenon, and the wide and rapid adoption of deep learning by practitioners brought unintended consequences in many situations such as in the infamous case of Google Photos' racist image recognition algorithm; thus, necessitated the utilization of the quantified uncertainty for each prediction. There have been recent efforts towards quantifying uncertainty in conventional deep learning methods (e.g., dropout as Bayesian approximation); however, their optimal use in decision making is often overlooked and understudied. Thus, I present a mixed-integer programming framework for selective classification called MIPSC, that investigates and combines model uncertainty and predictive mean to identify optimal classification and rejection regions. I also extend this framework to cost-sensitive settings (MIPCSC) and focus on the critical real-world problem, online fraud management and show that my approach outperforms industry standard methods significantly for online fraud management in real-world settings. / Dissertation/Thesis / Doctoral Dissertation Computer Science 2019

Terrorism threatens both international peace and security and is a national concern. It is believed that terrorist organizations rely heavily on a few key leaders and that destroying such an organization's leadership is essential to reducing its influence. Martonosi et al. (2011) argues that increasing the amount of communication through a key leader increases the likelihood of detection. If we model a covert organization as a social network where edges represent communication between members, we want to determine the subset of members to remove that maximizes the amount of communication through the key leader. A mixed-integer linear program representing this problem is presented as well as a decomposition for this optimization problem. As these approaches prove impractical for larger graphs, often running out of memory, the last section focuses on structural characteristics of vertices and subsets that increase communication. Future work should develop these structural properties as well as heuristics for solving this problem.

Improving performance of wind turbines through effective control strategies to reduce the power generation cost is highly desired by the wind industry. The majority of the literature on performance of wind turbines has focused on models derived from principles versed in physics. Physics-based models are usually complex and not accurate due to the fact that wind turbines involve mechanical, electrical, and software components. These components interact with each other and are subjected to variable loads introduced by the wind as well as the rotating elements of the wind turbine. Recent advances in data acquisition systems allow collection of large volumes of wind energy data. Although the prime purpose of data collection is monitoring conditions of wind turbines, the collected data offers a golden opportunity to address most challenging issues of wind turbine systems. In this dissertation, data mining is applied to construct accurate models based on the turbine collected data. To solve the data-driven models, evolutionary computation algorithms are applied. As data-driven based models are non-parametric, the evolutionary computation approach makes an ideal solution tool. Optimizing wind turbines with different objectives is studied to accomplish different research goals. Two research directions of wind turbines performance are pursued, optimizing a wind turbine performance and optimizing a wind farm performance. The goal of single wind turbine optimization is to improve wind turbine efficiency and its life-cycle. The performance optimization of a wind farm is to minimize the total cost of operating a wind farm based on the computed turbine scheduling strategies. The methodology presented in the dissertation is applicable to processes besides wind industry.

Consider the problem of maximizing the revenue generated by tolls set on a subset of arcs of a transportation network, where origin-destination flows (commodities) are assigned to shortest paths with respect to the sum of tolls and initial costs. This thesis is concerned with a particular case of the above problem, in which all toll arcs are connected and constitute a path, as occurs on highways. Further, as toll levels are usually computed using the highway entry and exit points, a complete toll subgraph is considered, where each toll arc corresponds to a toll subpath. Two variants of the problem are studied, with or without specific constraints linking together the tolls on the arcs. The problem is modelled as a linear mixed integer program, and proved to be NP-hard. Next, several classes of valid inequalities are proposed, which strengthen important constraints of the initial model. Their efficiency is first shown theoretically, as these are facet defining for the restricted one and two commodity problems. Also, we prove that some of the valid inequalities proposed, together with several constraints of the linear program, provide a complete description of the convex hull of feasible solutions for a single commodity problem. Numerical tests have also been conducted, and highlight the real efficiency of the valid inequalities for the multi-commodity case. Finally, we point out the links between the problem studied in the thesis and a more classical design and pricing problem in economics. / Considérons le problème qui consiste à maximiser les profits issus de la tarification d’un sous-ensemble d’arcs d’un réseau de transport, où les flots origine-destination (produits) sont affectés aux plus courts chemins par rapport aux tarifs et aux coûts initiaux. Cette thèse porte sur une structure de réseau particulière du problème ci-dessus, dans laquelle tous les arcs tarifables sont connectés et forment un chemin, comme c’est le cas sur une autoroute. Étant donné que les tarifs sont habituellement déterminés selon les points d’entrée et de sortie sur l’autoroute, nous considérons un sous-graphe tarifable complet, où chaque arc correspond en réalité à un sous-chemin. Deux variantes de ce problème sont étudiées, avec ou sans contraintes spécifiques reliant les niveaux de tarifs sur les arcs. Ce problème peut être modélisé comme un programme linéaire mixte entier. Nous prouvons qu’il est NP-difficile. Plusieurs familles d’inégalités valides sont ensuite proposées, celles-ci renforçant certaines contraintes du modèle initial. Leur efficacité est d’abord démontrée de manière théorique, puisqu’il s’agit de facettes des problèmes restreints à un ou deux produits. Certaines des inégalités valides proposées, ainsi que plusieurs contraintes du modèle initial, permettent aussi de donner une description complète de l’enveloppe convexe des solutions réalisables d’un problème restreint à un seul produit. Des tests numériques ont également été menés, et mettent en évidence l’efficacité réelle des inégalités valides pour le problème général à plusieurs produits. Enfin, nous soulignons les liens entre le problème de tarification de réseau étudié dans cette thèse et un problème plus classique de tarification de produits en gestion.