This thesis studies Production Scheduling in a multistage hybrid flowshop facility. It first states the general Production Planning and Scheduling problem and highlights some drawbacks of classical solutions. A theoretical decomposition-based approach is introduced whose main issue is to overcome non-efficient capacity utilization. By using Branch and Bound methods, an in-depth analysis of the scheduling part of the system is then carried out throughout the study and development of upper and lower bounds as well as branching schemes. Already-existing and new heuristics are presented and compared on different shop floor configurations. Five different heuristic approaches are studied. By scheduling the HFS one stage at a time the first approach uses different stage sequencing orders. The second and third approaches are mainly list heuristics. The second approach uses ideas derived from the multistage classical flowshop with a single machine per stage, while the third approach uses classical dispatching priority rules. The fourth and fifth approaches, respectively, use random scheduling and local search techniques. Statistical analysis is carried out in order to compare the heuristics and to select the best of them for each shop configuration. Already-existing and new lower bounds on the single stage subproblem are also presented and compared. Three new lower bounds are developed: a dual heuristic based bound, a partially preemptive bound and a heuristic for the so-called subset bound. Some of these lower bounds use a network flow algorithm. A new version of the “Preflow Push” algorithm which runs faster than the original one is presented. The best lower bounds are selected based on numerical tests. Two branch and bound algorithms are presented, an improved version of the sequence enumeration method and a generalization of the so-called interval branching method, along with several bounding strategies. Based on the upper and lower bound studies, several branch and bound algorithms are presented and compared using numerical tests on different shop floor configurations. Eventually, an Object Model for Scheduling Algorithm Implementations (OMSAI), that has been used for the computer implementation of the developed algorithms, is presented.

This thesis studies Production Scheduling in a multistage hybrid flowshop facility. It first states the general Production Planning and Scheduling problem and highlights some drawbacks of classical solutions. A theoretical decomposition-based approach is introduced whose main issue is to overcome non-efficient capacity utilization. By using Branch and Bound methods, an in-depth analysis of the scheduling part of the system is then carried out throughout the study and development of upper and lower bounds as well as branching schemes. Already-existing and new heuristics are presented and compared on different shop floor configurations. Five different heuristic approaches are studied. By scheduling the HFS one stage at a time the first approach uses different stage sequencing orders. The second and third approaches are mainly list heuristics. The second approach uses ideas derived from the multistage classical flowshop with a single machine per stage, while the third approach uses classical dispatching priority rules. The fourth and fifth approaches, respectively, use random scheduling and local search techniques. Statistical analysis is carried out in order to compare the heuristics and to select the best of them for each shop configuration. Already-existing and new lower bounds on the single stage subproblem are also presented and compared. Three new lower bounds are developed: a dual heuristic based bound, a partially preemptive bound and a heuristic for the so-called subset bound. Some of these lower bounds use a network flow algorithm. A new version of the “Preflow Push” algorithm which runs faster than the original one is presented. The best lower bounds are selected based on numerical tests. Two branch and bound algorithms are presented, an improved version of the sequence enumeration method and a generalization of the so-called interval branching method, along with several bounding strategies. Based on the upper and lower bound studies, several branch and bound algorithms are presented and compared using numerical tests on different shop floor configurations. Eventually, an Object Model for Scheduling Algorithm Implementations (OMSAI), that has been used for the computer implementation of the developed algorithms, is presented.

Les travaux présentés dans cette thèse abordent la minimisation de coûts de production dans uneindustrie de produits douches et capillaires. Dans cette industrie, le processus de production inclus deuxétapes successives : la fabrication des lots de produit cosmétique et le stockage intermédiaire de ces derniers.Les coûts de production sont essentiellement liés aux opérations de lavage des ressources de fabrication etde stockage intermédiaire. Ces opérations de lavage doivent être effectuées lors de la succession des lots vuleur différentes caractéristiques physiques (couleur, viscosité,...) et chimiques (contenus chimiques,...).Ce problème est décomposé en deux sous-problèmes. Le premier consiste en l'optimisation du stockageintermédiaire. Le site dispose de plusieurs cuves de stockage, de différentes capacités, disposées en parallèle.Le rôle de ces cuves de stockage est de contenir temporairement les lots. Résoudre ce problème équivautà calculer les affectations des lots sur les cuves ainsi que leur date de début de transfert. L'objectif est deminimiser le nombre d'opérations de lavages des cuves de stockage.Le second sous-problème consiste à optimiser la fabrication des lots. Le site comprend plusieurs sallesde fabrication disposées en parallèle. Chaque salle de fabrication est constituée par plusieurs machinesorganisées en Flowshop Hybride. Pour résoudre ce problème, il faut calculer une affectation des lots sur lessalles de fabrication et les ordonnancer sur les machines de celles-ci. L'objectif est de minimiser le nombred'opérations de lavage induites par la succession des lots sur les machines.Nous proposons de résoudre le sous-problème d'optimisation du stockage intermédiaire en premier lieu,pour ensuite résoudre le sous-problème d'optimisation de la fabrication. Nous proposons et expérimentonsplusieurs méthodes heuristiques (gloutons, colonies de fourmis, méthodes arborescentes tronquées, méthodes dédiées) pour la résolution de chaque sous-problème. Les meilleures méthodes de résolution sontdestinées à être intégrées dans un logiciel de planification de la production quotidienne. / The work presented in this thesis addresses the minimization of production costs in an industry ofshowers and hair products. In this industry, the production process consists in two successive steps : themaking of cosmetic products and the intermediate storage of these latter. Production costs are mainlyrelated to cleaning operations of the making and the storage resources. These cleaning operations must beperformed in the sequence of two different batches of cosmetic product because of their different physical(color, viscosity, ...) and chemical (chemical contents,..) characteristics.This problem is decomposed into two sub-problems. The first one is the optimization of intermediatestorage. The shop is made up of parallel storage tanks of various capacities. These storage tanks haveto temporarily store the batches. Solving this problem is equivalent to calculating the assignment of thebatches on the storage tanks and their starting date of transfer. The objective is to minimize the numberof cleaning operations of the storage tanks.The second sub-problem is the optimization of the making process of the batches. The shop gathersseveral making units arranged in parallel. Each making unit consists in multiple mixing machines organizedin hybrid flowshop. To solve this problem, we have to calculate an assignment of the batches on the makingunits and their schedule on the mixing machines. The objective is to minimize the number of cleaningoperations.We propose to solve the sub-problem of optimization of the intermediate storage first, and then solvethe sub-problem of the optimization of the making process. We propose and experiment several heuristics(greedy, ant colonies, truncated tree methods, dedicated methods) for solving each sub-problem. The bestsolution methods are designed to be integrated into a software production planning.

This paper focuses on characterization of scheduling in systems with one or multiple machines. There are different types of tasks given, with which we can encounter in scheduling. At the beginning, introduce basic concepts of production scheduling. The second chapter is a flowshop problem with its history and projections for the flowshop problem. In the next chapter is a modification of flowshop problem called hybrid flowshop, which is divided according flexibility to hybrid flowshop with processing flexibility and hybrid flowshop with routing flexibility. Another chapter is open shop problem, which have some differences compare with the above mentioned types. The last chapter is a job shop, stating in a graphical solution for two machines and a brief description of the algorithm Shifting bottleneck.

Dans cette thèse, nous avons traité les problèmes d'ordonnancement d'ateliers de type flowshop hybride à deux étages avec machines à traitement par batches sur le second étage et compatibilité entre les tâches.Les durées opératoires des tâches sont données par des intervalles, et les tâches sont dites compatibles si elles partagent une même durée d'exécution. Pour le problème de minimisation de la date de fin d'ordonnancement de ce type d'atelier, nous avons développé 6 heuristiques à performances garanties. D'après les expériences réalisées, ces heuristiques sont efficaces sur de grandes instances. Pour les petites instances, nous avons présenté deux méthodes exactes de type procédures par séparation évaluation qui permettent de résoudre des instances de 20 tâches. Nous avons également développé un schéma d'approximation polynomial (PTAS) utilisable lorsque les durées d'exécution sur le premier étage sont identiques. En complément de ces travaux, nous avons également étudié d’autres problèmes de minimisation de critères réguliers sur une machine à traitement par batches. Nous avons développé des algorithmes de programmation dynamiques pseudo-polynomiaux pour les problèmes de minimisation de la somme des dates de fin d'exécution et pour les problèmes avec dates de fin souhaitées. Afin de compléter ces résultats de complexité, nous avons montré la NP-complétude des problèmes avec dates de fin souhaitées / This thesis deals with 2-stages hybrid flowshop scheduling problems with batching machines on the second stage and compatibility constraints. The processing times of tasks are given by intervals and tasks are compatible if they share a same second stage processing time. We developped 6 heuristics with worth-case analysis for the makespan minimization problem. The experimental results show that these heuristics give good schedules with an average gap of 1\% on 200 task instances. For small instances, we presented 2 exact methods, Branch \& Bounds, which solves up to 20 task instances. For the particular case with identical processing times on first stage and uniform processing time intervals we developped a Polynomial Time Approximation Scheme (PTAS). The second part of this thesis deal with scheduling problems on one batching machine with infinite capacity and regular criteria minimization. We developped pseudo-polynomial dynamic programming algorithm for minimization total completion time, maximal lateness and total tardiness. Finally we show the NP-completeness of problems with due dates

O presente trabalho aborda o ambiente de produção conhecido como flowshop híbrido. Devido a crescente complexidade dos sistemas de produção, este ambiente é frequentemente encontrado em situações reais de manufatura. No caso estudado existem estágios em série e em cada estágio existe um número de máquinas idênticas em paralelo. Os tempos de processamento em cada estágio são dependentes da tarefa, já a rota através do sistema é a mesma para todas as tarefas. O objetivo é minimizar o atraso total, ou seja, a soma do atraso de todas as tarefas. Um modelo de programação linear inteira mista é apresentado para este problema e, dada a sua complexidade, ele é abordado através de uma meta-heurística relativamente nova e que, conforme revisão da literatura, nunca foi aplicada a este problema. Conhecida por BRKGA (Biased Random-Key Genetic Algorithm), este método codifica as soluções de maneira a obter um melhor desempenho em comparação com algoritmos genéticos tradicionais. Com o objetivo de avaliar a melhor estratégia, são propostas diversas versões de BRKGA para o problema considerado. Estas versões buscam explorar características das melhores heurísticas construtivas da literatura, dentre estas: ordens direta e inversa de programação das tarefas dentro do ambiente produtivo, identificação do estágio gargalo e diferenciação da programação do gargalo dos demais estágios. Experimentos computacionais foram realizados com 432 problemas teste de grande porte. Os métodos apresentados são comparados entre si e os resultados mostraram que uma versão do BRKGA se destaca frente às demais, visto que ela atingiu o melhor resultado em 61% dos problemas. Destaca-se que o método de melhor desempenho da literatura obteve a melhor solução em apenas 15% dos problemas. Devido às dimensões dos problemas teste da literatura, não foi possível encontrar suas soluções ótimas. Deste modo, este trabalho propõe um novo limitante inferior para o mínimo atraso total. Além disso, 576 novos problemas teste de menores dimensões são propostos e seus resultados ótimos são utilizados para aprofundar as comparações. Os resultados deste experimento indicaram que o BRKGA proposto apresentou um bom desempenho visto que, na média, seus resultados estão apenas a 2,4% dos resultados ótimos. / This work addresses a scheduling problem in hybrid flowshops. Due to the increasing complexity of production systems, this production environment is often encountered in real manufacturing situations. In hybrid flowshops, there are stages in series and, in each stage, a number of similar parallel machines. Processing times in each stage are dependent on the job, and the route through the system is the same for all jobs. The objective is to minimize the total tardiness, that is, the sum of all jobs tardiness. A mixed integer linear programming model is presented for the problem considered. Given its complexity, this problem is approached by a relatively new meta-heuristic, known as BRKGA (Biased Random-Key Genetic Algorithm). A literature review showed that BRKGA had never been applied to this problem. The BRKGA codes solutions in order to obtain a better performance compared with traditional genetic algorithms. Several versions of BRKGA were developed in order to evaluate the best strategy to solve the problem considered. These versions aim to exploit features of the best constructive heuristic from the literature, among them: scheduling jobs in direct and inverse order within the production environment, identification of the bottleneck stage and distinction of the bottleneck stage schedule from the others. Computational experiments were conducted with 432 large instances. The methods were compared and the results showed that one of these versions stood out against the others. This version achieved better results in 61% of instances, while the best heuristic from the literature achieved 15%. Due to the size of these instances, optimal solutions were not found. Therefore, this work develops a new lower bound for the minimum total tardiness. Additionally, in order to find optimal results, a set of 576 new instances is proposed. This experiment indicated that the BRKGA proposed performed well since, on average, their results are only 2.4% away from the optimal results.

O presente trabalho aborda o ambiente de produção conhecido como flowshop híbrido. Devido a crescente complexidade dos sistemas de produção, este ambiente é frequentemente encontrado em situações reais de manufatura. No caso estudado existem estágios em série e em cada estágio existe um número de máquinas idênticas em paralelo. Os tempos de processamento em cada estágio são dependentes da tarefa, já a rota através do sistema é a mesma para todas as tarefas. O objetivo é minimizar o atraso total, ou seja, a soma do atraso de todas as tarefas. Um modelo de programação linear inteira mista é apresentado para este problema e, dada a sua complexidade, ele é abordado através de uma meta-heurística relativamente nova e que, conforme revisão da literatura, nunca foi aplicada a este problema. Conhecida por BRKGA (Biased Random-Key Genetic Algorithm), este método codifica as soluções de maneira a obter um melhor desempenho em comparação com algoritmos genéticos tradicionais. Com o objetivo de avaliar a melhor estratégia, são propostas diversas versões de BRKGA para o problema considerado. Estas versões buscam explorar características das melhores heurísticas construtivas da literatura, dentre estas: ordens direta e inversa de programação das tarefas dentro do ambiente produtivo, identificação do estágio gargalo e diferenciação da programação do gargalo dos demais estágios. Experimentos computacionais foram realizados com 432 problemas teste de grande porte. Os métodos apresentados são comparados entre si e os resultados mostraram que uma versão do BRKGA se destaca frente às demais, visto que ela atingiu o melhor resultado em 61% dos problemas. Destaca-se que o método de melhor desempenho da literatura obteve a melhor solução em apenas 15% dos problemas. Devido às dimensões dos problemas teste da literatura, não foi possível encontrar suas soluções ótimas. Deste modo, este trabalho propõe um novo limitante inferior para o mínimo atraso total. Além disso, 576 novos problemas teste de menores dimensões são propostos e seus resultados ótimos são utilizados para aprofundar as comparações. Os resultados deste experimento indicaram que o BRKGA proposto apresentou um bom desempenho visto que, na média, seus resultados estão apenas a 2,4% dos resultados ótimos. / This work addresses a scheduling problem in hybrid flowshops. Due to the increasing complexity of production systems, this production environment is often encountered in real manufacturing situations. In hybrid flowshops, there are stages in series and, in each stage, a number of similar parallel machines. Processing times in each stage are dependent on the job, and the route through the system is the same for all jobs. The objective is to minimize the total tardiness, that is, the sum of all jobs tardiness. A mixed integer linear programming model is presented for the problem considered. Given its complexity, this problem is approached by a relatively new meta-heuristic, known as BRKGA (Biased Random-Key Genetic Algorithm). A literature review showed that BRKGA had never been applied to this problem. The BRKGA codes solutions in order to obtain a better performance compared with traditional genetic algorithms. Several versions of BRKGA were developed in order to evaluate the best strategy to solve the problem considered. These versions aim to exploit features of the best constructive heuristic from the literature, among them: scheduling jobs in direct and inverse order within the production environment, identification of the bottleneck stage and distinction of the bottleneck stage schedule from the others. Computational experiments were conducted with 432 large instances. The methods were compared and the results showed that one of these versions stood out against the others. This version achieved better results in 61% of instances, while the best heuristic from the literature achieved 15%. Due to the size of these instances, optimal solutions were not found. Therefore, this work develops a new lower bound for the minimum total tardiness. Additionally, in order to find optimal results, a set of 576 new instances is proposed. This experiment indicated that the BRKGA proposed performed well since, on average, their results are only 2.4% away from the optimal results.

No description available.

Due to the current trends in business as the necessity to have a large catalogue of products, orders that increase in frequency but not in size, globalisation and a market that is increasingly competitive, the production sector faces an ever harder economical environment. All this raises the need for production scheduling with maximum efficiency and effectiveness. The first scientific publications on production scheduling appeared more than half a century ago. However, many authors have recognised a gap between the literature and the industrial problems. Most of the research concentrates on optimisation problems that are actually a very simplified version of reality. This allows for the use of sophisticated approaches and guarantees in many cases that optimal solutions are obtained. Yet, the exclusion of real-world restrictions harms the applicability of those methods. What the industry needs are systems for optimised production scheduling that adjust exactly to the conditions in the production plant and that generates good solutions in very little time. This is exactly the objective in this thesis, that is, to treat more realistic scheduling problems and to help closing the gap between the literature and practice. The considered scheduling problem is called the hybrid flowshop problem, which consists in a set of jobs that flow through a number of production stages. At each of the stages, one of the machines that belong to the stage is visited. A series of restriction is considered that include the possibility to skip stages, non-eligible machines, precedence constraints, positive and negative time lags and sequence dependent setup times. In the literature, such a large number of restrictions has not been considered simultaneously before. Briefly, in this thesis a very realistic production scheduling problem is studied. Various optimisation methods are presented for the described scheduling problem. A mixed integer programming model is proposed, in order to obtai / Urlings ., T. (2010). Heuristics and metaheuristics for heavily constrained hybrid flowshop problems [Tesis doctoral no publicada]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/8439 / Palancia

Made available in DSpace on 2016-06-02T19:53:31Z (GMT). No. of bitstreams: 1 LUGO_Pedro_2013.pdf: 2147400 bytes, checksum: a0c7948826b7f243c447b99fd96c3388 (MD5) Previous issue date: 2013-09-12 / Financiadora de Estudos e Projetos / This research studies the hybrid flowshop scheduling problem. In this production configuration, we have a set of jobs that has to be processed in a set of stages. At every stage we have a set of parallel machines available to process the jobs. All jobs have to be processed following the same production flow, from the first to the last stage. Every job has to be processed on one machine at each stage and each machine can process at most one job at a time. Some constraints commonly found in real production systems as unrelated parallel machines, limited buffers, sequence-dependent setup times (both anticipatory and non-anticipatory), machine eligibility, transportation times and release times for machines are also taken into account. The optimization criterion is the makespan, whose minimization is related to the efficient use of production resources. A mixed integer programming model is proposed and solved by the commercial solver CPLEX. The computational evaluation results indicate that the model is suitable just to solve instances up to nine jobs and five stages. Therefore, to solve larger instances (50-100 jobs), several heuristics and an iterated local search (ILS) algorithm are proposed and evaluated computationally. The results indicate that the ILS is able to obtain good quality solutions in short computation times. / Este trabalho estuda o problema de programação da produção em sistemas Flowshop híbrido. Nesta configuração de produção há um conjunto de tarefas que deve ser processado em um conjunto de estações, nas quais um determinado número de máquinas paralelas encontra-se disponível para o processamento das tarefas. Todas as tarefas devem ser processadas seguindo o mesmo fluxo de produção, desde a primeira até a última estação. Cada tarefa deve ser processada em uma máquina de cada estação e cada máquina pode processar, no máximo, uma tarefa por vez. Algumas restrições comumente encontradas em sistemas de produção reais, como máquinas paralelas não relacionadas, buffers limitados, tempos de preparação dependentes da sequência (antecipatórios e não antecipatórios), elegibilidade de máquinas, tempos de transporte e tempos de liberação das máquinas, também são consideradas. O critério de otimização é o makespan, cuja minimização está diretamente relacionada com a utilização eficiente dos recursos de produção. Um modelo de programação inteira mista é proposto e resolvido através do solver comercial CPLEX. Os resultados da avaliação computacional indicam que o modelo é viável somente para resolver instâncias de até nove tarefas e cinco estações. Desta forma, para resolver instâncias de maior tamanho (50-100 tarefas), várias heurísticas e uma meta-heurística de busca local iterada (ILS, Iterated Local Search) são propostas e avaliadas computacionalmente. Os resultados indicam que o ILS é capaz de obter soluções de boa qualidade em curtos tempos computacionais.