Automotive assembly plants work on a pre-planned job sequence in order to optimize the performance of the assembly line. However, the job sequence becomes scrambled due to factors such as plant layout, process design, variability and uncertainty. Assembly plants use either a mix-bank or an automatic storage and retrieval system to regenerate the sequence before final assembly. A mix-bank, which is a set of parallel lanes, is the most common method used in the automotive industry to reconstruct the sequence. Only the first vehicles on the lanes are available for sequencing in a mix-bank set-up. Hence the lane selection policy and the lane configuration of a mix-bank play crucial roles in recreating the sequence. This thesis addresses the problem of identifying a superior lane selection policy for a mix-bank re-sequencing area. Simulation models of a re-sequencing area are used to evaluate lane selection policies. Varying the lane configurations and the nature of sequence tests the effectiveness of the selection policies.

This work develops optimal assembly sequences for modular building blocks. The underlying concept is that an automated device could take a virtual shape such as a CAD file, and automatically decide how to physically build the shape using simple, identical building blocks. This entails deciding where to place blocks inside the shape and generating an efficient assembly sequence that a robot could use to build the shape. The blocks are defined in a general, parameterized manner such that the model can be easily modified in the future. The primary focus of this work is the development of methods for generating assembly sequences in a time-feasible manner that ensure static stability at each step of the assembly. Most existing research focuses on complete enumeration of every possible assembly sequence and evaluation of many possible sequences. This, however, is not practical for systems with a large number of parts for two reasons: (1) the number of possible assembly sequences is exponential in the number of parts, and (2) each static stability test is very time-consuming. The approach proposed here is to develop a multi-hierarchical rule-based approach to assembly sequences. This is accomplished by formalizing and justifying both high-level and mid-level assembly rules based on static considerations. Application of these rules helps develop assembly sequences rapidly. The assembly sequence is developed in a time-feasible manner according to the geometry of the structure, rather than evaluating statics along the way. This work only evaluates the static stability of each step of the assembly once. The behavior of the various rules is observed both numerically and through theory, and guidelines are developed to suggest which rules to apply. A secondary focus of this work is to introduce methods by which the inside of the structure can be optimized. This structure optimization research is implemented by genetic algorithms that solve the multi-objective optimization problem in two dimensions, and can be extended to three dimensions.

No / This paper studies the selective maintenance problem for multi-state systems with structural dependence. Each component can be in one of multiple working levels and several maintenance actions are possible to a component in a maintenance break. The components structurally form multiple hierarchical levels and dependence groups. A directed graph is used to represent the precedence relations of components in the system. A selective maintenance optimization model is developed to maximize the system reliability in the next mission under time and cost constraints. A backward search algorithm is used to determine the assembly sequence for a selective maintenance scenario. The maintenance model helps maintenance managers in determining the best combination of maintenance activities to maximize the probability of successfully completing the next mission. Examples showing the use of the proposed method are presented.

Dans le contexte économique actuel, il faut proposer des produits personnalisés dequalité, à faible coût et dans des délais de plus en plus courts. La société MABI a choisi devoir chacune de ces contraintes comme une opportunité de repenser ses produits en misantsur l’innovation. Il faut alors optimiser certaines tâches routinières d’ingénierie afin dedégager du temps pour la conception des nouveaux produits. Le travail de recherche réalisés’inscrit dans le cadre d’une thèse en convention CIFRE en partenariat entre la société MABIet le laboratoire IRTES-M3M de l’UTBM. MABI conçoit, assemble, commercialise et assurele service après-vente de produits propres dans le domaine de la protection et la rénovationdes bâtiments. Ses besoins d’amélioration concernent le processus de développement deproduits qui doivent répondre aux besoins des clients tout en respectant des contraintesd’assemblage spécifiques à l’entreprise. La finalité industrielle de la thèse consiste à décliner au niveau du domaine du« Produit », la méthodologie générique élaborée sur la base de notre travail de recherchescientifique. A ce niveau, notre problématique scientifique consiste à rendre opérationnel etdynamique le modèle Multi-Domaines et Multi-Vues (MD-MV), structuré de manière «plutôtstatique», en y apportant des éléments de raisonnement contribuant à créer des interactionsinter-domaines et inter-points de vue. Pour ce qui est du domaine du « Produit », il endécoule la méthodologie FARD (Functional And Robust Design) qui vise à concevoir et àgénérer rapidement l’ensemble des variantes de produits d’une même famille modulaire touten assurant le respect des besoins clients (conception fonctionnelle) et des contraintesd’assemblage à travers une aide à la décision pour le choix de la séquence d’assemblage,contribuant ainsi à créer une interaction dynamique avec le domaine du « Process ». Quatrethèmes de recherche sont abordés : la modularité, la conception fonctionnelle, la conceptionpour l’assemblage (dès les phases amont du processus de conception) et la simulation(accélérée grâce au paramétrage du maillage). Habituellement, le domaine de la modularitéest souvent associé à celui de la conception fonctionnelle ou encore à celui de la conceptionpour l’assemblage, mais rarement les trois ensemble, ce qui constitue la spécificité de nostravaux. Enfin, l’aspect paramétrique de la méthodologie FARD, à travers les liens établisentre les quatre thèmes de recherche évoqués précédemment, rend possible la générationrapide des produits d’une même famille à partir d’un produit générique et ainsi de gagner dutemps de conception, en vue d‘atteindre nos objectifs de conception routinière « HautementProductive ». Trois cas d’études industriels et académiques illustrent l’application et lafaisabilité la méthodologie FARD... / In current economic context, enterprises must provide quality custom products at alower cost and a shorter delay. MABI Company chose to consider these constraints as anopportunity to rethink its products through innovation. Then certain routine tasks must beoptimized to free up time in order to have more time to innovate and design new products.This thesis is part of a CIFRE partnership between the MABI Company and the IRTES-M3Mlaboratory at UTBM. MABI designs, assembles, sells and provides after-sales service ofproducts in the field of the protection and the renovation of buildings. MABI needs ofimprovement are in the development process of its products that must meet customer needs,while respecting its assembly constraints.The industrial purpose of the thesis is to decline in the Product domain, the genericmethodology developed on the basis of our scientific research work. At this level, ourscientific problematic is to make operational and dynamic the Multi-Domains and Multi-Viewsdesign model (MD-MV), structured in a "rather static" way, and to enriched these models byadding reasoning procedures. It follows the FARD methodology (Functional And RobustDesign) which aims to design and quickly generate variants of a modular product family whileensuring compliance with customer requirements (functional design) and assemblyconstraints. Four domains are covered: modularity, functional design, design for assembly (atthe early stages of the design process) and simulation (accelerated through theparameterisation of the mesh). Usually the domain of modularity is often associated withfunctional design or with the design for assembly, but rarely the three together, thatconstitutes one of our added values. Finally, the parametric aspect of the FARDmethodology, that is the link between the four domains, allows accelerated the generation ofproducts of the same family from a generic product and thus saving design time to achieveour goal of "High Productive" routine design. Three industrial and academic case studiesillustrate the application and the feasibility of the FARD methodology...