Theses of the Industrial Engineering Center

Enterprises are now operating in an environment where market is more open, globalised, and competitive. Due to this new trend of market, enterprises have started involving in many industrial networks in order to maintain their business under such pressures.

The integration of business partners deeply depends on the ability to capture and share information seamlessly among information systems (ISs) of different enterprises. The MISE (Mediation Information System Engineering) project has been evolved in order to tackle this problem by providing an information technology solution for supporting the enterprise interoperability through ISs. It is developed on the basis of the MDE (Model Driven Engineering). This dissertation addresses the business level of the interoperability, and the CIM (Computer Independent Model) of the MDE. Its main objective is to develop a knowledge-based system for supporting the design of collaborative processes that conform to the BPMN (Business Process Modeling Notation).

We propose to work at the higher abstraction level than the CIM for capturing knowledge which allows to characterise collaboration by basing on the perspectives and experiences of business partners. We use the existing knowledge (instances about business processes) from the MIT Process Handbook for moving down to the CIM level. The prototype of our knowledge-based system is also developed in order to validate and evaluate the approach.

Our objective is to optimize the Order Management Process performance in case of stock-out situation. To reach this goal, we propose a method and a tool to integrate Supply Chain and Demand Chain.

Concretely, the thesis proposes a model known as “Advanced Availbale To Promise” for decision-making with the order management in the event of stock-out. This multicriterion decision-making tool makes possible to take into account the different points of view from the actors as well as the possibilities of flexibility to fulfill the orders (substitution, partial delivery, multi site). The parameters setting of the tool makes it possible to implement various strategies to manage the shortage and fulfill the orders: efficient, responsive, effective, agile and leagile..

In order to measure and to evaluate the performance of the results, the thesis also proposes a three-dimensional framework.

Information systems for pharmaceutical industry are now obliged to respect regulatory requirements imposed by this branch of industry. These requirements are listed in documents, as for example the GxP for Europe. In a concomitant way, organization can have deployed software process improvement methods, based on good practices repositories. The Capability Maturity Model Integration (CMMI) is one of them. It is often used in the area of information systems. These two types of documents, coming from sectorial expertise for the first one and from business expertise for the second one, are overlapping. To be efficient, their deployment can not be done without analysing this overlapping and needs the realization of a policy stemming from the two documentary sources for quality assurance. The aim of our researches is to measure the degree of correspondences between these two quality repositories.

To bring an answer to this question, we apply a model driven engineering method. We realize a mapping between models of those two categories of documents. Its aim is to put into relation comparable knowledge. Then, we propose a method to build the target repository, based on logical operators use in order to identify a minimal set of good practices required by pharmaceutical regulation. The approach is deployed on an industrial study case composed by 175 regulatory requirements and more than 800 practices, coming from CMMI but also from other repositories (ISO 9000, ITIL...).

In a collaborative context, the integration of industrial partners deeply depends of the ability of their Information Systems (IS) to interact efficiently. In this document we propose to tackle this point according to the point of view of IS interoperability. Interoperability of partners’ IS is based on two main aspects: (i) the fact that partners’ IS respect SOA (Service-Oriented Approach) concepts and (ii) the support of a Mediation Information System (MIS) able to put IS together in one single merging System of Systems (SoS).

We propose to design such a MIS according to MDA (model-Driven Approach) principles. We aim at using business model (the needs) to design a logical model of a solution (logical architecture). The business model is a collaborative business model (in BPMN, at the CIM level), while the logical model is a MIS model (using UML, at the PIM level). This document presents the theoretical aspects of this subject, the mechanisms of transformation and the dedicated translation rules. Finally, we show the prototype of a demonstration tool embedding the transformation rules and running those principles.

In a rapidly changing economic and industrial context, this thesis presents a cooperation support approach for the distributed planning of capacities within a supply chain. We propose a simulation tool allowing the evaluation of planning behaviors of partners acting in more or less cooperation for various scenarios of market behavior.

Applying decision and game theories, we then introduce a risk evaluation of each supply chain partner as well as the supply chain as a whole. Its objective is to direct the choice of the decision makers as regards planning policies using a co-operative risk management. An industrial application, in the electronics and telecommunications sector, provides an implementation of our proposals.

The research work presented in this thesis deals with assistance to design and configuration tasks. An integration of different existing concepts of constraint programming has been achieved. This integration has been tested on an implementation based upon syntaxic trees. The syntaxic trees allow to express different kinds of CSP (Constraint Satisfaction Problem) which model design or configuration problems.

The first part presents the fields of design and configuration, and aims at identifying the needs for decision aid: different kinds of parameters (discrete and continuous), hierarchical organisation and optionnal elements. Different constraint-based approaches which may fulfill any need are then detailed.

The second part presents the RCSP (Relevancy CSP), which are an integration of different CSP from the literature seen in the first part. Some recommendations for modeling design or configuration problems are set up. The implementation is then presented, on the one hand for CSP processing and on the other hand for RCSP processing.

Our research deals with risk identification and evaluation in information system projects. We have elaborated a frame based on the different sources of risks in IS projects. This frame is included in the methodology we propose for the interviews and for the analysis of the information we gather in order to identify risks in such projects.

We have elaborated it from Social Science and Engineering, demonstrating, in doing so, their complementarity.

Thanks to this methodology, we were able to analyse real IS projects. Through the methodology of risk evaluation, that we propose in our research, we seek to rank the risks in order of their priority to determine their acceptability.

Our methodology of risk evaluation is based on the Analytical Hierarchy Process (AHP). We have paid detailed attention to the way in which the results of risks evaluation are rendered.

This work deals with SMEs networks: a set of relatively small companies, that decide to share some objectives. The study is limited to SMEs production networks which manage their collaborative work as a shared project.: a complete whole of activities and operations which must be carrieds by the network in a definite time. Each company preserves individual activities out of the collective project.

Thus, if classical project management theory and practice could certainly be applied to the shared project, by definition, sources of uncertainty are always more difficult to handle in such organizations, and risk management is emphasized by the need to react. Reactivity has to be rapidly included in the planning phase of project design. That is the reason why integration of risk to project management appears to this work as a key factor to conceive the management framework.

Unfortunately, if management levels of networks have some similarities with the ones of a single enterprise, coordination of collective activities in a network at least leads to consider some adaptations and recommendations.

The goal of this work is to search the main features of a network management framework based on project and risk management.

The framework is realized from an empirical study in a twelve SMEs manufacturing network from the aeronautical sector, located in Toulouse, France. The study consisted in analyzing project documentation and developing interviews with project managers. The proposition is based on the results of this study.

In this PhD thesis, we deal with two axes of industrial industrial, implemented on the same industrial case of study.

First of all, we focus on the analysis of industrial problems, using previous work concerning both enterprise modeling and continuous improvement. We propose an analysis/action grid based on enterprise modeling in order to identify the problems of the system and improvement tools to use. An application is presented based on our industrial problem.

Furthermore, a new scheduling problem results from the previous application: a shop with parallel flexible machines requiring secondary ressources (tools and operator interventions), with set-up but without precedence constraints between operations of a job. We will study the total job tardiness criterion. Two approaches are proposed to solve this problem: by heuristic and by simulated annealing. For the heuristic approach, a new rule called ATCTRS has been developped. Its aim is to make a compromise between the tardiness of an operation and the occupation of the machine. For the meta-heuristic approach, we mainly study a neighborhood structure fitting the problem to solve.

The work presented in this thesis deals with interactive aiding design of process. The design support tool is based on a knowledge model described as a Constraint Satisfaction Problem (CSP). The solutions search being interactive, the tool makes use of CSP filtering methods. The range of knowledge used leads us to integrate different CSP types (discrete, continuous, mixed, dynamic) as well as their filtering methods (arc-consistency, 2B-consistency, discretization). This work is based on an industrial problem relevant to the design of heat treatment operations.

The first part gives the backgroound of our works and preents the industrial probllem. In a second part, a typology of the constraints necesary to build knowledge models is made. This typology allows usto identify the different CSP types to use. Each one is presented in terms of state of art, needs and adopted solutions. Taking into account 2D abacus in constraints-based models needs the use of a specific data structure named quad tree. The integration of this data structure does not raise any particular problem for the abacus described as a single numerical constraint. We extend it in order to take into account abacus described as piecewise numerical constraints. The last part presents the general architecture of a constraint-based model built with heat treatment experts and how the filtering engines are gathered.

This research work proposes a formal framework to manage the project of design and exploitation of an ERP in SMEs/SMIs. The study covers two scientific disciplines : industrial engineering and software engineering.

Stakes and risks in the implementation of an information system are substantial. In order to improve the quality of projects in terms of cost, time and performance, we contribute to the progress of knowledge with a set of new elements which facilitate project design and management. We propose a method of engineering of these projects based on the use of enterprise and reference models.

First, we work on the use of a single formalism, the UML language (Unified Modeling Language), to model the company orgnization and software functionalities. The UML language makes it possible to represent the physical, logical or functional points of view of these systems.Moreover, the UML language allows the user to have various levels of granularity with the representation, with respect the user's expectations.

Second, we deal with the definition of a formal framework for ERP project management. Within this reference framework, we try to allocate enterprise and reference models. We defined mechanisms to modify the models and to integrate the in order to obtain the right parameter setting for the software solution. The customization corresponds to the right adequacy between the organization and the information system. A case study is used in order to demonstrate the relevance of our approach for the last two parts of the thesis.