Modéliser et analyser les risques de propagations dans les projets complexes : application au développement de nouveaux véhicules

摘要

The management of complex projects requires orchestrating the cooperation of hundreds of individuals from various companies, professions and backgrounds, working on thousands of activities, deliverables, and risks. As well, these numerous project elements are more and more interconnected, and no decision or action is independent. This growing complexity is one of the greatest challenges of project management and one of the causes for project failure in terms of cost overruns and time delays. For instance, in the automotive industry, increasing market orientation and growing complexity of automotive product has changed the management structure of the vehicle development projects from a hierarchical to a networked structure, including the manufacturer but also numerous suppliers. Dependencies between project elements increase risks, since problems in one element may propagate to other directly or indirectly dependent elements. Complexity generates a number of phenomena, positive or negative, isolated or in chains, local or global, that will more or less interfere with the convergence of the project towards its goals. The thesis aim is thus to reduce the risks associated with the complexity of the vehicle development projects by increasing the understanding of this complexity and the coordination of project actors. To do so, a first research question is to prioritize actions to mitigate complexity-related risks. Then, a second research question is to propose a way to organize and coordinate actors in order to cope efficiently with the previously identified complexity-related phenomena.The first question will be addressed by modeling project complexity and by analyzing complexity-related phenomena within the project, at two levels. First, a high-level factor-based descriptive modeling is proposed. It permits to measure and prioritize project areas where complexity may have the most impact. Second, a low-level graph-based modeling is proposed, based on the finer modeling of project elements and interdependencies. Contributions have been made on the complete modeling process, including the automation of some data-gathering steps, in order to increase performance and decrease effort and error risk. These two models can be used consequently; a first high-level measure can permit to focus on some areas of the project, where the low-level modeling will be applied, with a gain of global efficiency and impact. Based on these models, some contributions are made to anticipate potential behavior of the project. Topological and propagation analyses are proposed to detect and prioritize critical elements and critical interdependencies, while enlarging the sense of the polysemous word “critical."The second research question will be addressed by introducing a clustering methodology to propose groups of actors in new product development projects, especially for the actors involved in many deliverable-related interdependencies in different phases of the project life cycle. This permits to increase coordination between interdependent actors who are not always formally connected via the hierarchical structure of the project organization. This allows the project organization to be actually closer to what a networked structure should be. The automotive-based industrial application has shown promising results for the contributions to both research questions. Finally, the proposed methodology is discussed in terms of genericity and seems to be applicable to a wide set of complex projects for decision support.