Development of a tool for automatic generation of uncertain systems behavioral models

The dynamic landscape of engineering and technology continuously demands innovative solutions to complex problems. As companies strive for efficiency, cost-reduction and sustainability, automation and uncertainty modelling are becoming pivotal matters in the industrial world. This thesis presents the results of a six-months internship in the Alten Rennes Innovation Department. The primary objective of this internship was finding innovative solutions for the management of requirements in systems engineering. Numerous studies have been conducted on systems failure causes. D Laurie Hughes et al. propose an analysis of phenomenon with the goal of identifying the trends and the main weaknesses of the current approach. In this work, they analyse case studies, theoretical studies, empirical studies, and other documents that aim at proposing an interpretation of data as well, for a total of more than 60 documents. Such a vast bibliography comes with significant differences even in base concepts like the definition of “project failure”, “project success”, and in failure classification. Despite this complex scenario, they identify some factors that show up frequently, regardless of the theoretical framework. Requirements management, together with change management are referred to as two of the most frequent failure causes in system engineering. Some complementary information can be found in Nasa’s report on the 14th Annual International Symposium (2004) conference proceedings. The document proposes an analysis of the error-fixing cost vs. project phase trend. Nasa indicates five main project phases: requirements, design, manufacturing, integration/test, operations. Three different methods are proposed to estimate the error-fixing escalation as the project advances. While the three methods are based on different philosophies, they reveal the same qualitative behaviour. Normalising the data with the requirement phase error-fixing cost, the following phases always have positive cost factors. In other words, costs rise according to all models. Moreover, the increase is significantly steep. The most favourable method evaluates the test phase at least 21 times more expensive than the requirement one, going up to 78 times according to the most pessimistic one. These studies present a scenario with clear issues that Alten wants to address. Establishing a solid, automated procedure to deal with requirements would drastically reduce mistakes in the early stages of a project, which would mean optimising time, resources, and money. At the same time, uncertainty modelling results nowadays mandatory to build reliable requirement models. Achieving such a process would allow to dedicate more time, money, and energies to system engineering aspects that have been apparently overlooked up to now: change and requirements management.