Real tool for understanding the causes and consequences of damaging or catastrophic events, the models of accident took place at the heart of critical approaches to accident investigations, and / or risk assessments within the socio-technical systems. Their effectiveness has been proven and many organizations are using them for answering a need for safety and performance requirements and progress. Many models of accident have been developed. Originally based on the principle of representation of a chain of events linked by cause and effect relationships, contemporary models are now based, for a minority even on systems theory and control theory, thus promoting systems analysis and a holistic view of accidents and disasters. Among latest modeling approaches, systemic accident model called STAMP (Systems-Theoretic Accident Modeling and Processes), developed at the Massachusetts Institute of Technology, and based on the concepts of "constraint", "structure" and "Model of process" was chosen in the framework of this thesis. This model assumes that accidents result from inadequate and ineffective control of the structure of a socio-technical system leading them to an accidental state. The thesis has three objectives. The first is to capture the epistemological, theoretical and methodological fundamentals underlying the STAMP. The second is the effective implementation of the model in the context of risk assessment of an innovative industrial process to clean up contaminated sediments. The third aims to assess the contributions and limitations of the model and propose with regard to the limits identified proposals to improve both the theoretical and methodological aspects.