This thesis is part of the European project PAPYRUS (7th FWP (Seventh Framework Program) and it concern the developments of models and tools for the analysis of an industrial process in interaction with system performance indicators. Thus, the developments of Plug \& play algorithms for fault diagnosis. More specifically, the first objective of the thesis is to propose models and criteria, which allow, for large complex systems, whether the objectives expressed in terms of performance (cost, dependability, etc.) are achievable. Within the causality modeling system, a transfer entropy based method is proposed to identify the causality model of a system from data. We also focused on the influence of different faults on system performance reachability. The tools used are mainly based on graphical approach analysis in parallel with statistical tools. The second objective concerns the implementation of algorithms for faults diagnosis. A hierarchical fault diagnosis process based on causality model of the system is implemented. This step also allows the evaluation of the system performance. We applied our methods on the PAPYRUS project plant (board machine Stora Enso IMATRA in Finland).
