The constant evolution of the environment of a system leads to detect and interpret all the various events susceptible to modify the performances of the system, i.e. its functionality, structure and / or behaviour, particularly during its exploitation phase. To cope with such a problem, this paper focuses on the appropriateness of production system multi-modelling (multilevel and multi-view modelling) with the context of use of the system, and its evaluation. It proposes an approach which provides analysts with the appropriate view of a production system, for control assessment. This approach is based on the comparison between the Performance Indicators – System (performance evaluated a posteriori when the measurement comes from the real system) and the Performance Indicators – Model (performance evaluated a priori when the measurement translates the behavior of a model of the system).