We propose modelling tools for computer aided supervision of continuous processes. Our modelling approach aims at including some explanation capabilities in a qualitative modelling of processes behaviour to advise operators. The model is declarative and is represented as a causal graph in which the nodes are the most relevant supervision variables and the arcs model the causal links between these variables. The arcs allow to compute the behaviour of each variable through behaviour constraints named Qualitative Transfer Functions (QTFs). QTFS are approximations of some simple transfer function response curves to standard inputs. The a priori well known shapes of these behaviours are parametrized with some data such as gain, delay and time response. We propose a QTFs library which allows to model simple representations between two variables and more complex systems such as control loops and balanced variables. The qualitative model is used by a supervision system to compute behaviour simulation, truth maintenance on deductions and on the model structure and to provide action advices to operators.
