Declarative Modelling for Process Supervision

Abstract Our work is a contribution to computer aided supervision of continuous processes. It is inspired by an area of Artificial Intelligence : qualitative physics. Here, supervision is based on a model which continuously provides operators with a synthetic view of the process; but this model is founded on general principles of control theory rather than on physics. It involves concepts such as “high gain” or “small time response”. It helps in linking temporally the evolution of various variables. Moreover, the model provides predictions of the future behaviour of the process, which allows action advice and alarm filtering. This should greatly reduce the famous “cognitive overload” associated to any complex and dangerous evolution of the process.