A causal reasoning for the simulation of continuous industrial processes

This report describes an on-line simulation tool to be integrated in a supervision support system for industrial continuous processes. The aim is to provide operators with the future behaviour of the process after significant modifications have been detected on some inputs or on measurable disturbances. A nuclear waste processing plant is used to illustrate the method: the process is modeled by a causal graph, whose nodes are the variables relevant for the operators, and the arcs the cause-effect relationships between them. Each of the arcs support a qualitative transfer function (QTF), parameterized by a delay, a static gain and a settling time. This model is the knowledge base used by the simulator. The evolution of a variable is represented by a piecewise linear function. The simulation algorithm aims to propagate the evolutions from a variable into another one in the graph thanks to the QTFs. It leads to the concept of event, a basic function constituted with a step and a ramp. 38 fig., 6 ref