Operating mode recognition. Application to a grinding mill process.

Process monitoring needs the development of data analysis tools aiming at recognizing, at each time instant, system operating mode using the measurement collected on the system. This communication aims at presenting a method relying on measurement analysis, able to identify operating modes without the knowledge of the mathematical models describing these modes. The proposed method relies on the writing of a global model combining, in a multiplicative way, the models describing the different modes. The parameters of this global model are then numerically identified from the available set of measurements. The sensitivity analysis of the global model with regard the input/output variables then provides an indicator to identify, at each instant, the current operating mode. The proposed method is applied on a simplified model of a grinding mill.