Process knowledge acquisition and control by quantitative and qualitative complementarity

Abstract

Even when an autoregressive model[1,2] is created and computer control is effected based on it, the subsequent measured values are sometimes imperfect due to disturbances in the process and noise in the measurements. This paper proposes an approach for overcoming this drawback of tight control by an AR model when it is impossible to carry out computer online control based on an autoregressive model. This approach in the broad sense of the term involves robust control in which model-based deep knowledge based on an existing AR model or mathematical model is used and converted to fuzzy qualitative oontrol. As an actual example we discuss a cement rotary kiln process, and we present an approach for process disturbances and incomplete measured values by transforming quantitative control into qualitative control and also making use of hidden information that cannot be abstracted without sensor fusion. As a feature of this method we discuss the effectiveness and purpose of the paradigm in which one does not quantify a qualitative model but rather goes in the opposite direction of qualitizing a quantitative model.