Research Of Predictive Control Algorithm On The Cement Rotary Kiln Based On Fuzzy ARX Model

Cement rotary kiln is the core equipment of the NSP cement production line, themost critical calcinations’ physical and chemical reactions of the raw material in thecement production is conducting in the kiln; the control level of the kiln has a directimpact on the production, quality and energy consumption. The research on modeling andcontrol method for the cement rotary kiln has important theoretical and practicalsignificance to improve the control level of the cement rotary kiln. For that thecalcinations’ process of the NSP cement rotary kiln has the characteristics of nonlinear,multivariable and multi-delay, on the basis of analysis process to the sample data, thispaper uses the combined modeling theory and the predictive control theory to study thepredictive control method of the cement rotary kiln based on fuzzy ARX model, specificresearch work are shown as follows:First of all, according to the calcined mechanism of the cement rotary kiln, thecontrol factors that affect the quality and energy consumption of the cement clinker isanalyzed, the variation between the parameters are studied. The correlation functionmethod is used to analyze the test data of control parameters, and then the modelingvariables of the rotary kiln system are determined. The nitrogen oxides content in thesmoke room that is at the end of rotary kiln, oxygen content and temperature at the end ofthe rotary kiln are set as the control targets, and the amount of feeding coal at the head ofthe kiln, the baffle opening of high temperature fan and feeding speed area are set as theoperating variables. All these works provide the basis for the establishment of theprediction model for the cement rotary kiln calcinations’ process.Secondly, according to the combined modeling idea based on the separationidentification of nonlinear characteristic and dynamic linear characteristic in systemsteady state, the real gain correction method of ARX model based on T-S fuzzy model isproposed and the variant gain fuzzy ARX suiting for the multi-variable nonlinear systemis worked out. On basis of data driver thought, this paper puts forward a data driven fuzzyidentification improved algorithm and establishes cement rotary kiln calcining system steady model under the condition that the system test data can not cover all input variableconstraint range. Use the push difference method to calculate the actual system dynamicgain for real-time correction of multi-delay ARX model gain and improve the accuracy ofthe model, and then put forward the variable gain fuzzy ARX model method formulti-delay multivariable nonlinear system in order to solve the modeling problem ofmuch delay characteristics in the rotary kiln. Based on the separately preprocessing ofsteady state and dynamic identification data in the cement rotary kiln, the control systemprediction model of cement rotary kiln calcining process is established.Furthermore, the sequential quadratic programming optimization algorithm isutilized to dynamically solve the input change of multivariable nonlinear system and apredictive control algorithm based on multivariable fuzzy ARX model is developed. Onthis basis, the coefficient matrix interpolation combination method is proposed, and themulti-delay is added into the multi-step prediction output expression of the multivariablenonlinear system. And then a predictive control algorithm of the multi-delay multivariablefuzzy ARX model is designed so as to realize the multi-time delay control of rotary kilnsintering process, and then the prediction control algorithm for the cement rotary kilncalcine system was designed.Finally, using the DCS system to collect the test data of cement rotary kiln, and thedata is modeled and analyzed. According to the proposed multi-time delay variable gainfuzzy ARX model and its prediction algorithm of cement rotary kiln in this paper, thepredictive control experiment of cement rotary kiln is performed. The experimentalresults can reveal that the feasibility and validity of the proposed algorithms in this paper.