Modeling Methods Of Cement Grinding Production Process

The production process of cement mainly includes three steps of raw material preparation,clinker firing and cement grinding.The cement grinding production process,as the last stage of the cement production,not only directly determines the quality of finished cement product,but also is the most energy-consuming stage in the entire cement production process.Accurate modeling of the cement grinding production process is helpful to controlling the cement production process more effectively.This thesis studies the modeling method of the cement grinding production process based on the actual production data of a cement grinding system.The main work of this thesis includes:(1)Analyzing the production of cement grinding system.A full understanding of the operation principle of the equipment and familiarity with the grinding process are very important to modeling the production process.With the progress of technology,the cement grinding process of different cement plants are different.This thesis focuses on the analysis of the cement grinding process of a cement plant in Tai’an City,Shandong Province.(2)Analyzing the key parameters of the cement grinding system.There are a total of 44 control variables and controlled variables related to the cement grinding process.When constructing the model,if all variables are included,an extremely complex model will be obtained.Therefore,before constructing the model,it is necessary to analyze the core variables of the system and determine the main input and output parameters.The most important controlled variables of the cement grinding system are the amount of circulating materials,the current of the mill bucket,the differential pressure of the mill,and the power/current of the mill.The control variables include the total feeding amount,grinding pressure,and amount of water,powder separator speed,powder separator current,exhaust fan speed,exhaust fan current,valve opening,mill vibration,material layer thickness and mill air temperature.(3)Dividing the cement grinding into several subsystems.The equipment and parameters involved in the cement grinding system are too complicated.Considering all of them in one system will greatly increase the difficulty of modeling.Therefore,the cement grinding system is divided into multiple relatively independent subsystems before the modeling work is carried out.The modeling of the subsystems is constructed separately.Subsystem division is carried out from the perspective of process mechanism and data analysis at the same time,the models are established respectively,and finally the accuracy of the models is compared.(4)Modeling the cement grinding production process.In this thesis,we first establish a model of the cement grinding system based on the ARIMA model and the NARX model,and then establish a neural network model for the four divided subsystems based on the long and short-term memory network model,and realizes the amount of circulating materials and the current of the mill bucket.This thesis realizes single-step and multi-step predictions of the four controlled variables of circulating material volume,current of the grinding bucket,grinding differential pressure and mill power/current.Among them,the error of the prediction result of the long and shortterm memory neural network model is in the acceptable range.The prediction result of the ARIMA model shows that the model is only suitable for single-variable short-term forecasting,and is not suitable for the cement grinding system,which is a complex system with many input variables.In the process of modeling,due to too many input variables related to the power/current of mill,the principal component analysis(PCA)is performed on the input variables during the data processing,and the input variables are analyzed according to the magnitude of the principal component contribution rate.The above work reduces the dimensionality of the input data in the neural network model and reduces the complexity of model training.(5)Development of model identification software for cement grinding system.In order to realize the visualization of the model identification work,in this thesis,we simulate the online identification process of the cement grinding system and develop a software for modeling the cement grinding process in the MATLAB environment.The design and development of this software is discussed in details.