Research On Temperature Prediction Method Of Rotary Kiln Based On Nonlinear Dynamic Time Delay Analysis And Machine Learning

Over the past few decades,China has vigorously carried out infrastructure construction,and cement manufacturing industry has entered the golden stage of development.However,with the continuous promotion of China’s supply-side structural reform and “carbon peak,carbon neutralization” goal in recent years,a large number of cement manufacturing plants are facing the urgent task of intelligent transformation and upgrading.At present,China’s cement production enterprises generally have the problems of lack of high-quality products and high production energy consumption.As the core equipment of cement production industry,cement rotary kiln calcines cement raw materials at high temperature,which has a key impact on the production process.The temperature in the kiln is the main index to monitor the running state of the rotary kiln.How to get the stable and accurate temperature in the kiln is of great significance to the stable running of the rotary kiln,so as to improve the product quality,reduce equipment failure and reduce energy consumption.In actual production,physical monitoring equipment is easy to produce fluctuation,abnormality and even failure in the harsh and complex environment inside the rotary kiln,so it can not monitor the temperature of the rotary kiln in real time and stably.The production process lacks a short-term prediction model to predict the future temperature trend,and the temperature anomaly early warning cannot be carried out.In order to solve the above problems,this thesis constructs a data-driven real-time temperature prediction model of cement rotary kiln through data mining algorithm and machine learning algorithm,so as to realize timely and accurate temperature prediction in the kiln.Using the characteristics and advantages of the data model,a multi-step prediction model is established to realize the multi-step prediction of short-term temperature in the future.On this basis,the intelligent temperature prediction and early warning system of rotary kiln is further designed and realized.The specific research contents are as follows:（1）The process flow of cement rotary kiln is introduced in detail,which lays the field background knowledge for temperature modeling.In the first place,the whole cement production process is briefly introduced,and the process of each subsystem is briefly described.Secondly,the most important firing system is summarized,and the process flow of the main working equipment of the system is introduced.Futhermore,the process of the most important equipment-cement rotary kiln is introduced in detail,which reveals the importance of obtaining the temperature in the kiln.Finally,through the above process analysis,the factors affecting the temperature of cement rotary kiln are found out,and three typical factors are analyzed in detail.It shows that each factor is coupled with each other,and the system has large time delay and nonlinearity,which lays a solid foundation for subsequent feature selection and model establishment.（2）Considering the nonlinear and large time-delay characteristics of rotary kiln,the temperature of rotary kiln is modeled and predicted based on the Extreme Learning Machine（ELM）algorithm and the dynamic time-delay analysis method.First of all,a feature dimension reduction method based on static time delay is proposed,and the Minimum Redundancy Maximum Relevance（m RMR）feature selection is carried out based on the feature space reconstructed by static time delay.Moreover,considering the complex disturbance of rotary kiln and the dynamic change of time delay,based on the traditional dynamic time delay analysis method for linear characteristics,a nonlinear dynamic time delay analysis method based on mutual information（MIDTA）is proposed,and the effectiveness of the algorithm is verified by numerical simulation experiments.Last,based on the above research,a prediction model based on MIDTA-ELM is established.The model includes data preprocessing process and an online dynamic time delay estimation method.The model is applied to the temperature prediction test of rotary kiln,and the experimental results show that the accuracy of the model is significantly improved compared with the unmodified model.（3）In order to realize the short-term multi-step prediction of rotary kiln temperature,the Hierarchical Extreme Learning Machine（HELM）is improved by Sparrow Search Algorithm（SSA）,and the rotary kiln temperature prediction model is established combined with the previously proposed dynamic time-delay analysis method.Firstly,the principle of the Extreme Learning Machine as Autoencoder is introduced.On this basis,the mathematical principle and training process of HELM are introduced.Secondly,the optimization principle and search strategy of SSA are introduced.The number of hidden layer nodes of HELM is optimized by SSA,and the optimization framework of HELM by SSA is established.Finally,combined with the previously proposed MIDTA algorithm,the MIDTA-SSA-HELM prediction model is constructed.Based on the strong fitting ability of HELM,the single-step prediction and multi-step prediction tests of rotary kiln temperature are carried out.The test results show that compared with the unmodified model and some traditional prediction models,the model has higher prediction accuracy and is suitable for the future multi-step prediction of rotary kiln temperature.（4）Combined with the algorithm proposed in the above research as the theoretical basis,the temperature prediction and early warning system of cement rotary kiln is designed and developed by using MATLAB software.To begin with,the system requirements are analyzed in detail,and the main functions of the system are obtained.The next,the overall framework of the system is designed,and each functional module is designed in detail according to the algorithm model proposed in this thesis.In the end,MATLAB is used to develop and deploy the system,realize the accurate calculation and visualization of rotary kiln dynamic time delay,real-time temperature and multi-step temperature,and realize the real-time monitoring and early warning of temperature and related variables.The system realizes the multi-functional prediction and early warning of the temperature of cement rotary kiln,and can provide effective help for operators.