Research And Development Of Process Fault Diagnosis System For Cement Clinker Preparation Process Based On Machine Learning

The cement industry is an important basic raw material industry for the development of China’s national economy and an important support for the improvement of human living conditions and the development of a circular economy.Cement clinker preparation is the core and key link in the cement production process,and its process is accompanied by complex physical and chemical reactions.If faults occur in the cement clinker preparation process,it not only affects the quality of cement but also causes huge economic losses to the whole cement production.Troubleshooting cement clinker preparation is of great significance to improve the quality of cement production and enhance the economic efficiency of enterprises.This essay focuses on the diagnosis of process faults in cement clinker preparation processes as follows:Firstly,for the situation that the data set corresponding to the process fault in the cement clinker preparation process is noisy or abnormal,etc.,a data pre-processing method of multirule fusion is proposed.Firstly,for the problem of the singularity of existing outlier rejection criteria,a multi-rule fusion data outlier rejection method is given to ensure the accuracy of the subsequent fault diagnosis data;then,based on data outlier rejection,the data is further filtered to eliminate the influence of random interference caused by environmental factors in the cement production process;finally,the normalization operation is performed on the filtered data to ensure The accuracy of the input data of the fault diagnosis model is ensured.Secondly,for the problem of cement clinker preparation process fault diagnosis,a cement clinker process fault diagnosis method integrating XGBoost and Bayesian optimization is proposed.First,the feature selection method of random forest is used to analyze the correlation between the process fault of the cement clinker preparation process and the key parameters of the production process;then,based on the correlation analysis results,an XGBoost fault diagnosis model with the input as the key parameters and the output as the process fault is established,and the hyperparameters of the fault model are optimized by using Bayesian optimization method;finally,based on the real-time collected key parameter state Finally,the fault diagnosis and identification are performed based on the real-time collected key parameter state data.Compared with the existing fault diagnosis methods,the fault diagnosis accuracy of the proposed method reaches 97.73%,which verifies the effectiveness of the proposed method.Thirdly,aiming at the problem of real-time diagnosis and identification of cement clinker preparation process faults,a cement clinker preparation process fault diagnosis system is developed.Firstly,the overall architecture and functional architecture of the cement clinker preparation process fault diagnosis system is given around the actual demand of cement production,which contains five functional modules of data acquisition,data pre-processing,fault diagnosis,fault monitoring,and data management;then,the framework of B/S development mode and C# and Python as the main programming languages are used to complete the cement clinker preparation process fault diagnosis system The development work was completed with B/S development framework and C# and Python as the main programming language.This system has practical application value for guaranteeing stable cement production.