| Cement clinker calcination as the core of the cement production process,its operating state directly affects the clinker quality and energy consumption.At present,cement enterprises mainly use"manual fire watching"and supplemented by process variables to identify the state of the clinker calcination process,and then knowledge workers set the value of the key variables of the control loop to complete the control of the cement clinker calcination process.As the cement clinker calcination process involves complex physicochemical reactions,and the process is characterized by complex and variable states and key parameters that cannot be detected online,the results of state identification,free calcium oxide(f-Ca O)content and production energy consumption in manual operation mode are subject to human subjective factors,making it difficult to ensure stable and efficient operation of the cement clinker calcination process.Aiming at the problems of difficult state identification of cement clinker calcination process,unstable f-Ca O content and high energy consumption in production,relevant research work is carried out around the state identification and optimization methods of cement clinker calcination process,and the main research contents are as follows.(1)Research on the state identification method of cement clinker calcination process based on steady-state detection and flame image features.A steady-state detection method based on trend feature extraction is proposed for the characteristics of cement clinker calcination process with alternating dynamic and steady-state operation.Combining empirical modal decomposition,sample entropy and least-squares fitting methods,the trend features of the state variables are extracted,and the information amount characterizing the trend changes is used to construct the steady-state test index to determine the steady-state of the cement clinker calcination process.On the basis of steady-state detection,the method of cement clinker calcination state identification based on flame image features is studied.An Otsu-Kmeans-based image segmentation method is proposed to segment the target regions of flame images,which can effectively solve the problems of blurred boundaries and mutual overlap between target regions.The support vector machine(SVM)state recognition model is established by extracting the target region features of flame images.The experimental results show that the SVM model based on the flame image features achieves an accuracy of 93.33%for the clinker calcination state recognition based on the steady-state detection of the cement clinker calcination process achieved by the steady-state detection method proposed in this paper.Under the same experimental conditions,the state recognition accuracy of probabilistic neural network(PNN)model and back propagation neural network(BPNN)model is 83.3%and 80%,respectively,and the model computation time is longer.The state recognition scheme in this paper provides an effective solution to realize the transformation from"manual fire watching"to"machine fire watching"in cement enterprises.(2)Research on f-Ca O content prediction model based on multi-model fusion strategy.A multi-model fusion modeling strategy based on K-Means++clustering and multi-kernel relevance vector machine(MKRVM)is proposed to construct the f-Ca O content prediction model in view of the problems that f-Ca O content is difficult to be detected online and the prediction performance of a single model is not satisfactory under complex multi-conditions.The K-Means++clustering method was used to divide the raw material triple rate value,raw material fineness and raw material feeding amount into sample working conditions and determine the training sample subsets,which effectively fused the data characteristics under different raw material sample working conditions.On this basis,the MKRVM-based f-Ca O content prediction model was established,and the final f-Ca O content prediction values are obtained by the weighted fusion of multiple models.The multi-model fusion modeling strategy and the MKRVM algorithm were experimentally validated,respectively.The results show that compared with the modeling approach using a single model(Single-MKRVM),its mean absolute error(MAE)is reduced by 25%,root mean square error(RMSE)by 18%,thayer inequality coefficient(TIC)by 27%,and coefficient of determination(R~2)by 29%.Compared with the single kernel function RVM model(Multi-RVM)and SVM model(Multi-SVM),the MAE is reduced by 8%and 12%,the RMSE is reduced by 10%and 16%,the TIC is reduced by 17%and 24%,and the R~2is improved by 21%and 26%,respectively.The problem of difficult online detection of f-Ca O content is solved,and the prediction accuracy and generalization ability of the model under complex multi-conditions are improved,providing a model basis for the study of cement clinker calcination process optimization problems.(3)Research on the intelligent optimization method of cement clinker calcination process.Aiming at the problems of unstable clinker quality and high energy consumption in cement clinker calcination process,an intelligent optimization method of cement clinker calcination process based on the improved Gray Wolf algorithm(Multi-GWO)is proposed.Based on process analysis of the data relationship between f-Ca O content and energy consumption,the f-Ca O content can reflect both clinker quality and energy consumption.The multi-objective optimization problem of improving f-Ca O content qualification rate and reducing energy consumption can be regarded as a single-objective problem with f-Ca O content as the optimization objective.Based on the f-Ca O content prediction model,the optimization objective function is established with f-Ca O content as the optimization objective,and the Multi-GWO intelligent optimization method is used to solve the problem and obtain the optimal value of f-Ca O content to meet the current working conditions,as well as the corresponding set values of rotary kiln clinker calcination temperature and decomposer outlet temperature.The simulation results show that the proposed Multi-GWO optimization method is better than the standard GWO algorithm in terms of finding accuracy and stability,and the f-Ca O content qualification rate can be improved by 8.34%,and the f-Ca O value is closer to the upper limit value of 1.5%in the qualified range of process requirements,which provides effective guidance for improving the f-Ca O content qualification rate and reducing energy consumption.(4)The software of cement clinker calcination process state identification and optimization guidance system was designed and applied in the field based on multimodal humanoid intelligent controller.The application results show that compared with manual operation,the clinker quality index f-Ca O content qualification rate was increased by 12.5%and the coal consumption was reduced by 1.5%,which can realize the production demand of cement enterprises to improve f-Ca O content qualification rate and reduce coal consumption,and has good engineering application value. |
PDF Full Text Request