Operating Mode Recognition With Multiple Time Scales And Intelligent Control Of Thermal State Parameters For Sintering Process

Iron ore sintering is an essential raw material preparation process in the iron and steel industry.The produced sinter ore is the primary raw material for blast furnace ironmaking.The quality and yield of sinter ore directly affect the cost and benefit of iron and steel production.The primary chemical reaction in the sintering process is the combustion reaction of the solid fuel in the material layer.The sintering thermal state is a comprehensive reflection of the combustion of the material layer.The operating mode is a direct presentation of the sintering thermal state.The operator will make different control decisions for different operating modes to meet production requirements.The ignition temperature and the burn-through point are the two critical parameters of the sintering thermal state.The prerequisite for producing high-quality sinter ore is that these critical parameters are within a reasonable range.Therefore,realizing the stable control of these core parameters under different operating modes is of great significance to improving the process stability and production efficiency.The sintering process has the characteristics of numerous parameters,complex mechanisms,multiple time scales,and multiple operating modes.These characteristics bring significant challenges to the operating mode recognition and the stable control of the sintering thermal state parameters.This thesis comprehensively uses time-series data analysis,mechanism analysis,intelligent modeling,intelligent decision-making,and intelligent control methods to focus on the research of the operating mode recognition with multiple time scales for the sintering process,the intelligent decision-making for the abnormal operating mode,and the intelligent control for the sintering thermal state parameters.The main research contents and innovative work of this thesis are listed as follows.(1)A long-time-scale operating mode recognition method based on time series data clustering is presentedAn operating mode recognition method based on time series data clustering is presented for the similarity of time series data under long-time-scale operating mode.Spearman’s rank correlation analysis and information entropy analysis are combined to select the input parameters to reduce the influence of irrelevant parameters on the operating mode recognition.A fuzzy C-means clustering method based on dynamic time warping distance is proposed to precisely describe the similarity of time series data.The voting mechanism fuses operating mode recognition submodels that are constructed by the naive Bayes classification method to improve the effectiveness of the operating mode recognition model.Experiments are produced using the raw running data collected from an iron and steel enterprise.The experimental result shows that the proposed method can recognize the operating mode more effectively than traditional data point clustering methods,with an average accuracy of 79.40%.(2)A long-time-scale operating mode recognition method based on time series fluctuation interval prediction is presentedAn operating mode recognition method based on time series fluctuation interval prediction is presented for the fluctuation of time series data under long-time-scale operating mode.The principal component analysis method realizes the data dimensionality reduction of the detection parameters to reduce the influence of redundant data on the operating mode recognition.The time series is converted into the fluctuation interval using the fuzzy information granulation method to fully describe the fluctuation and data distribution of the time series.The fluctuation interval prediction model of the burn-through point oriented to operating mode recognition is built for the characteristics of different operating modes with varying fluctuation intervals of burn-through point.The experimental result based on the raw running data shows that the proposed method can effectively predict the fluctuation interval of the burn-through point,and can recognize the operating mode with an average accuracy of 77.70%.(3)An intelligent decision-making strategy with the fusion of multi-time-scales operating modes is presentedAn intelligent decision-making strategy with the fusion of multi-time-scales operating modes is presented for the problem that abnormal operating mode will result in low-output and poor-quality sinter ore.A short-time-scale operating mode recognition method based on fuzzy rules is presented to reflect the multi-time-scales operating mode comprehensively.This recognition method uses one-way analysis of variance to select input parameters that significantly affect operating modes and uses fuzzy C-means clustering to construct fuzzy rules for the operating mode recognition.An intelligent decision-making method for thermal state parameters based on priority is presented to improve the abnormal operating mode in the fusion result.The raw running data of the sintering process is used to verify the presented method.The result shows that the presented recognition method has a better recognition effect than the existing methods,and the intelligent decision-making strategy can make decisions according to the fusion result of multi-time-scales operating modes.(4)An intelligent control strategy of sintering ignition process based on the prediction of ignition temperature is presentedAn intelligent control strategy for the sintering ignition process based on the prediction of ignition temperature is presented to solve the stable control of ignition temperature under unstable gas pressure.A prediction model of ignition temperature based on the combination of mechanism analysis and parameter identification is constructed to predict the change of ignition temperature caused by gas pressure fluctuations.An intelligent switching controller based on operating experience is designed for the problem that the ignition temperature has different degrees of response to different gas valve adjustments.A rule-based gas flow controller is designed to solve the non-linear problem between the valve and the flow caused by the long-term use of the gas valve.The ignition temperature control experiment is performed using the raw running data.The experimental result shows that the presented control strategy can effectively control ignition temperature under unstable gas pressure.(5)A fuzzy control strategy of burn-through point based on the feature extraction of time-series trend is presentedA fuzzy control strategy for the burn-through point based on the feature extraction of the time-series trend is presented to solve the problem of stable control of the burn-through point in the presence of disturbances that are difficult to estimate.This strategy uses the Mann-Kendall test method to quantify the trend feature of the time series.The global trend feature estimates the influence of external interference on the burn-through point,and the local trend feature estimates the future change of the burn-through point.A fuzzy controller for burn-through point is designed to play the guiding role of trend feature on the control of the burn-through point,which realizes the effective combination of trend feature and operating experience.The control experiment for the burn-through point is constructed based on the semi-physical simulation system and the raw running data.The experimental result shows that the presented control strategy has a better control effect than the existing control strategies.The industrial application in the sintering plant of a large-scale iron and steel company in China shows that the proposed method can further improve the stability of the burn-through point.This thesis provides powerful guidance for operators to adjust the sintering process by studying the operating mode recognition with multiple time scales and intelligent control of thermal state parameters,which is of great significance for improving the stability of the sintering process and reducing production energy consumption.In future research,issues including the accuracy of operating mode recognition,the superiority of decision-making results,and the stability of thermal state parameters need to be further studied and explored.