Research On Kernel Partial Least Squares Based Monitoring Method For Molten Iron Quality In Blast Furnace

The steel industry has entered a stage of high quality development.The blast furnace ironmaking process is the upstream main process of the steel industry,and the quality of the molten iron produced directly determines the quality of the steel products in the subsequent process.Ensuring long-term stable and smooth operation and high-quality production of blast furnace ironmaking process play a vital role in the high-quality and sustainable development of the steel industry.Blast furnace ironmaking is a continuous and complex production process under high temperature,high pressure and closed environment.It has many process variables that can not only map out the current operating status of blast furnace,but also contain abundant product quality information.The quality condition of molten iron in blast furnace is reflected by the quality parameters of molten iron.However,the quality parameters of molten iron except the molten iron temperature(MIT)are difficult to be directly detected online and have time lag.In addition,the current monitoring technology of blast furnace ironmaking site is mostly combined with intuitive judgment and single variable indicator monitoring,which ignores the correlation between variables and the coupling relationship between different operating subsystems,which cannot timely and effectively monitor the production process.Therefore,using the internal relationship between the process variables which are easy to measure and the quality parameters which are difficult to measure,the fluctuations of the molten iron quality indicators in blast furnace can be monitored through the change of process variables,and the potential factors affecting the molten iron quality can be identified as early as possible,so as to provide valuable operational system adjustment information for improving the molten iron quality.It has important practical significance for ensuring the safe production of blast furnace ironmaking and improving the molten iron quality and economic benefit.In view of the above problems,supported by the National Natural Science Foundation Major Project "High performance operation control method and implementation technology of large blast furnace"(61290323),and on the basis of summarizing the development of process monitoring methods,combined with the characteristics of blast furnace ironmaking process,this paper carries out the research on kernel partial least squares(KPLS)based monitoring method for molten iron quality in blast furnace and conducts experiments on No.2 blast furnace of Liuzhou Iron and Steel Company in Guangxi province.The specific works are as follows:(1)Firstly,by analyzing the blast furnace ironmaking operation process and the molten iron quality parameters,36 variables of the main operating variables and state variables of blast furnace ironmaking process are selected as the input process variables of the monitoring model,and 5 quality parameters of molten iron which can comprehensively reflect the molten iron quality,including the silicon content([Si]),manganese content([Mn]),phosphorus content([P]),sulfur content([S])and molten iron temperature(MIT),are used as the output quality variables.The missing value and dimensionless processing of the experimental data are carried out.(2)In the anomaly detection of molten iron quality in blast furnace,aiming at the complex nonlinear characteristic of blast furnace ironmaking process and the problem of false alarms and missed detection in the monitoring process,a fault detection method of KPLS based on adaptive threshold is proposed according to the effectiveness of the kernel partial least squares method in quality monitoring and the superiority of the exponentially weighted moving average(EWMA)method in dealing with the problem of small changes and abrupt shifts.First of all,in order to accurately analyze the quality-related and quality-unrelated faults in the production process,a KPLS model is established according to the historical data.The Hotelling’s T2 and squared prediction error(SPE)statistics are used to monitor the operation conditions of process from different aspects.Secondly,combining the EWMA method based on the finite window length with the quality monitoring strategy,the derived adaptive threshold introduces the influence of historical data on the monitoring indicators,which improves the effect of fault detection.The validity of the proposed method is verified by the simulation experiment of Tennessee Eastman(TE)process.Finally,using the actual blast furnace data to carry out industrial test,the results show that the proposed method can effectively detect the abnormal conditions of molten iron quality.(3)In the anomaly diagnosis of molten iron quality in blast furnace,aiming at the problem of fault identification of blast furnace ironmaking process with strong nonlinearity and less prior knowledge of faults,a fault identification method of KPLS based on improved contribution rate is proposed.Firstly,in order to clarify the physical meaning of the existing fault identification method of KPLS based on contribution rate and simplify calculations,a scale factor variable is introduced into the new sample to construct monitoring indicator function.By performing Taylor approximation on the monitoring indicator function near the scale factor whose values are all one,two new statistics are obtained from the absolute value of the first-order partial derivative representing the contribution of each variable.Secondly,in order to further improve the accuracy of fault identification,the relative contribution rate is used to identify the final fault variables.The validity of the proposed method is verified by numerical simulation and TE process simulation of different fault types.Finally,conduct fault identification tests on the abnormal situations of the molten iron quality in blast furnace.The results show that the proposed method can accurately identify the cause variables of the abnormal operation in blast furnace.