Research And Application Of Soft Sensing Methord In The Process Of Naphtha Cracking To Ethylence

As a pivotal chemical product,ethylene production capacity is an important indicator of the development of petrochemical industry in a country and is also widely used in ecological field and agricultural field.The naphtha cracking method is a well-developed technology to produce ethylene.The process variables,such as temperature,pressure,and flow rate are the key factors affecting the process of cracking naphtha which can reflect the operating condition of the naphtha cracking unit.However,the measured data from sensors may be fault due to certain unexpected factors,such as device aging and tearing.Soft-sensing measurement method can effectively solve the problem that some key variables cannot be available in time due to technical difficulty or sensor failure.The accuracy of the model will be directly affected by the selection of variables and regression methods.Nonlinear correlations among the variables must be taken into consideration when the selecting auxiliary variables and regression methods.In this work,a new algorithm,MICSVR,is proposed,combing support vector regression(SVR)and maximum information coefficient(MIC),which is applied to measurement soft-sensing.Benefiting from the advantages of MIC in nonlinear correlation measurement between the process variables and target variable,the data redundancy can be avoided by selecting the appropriate modeling variables.Based on this,the SVR method is further applied to extract the correlations between the modeling variables and the target variable,by which a soft sensing model is established to predict the target variable and realize data reconciliation due to the sensor failure.First,data from the benchmark Tennessee Eastman process is researched to validate the proposed algorithm.The result shows that the SVR based on the maximum information coefficient is more effective than the traditional linear method.Second,this method is applied to the naphtha cracking unit.The result shows that the proposed algorithm can effectively realize the soft measurement of key variables and the data reconciliation when the sensor failure occurs.