Dry Point Of Aviation Kerosene Soft Sensing Realizing In The Rectification

In petrochemical enterprises, many important process indexes are difficult to be measured online or directly. For example, the dry point of aviation kerosene is an important index of operation quality. But current methods can not obtain the real time value of dry point efficiently. To solve the problem, this paper chooses the distilling instrument of Dalian Petrochemical Cooperation as the research object, investigates the technology of soft sensor for measuring the dry point of aviation kerosene online. This paper implements the approximation of dry point based on data sampled from technical scenes, which make possible operation optimization and quality control.Firstly, this paper chooses thirteen process variables as auxiliary variables based on the analysis of the technique principles of crude oil distilling and the communication with operators. Then through the internal computing platform and DCS of the enterprise, we obtain related sample data, which provides necessary conditions for online approximation. Because of many errors existing in engineering data, this paper adopts data revision technology to fix them. For building the model of soft sensor, this paper studies many methods, such as PCA, RBF NN, SVM and LS_SVM, etc. A method combining PCA and LS_SVM is proposed in this paper. To test theoretic analysis, emulation experiments are done based on sample data. From the analysis, some conclusions can be drawn as follows. PCA can extract characteristic information from data efficiently. Substituting equation constraint for traditional non-equation constraint, LS_SVM obtains object model parameters by solving a set of linear equations, which avoids the quadratic programming and shortens the time of modeling learning greatly. Compared with RBF and traditional SVM regression method, under the same sample condition, the method proposed in this paper has better abilities of model approach and generalization than RBF, and less running time than traditional SVM. Experiments show that using PCA and LSSVM, this paper achieves satisfied results. This paper also discusses the problem of revision of soft senor model and proposes some effective methods, which can be applied to industry production, In conclusion, methods presented in this paper lay the foundation of dry point online estimation of aviation kerosene, and make it feasible to implement advanced control on quality parameters of aviation kerosene.