论文部分内容阅读
There are so many complex factors affecting moisture content in oil in the working process of vacuum oil purification machine,which results in controlling production quality dtfficultly.An online quality estimation model,namely,soft sensor model,for the moisture content is presented so as to reduce the lag of traditional offiine measurement in the laboratory and improve the control precision for the machine.According to multiple easy-to-measure variables at different sensing points acquired from the configuration system,the hard-to-measure variable,the moisture content in oil,is computed by the soft sensor model.The rough set theory is firstly employed to compress the data for pretreatment,which can get rid of multicollinearity and reduce the dimension of input variables for the model.Then the Least Squares Support Vector Machine (LSSVM) is delivered for the moisture content soft senor nonlinear modeling.In the experiment,150 samples are employed for model training and other 150 samples for testing.Comparing LSSVM with BPNN and RBFNN based soft sensor models,results show that the proposed method features faster and more precise approximation ability.It has better performance of generalization for tracking the trend of moisture content variety in oil.