SOFT SENSOR OF FeO CONTENT IN SINTER BASED ON APSO AND LSSVM

The content of the sinter FEO is a comprehensive index of sinter production, sinter FeO content in real time to accurately predict the production has great significance in guiding the sintering. Through the multiple cycles of the cross-section images analysis, selection method based on the differential operator on cross-section images was proposed, it is the premise of content prediction. Then, a combination model of adaptive particle swarm optimizatioti (APSO) algorithm and least square support vector machine (LSSVM) algorithm that used to predict sinter FeO content was studied, and had comparison with the BPNN, LSSVM model. By match the prediction results with the real test results, it indicated that prediction fluctuation of APSO-LSSVM model is relatively stable, the average relative error is only 1.07%, R-MSRE of 0.56% and 1.59% respectively lower than the LSSVM and BPNN model, it has proved that APSO-LSSVM has a higher accuracy.