Abstract
The sintering process is a complex process with the characteristics of uncertainty, multivariable coupling, time-varying and time-delay. The Burning-Through-Point (BTP) , which is a import parameter in sintering process , is affected by many reasons and difficult to be controlled to the required precision by conventional control methods. This paper presents a new time-series forecasting methods, which is called Bayesian Least Squares Support Vector Machines（LS-SVM）. The method applies the Bayesian evidence flame work to infer automatically model parameters of LS-SVM regression.ALS-SVM model is proposed on the basis of the Bayesian LS-SVM models. Several intelligent forecasting key techniques of sintered ore’s chemical components including algorithms of nonlinear SVM in regression approximation, selection of kernel functions and parameters and standardizing of sample data,Bayesian evidence flame-work are studied ; and the control schedules of BTP based on interval optimization are analyzed. At last, a new intelligent forecasting system of BTP are designed and implemented. Experiment results show that the LS-SVM prediction designed within the Bayesian evidence framework consistently yields good generalization performances, which the method of combining Bayesian theory and LS-SVM is faster and more accurate for the BTP in compare with BP neural network and GM(1,1). DOI: http://dx.doi.org/10.11591/telkomnika.v11i8.308 7
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call