ε-SVR-Based Predictive Models of Energy Consumption and Performance for Sintering

Abstract

For realizing energy conservation and burdening optimization of sintering process in iron and steel enterprises, as to the predictive issues of energy consumption and performance indices, the Support Vector Machine for Regression (ε-SVR) was introduced into sintering production system. A general modeling mode was proposed and the predictive model of energy consumption and several performances like chemical compositions was established by history data of sintering. Then, this model was compared with several other methods such as multiple linear regressions, ELM, BPNN and RBFN in a case study. Results show that the ε-SVR method can achieve qualified prediction results rapidly with the best accuracy and time efficiency.

A Data-Driven Model for Energy Consumption in the Sintering Process

“…A key element of SVR is to introduce a loss function, e, on the basis of support vector machine which is used to find an optimal classification surface to minimize the error between all the sintering observations and the surface [37]. The principle of SVR used in sintering process has been elaborated in previously published paper [38], and the procedure of a baggingenhanced SVR (B-SVR) is similar to those of B-ELM. Since SVR is a strong learning machine, the performance of B-SVR needs to be validated in real case.…”

A new pattern based Petri net to model sintering production process

“…To handle the large time delay in sintering process, predictive control methods are now widely used for BTP control (Cheng, 2010; Greenwell and Vahidi, 2010; Kouro et al, 2009; Lu and Tsai, 2007; Nolde and Morari, 2010; Wu and Xu, 2007). There are also some other techniques to be used in modeling sintering process such as BPNN (back propagation neural network), RBF (radial basis function), SVM (support vector machine) and fuzzy network and ELM (extreme learning machine) (Feng et al, 2012; Huang et al, 2006; Li et al, 2016; Meng et al, 2012; Mengjoo et al, 2000; Wang et al, 2014). But some of these models are limited by the fact that a large number of physical parameters are difficult to obtain.…”