Prediction of Silicon Content in Blast Furnace Hot Metal Using Partial Least Squares (PLS)

Bhattacharya, Tathagata

doi:10.2355/isijinternational.45.1943

Cited by 107 publications

(55 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The recent study on evolving neural network models by a genetic algorithm 23) also found the tuyere heat loss, the specific blast volume and the gas permeability important. Also other investigators 20) have found permeability indices relevant for predicting the hot metal silicon content.…”

Section: Detection Of Relevant Inputs and Time Lagsmentioning

confidence: 99%

“…21,22) In most of these efforts, the input variables in the models have been selected on the basis of process knowledge, but in some of the papers more systematic approaches have been made. For instance, Waller and Saxén 12) applied an exhaustive search among linear finite impulse response (FIR) models using a large set of inputs with different time lags, Bhattacharya 20) used a partial least squares procedure to select relevant inputs, while Saxén et al 23) evolved sparsely connected neural network models of the silicon content by a genetic algorithm. However, in practically all approaches on nonlinear prediction of the silicon content by neural networks, a small set of potential inputs was always selected a priori.…”

Section: Introductionmentioning

confidence: 99%

“…Even though the main mechanisms behind the silicon transfer have been clarified, i.e., the gasification of SiO 2 from coke and coal ash to SiO(g) in the high temperature region at the tuyere level, and its reduction to silicon in the liquid iron, the most successful approaches to short-term prediction (e.g., on a tap to tap basis) of the hot metal silicon content are based on stochastic models because of uncertainty in the estimates of the internal state and dynamics of the high-temperature region. Numerous black-box models for the prediction of the hot metal silicon content have been developed, [8][9][10][11][12][15][16][17][18][19][20][21][22][23] and some recent papers have stressed the chaotic nature of the signal. 21,22) In most of these efforts, the input variables in the models have been selected on the basis of process knowledge, but in some of the papers more systematic approaches have been made.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Nonlinear Prediction of the Hot Metal Silicon Content in the Blast Furnace

2007

View full text Add to dashboard Cite

Section: Detection Of Relevant Inputs and Time Lagsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Nonlinear Prediction of the Hot Metal Silicon Content in the Blast Furnace

2007

View full text Add to dashboard Cite

“…And to achieve this, future information of silicon content is needed before blast furnace operators take corrective actions. This is one of the reasons that prediction of silicon content has attracted considerable research interests; among them are models like statistical model 1) and fuzzy models, 2,3) neural networks and nonlinear methods. [4][5][6] Recently, several researchers have paid attention to the chaotic and fractal characteristics of silicon content in ironmaking process, 7,8) which clearly indicates a strong nonlinearity.…”

Section: Introductionmentioning

confidence: 99%

Wiener Model Identification of Blast Furnace Ironmaking Process

et al. 2008

View full text Add to dashboard Cite

To account for the nonlinearity of blast furnace ironmaking process, a nonlinear Wiener model identification algorithm is presented. The system consists of a linear time invariant (LTI) subsystem followed by a static nonlinearity. The inverse of the nonlinearity is assumed to be a linear combination of known nonlinear basis functions and the linear subspace algorithm is used to identify the model. The inputs to the model are parameters regarded to be most responsible for the fluctuation of thermal state in blast furnace while the output to the model is silicon content in hot metal. The identified Wiener model is then tested on datasets obtained from No. 6 Blast Furnace from Baotou Steel. It is found that the blast furnace of concern is a short memory system, so that for each prediction the Wiener method is retrained. It is shown that the retrained model well improves the predictive accuracy.KEY WORDS: blast furnace iron-making; silicon content; Wiener model; subspace algorithm. the chamber, so that chemical reactions take place throughout the furnace as the material moves downward. The end products are usually molten iron and slag phases tapped from the bottom, and flue gases exiting from the top of the furnace.The critical operating parameters are the temperature and silicon content in hot metal. It has been shown that silicon content has a linear approximation relation with temperature. Furthermore, the silicon content is a good measure of the heat content and the course of the blast furnace process. Therefore, only silicon content is chosen as an output variable for the process model. In the meantime, a large number of variables are monitored and they become the basis for the blast furnace operators to judge the status of blast furnace. Since not all monitored process variables have an influence on the fluctuation of silicon content in hot metal, selections must be made. According to engineering knowledge and controllability of variables, we select out 7 process variables 20) as the model inputs and silicon content in hot metal as the model output to construct the predictive model. These process variables include: the quantity of blast, the temperature of blast, the pressure of blast, the quantity of coal powder, the pressure of top gas, the coke ratio and the quantity of oxygen. Delay time for each input variable for this specific blast furnace is given in Table 1. Subspace MethodsSubspace methods are a relatively recent development in the field of system identification which is most suitable for multi-variable identification. There are 3 main subspace methods, i.e., N4SID developed by van Overschee and De Moor, 15) CVA developed by Larimore 16) and MOESP developed by Verhaegen and Dewilde. 17) For the 3 algorithms, N4SID and CVA have been seen much more applications. The CVA algorithm was based on mathematical statistics and time series analysis methods while the N4SID algorithm is more closely related to linear system theory. For the present work, we use the N4SID algorithm to identify the linear block of...

show abstract

“…Thus, data-driven modeling is being investigated quite intensively recently in an attempt to solve this intractable problem. In the process of data-driven modeling, the frequently used tools include neural net, [6][7][8][9][10][11] partial least squares, 12,13) fuzzy mathematics, 14) nonlinear time series analysis, 15,16) chaos, [17][18][19] etc. The main motivation is that, most of these tools have universal nonlinear approximation capabilities and can approach any function in any precision.…”

Section: Introductionmentioning

confidence: 99%