A Fig-Based Method for Prediction Alumina Concentration

Introduction

To obtain metal from hydrated aluminum oxide, Al2O3.H2O, known as bauxite, these elements must be separated by electricity in the smelting process called as aluminum electrolytic process. This reaction takes place in large, carbon-lined steel cells, or pots, through which a direct electrical current is passed.

The bottom of each pot acts as a cathode, or negative electrode. Carbon blocks are suspended in the pot to serve as an anode, or positive electrode. Inside the pot, alumina is dissolved in a molten electrolyte, composed mainly of the mineral cryolite. The electrical current passing from the anode to the cathode causes the oxygen in the mixture to react with the carbon anode to form carbon dioxide, while the aluminum settles to the bottom of the pot to be siphoned off to Casting and Fabricating. The whole process of aluminum electrolysis is show as Figure 1.

Figure 1.

Electrolysis of aluminum

In the aluminum electrolysis process, the stability of the alumina concentration is key issue to maintain high efficiency (Tian, 2003). It is difficult that an accurate prediction model about alumina concentration which can not be directly measured is established due to high temperature and strong corrosive electrolyte. The alumina concentration shows the performance of a nonlinear time-varying and long delay because of the feeding rate and diffusion rate and cell temperature (Liu, 2008). Therefore, the prediction of the alumina concentration has been highlight field of the related research. Li (2004) proposes new fuzzy expert control method applied smart identification, multi-control mode and decision-making mechanism to achieve the alumina concentration prediction and real-time control. A prediction model based on wavelet neural network was proposed by Lee (2011). The cell resistance parameters were tracked to control alumina concentration in the ideal range through the adjustment of feeding amount. The GM(1,1)model is introduced into the aluminum concentration estimate (Zhang, 2008). The prediction method based on linear regression and orthogonal transform is determined to improve the accuracy of the alumina concentration forecast (Lin, 2010).. In recent years, Julier and Uhlmann proposed unscented kalman filter (UKF) method based on UT distribution (Julier, 2000, 2004; Pan, 2005), this method has capable of optimal estimation of nonlinear system and solve the problem of linear model and calculation Jaconbian matrix. However, the computation of model is still large when the dimension of input variables increases. Above methods can not solve this problem.

This paper presents a hybrid prediction model for an alumina concentration system in the context of cognitive informatics and cognitive computing (Wang, 2003, 2007, 2009, 2010, 2012; Wang et al., 2006). Granular computing is an emerging conceptual and computational paradigm for information processing, which plays a fundamental role within the field of computational intelligence (Zadeh, 1998, 2003). It concerns representing and processing complex information entities called “information granules”, which arise in the process of abstraction of data and derivation of knowledge from information. Granular computing may be viewed as an umbrella term covering theories, strategies, methodologies, techniques, tools, and systems that explore multilevel granularity in information processing, knowledge manipulation, and problem solving (Yao, 2008; Wang, 2009; Wang & Chiew. 2010). Within granular computing, a number of formal frameworks have been developed, among which the “Theory of Fuzzy Information Granulation”(TFIG) assumes a prominent position.

The aim of this research work is to construct a hybrid prediction model for alumina concentration. In this work, the original data will grain to decrease computation. Support vector machine is used on behalf of fuzzy particle of cell parameters regression forecast.

1. Fuzzy Information Granulation (Fig)

Zadeh defines granulation of an object A as a collection of granules of A, with a granule being a clump of objects of points, which are drawn together by indistinguishability, similarity, proximity or functionality. He also further gives the definition of data grain as: