Chuanhou Gao

Modeling of the Thermal State Change of Blast Furnace Hearth With Support Vector Machines

For the economic operation of a blast furnace, the thermal state change of a blast furnace hearth (BFH), often represented by the change of the silicon content in hot metal, needs to be strictly monitored and controlled. For these purposes, this paper has taken the tendency prediction of the thermal state of BFH as a binary classification problem and constructed a ν-support vector machines (SVMs) model and a probabilistic output model based on ν-SVMs for predicting its tendency change. A highly efficient ordinal-validation algorithm is proposed to combine with the F-score method to single out inputs from all collected blast furnace variables, which are then fed into the constructed models to perform the predictive task. The final predictive results indicate that these two models both can serve as competitive tools for the current predictive task. In particular, for the probabilistic output model, it can give not only the direct result whether the next thermal state will get hot or cool down but also the confidence level for this result. All these results can act as a guide to aid the blast furnace operators for judging the thermal state change of BFH in time and further provide an indication for them to determine the direction of controlling blast furnaces in advance. Of course, it is necessary to develop a graphical user interface in order to online help the plant operators.

A comparative analysis of support vector machines and extreme learning machines

The theory of extreme learning machines (ELMs) has recently become increasingly popular. As a new learning algorithm for single-hidden-layer feed-forward neural networks, an ELM offers the advantages of low computational cost, good generalization ability, and ease of implementation. Hence the comparison and model selection between ELMs and other kinds of state-of-the-art machine learning approaches has become significant and has attracted many research efforts. This paper performs a comparative analysis of the basic ELMs and support vector machines (SVMs) from two viewpoints that are different from previous works: one is the Vapnik-Chervonenkis (VC) dimension, and the other is their performance under different training sample sizes. It is shown that the VC dimension of an ELM is equal to the number of hidden nodes of the ELM with probability one. Additionally, their generalization ability and computational complexity are exhibited with changing training sample size. ELMs have weaker generalization ability than SVMs for small sample but can generalize as well as SVMs for large sample. Remarkably, great superiority in computational speed especially for large-scale sample problems is found in ELMs. The results obtained can provide insight into the essential relationship between them, and can also serve as complementary knowledge for their past experimental and theoretical comparisons.

Design of a multiple kernel learning algorithm for LS-SVM by convex programming

As a kernel based method, the performance of least squares support vector machine (LS-SVM) depends on the selection of the kernel as well as the regularization parameter (Duan, Keerthi, & Poo, 2003). Cross-validation is efficient in selecting a single kernel and the regularization parameter; however, it suffers from heavy computational cost and is not flexible to deal with multiple kernels. In this paper, we address the issue of multiple kernel learning for LS-SVM by formulating it as semidefinite programming (SDP). Furthermore, we show that the regularization parameter can be optimized in a unified framework with the kernel, which leads to an automatic process for model selection. Extensive experimental validations are performed and analyzed.

Binary Coding SVMs for the Multiclass Problem of Blast Furnace System

It poses a great challenge to control the blast furnace system, often meaning to control the components of the hot metal within acceptable boundary, such as the silicon content in hot metal. For this reason, this paper focuses on addressing the multiclass classification problem about the silicon change in hope of providing reasonable blast furnace control guidance. Through the proposed binary coding support vector machine (SVM) algorithm, a four-class problem, i.e., sharp descent, slight descent, sharp ascent, and slight ascent of the silicon content in hot metal, is reduced into two binary classification problems to solve. To heel, the confidence level about these classification results is also estimated. Reliable classification effect plus very few binary classifiers make the binary coding SVMs full of competitive power for practical applications, particularly when the confidence level is high. The four-class classification results can indicate not only the silicon change direction but also the rough silicon change amplitude, which can guide the blast furnace operators to determine the blast furnace control span together with the control direction in advance.

An optimal control method for applications using wireless sensor/actuator networks

The wireless sensor/actuator networks (WSANs) can be used for spatially distributed control systems. With smart sensors and actuators, the WSANs are able to not only sense the control system states and report measurements, but also perform control and actuation. This paper investigates WSANs on their ability of control. A centralized controller is introduced into WSANs to make up closed-loop control systems, in which control decisions are made based on global network-wide information. A model of the control and communication over WSANs is made theoretically, based on which we achieved an optimal control method. It is demonstrated by simulations that the control method proposed could stabilize the control system quickly.

Constructing Multiple Kernel Learning Framework for Blast Furnace Automation

This paper constructs the framework of the reproducing kernel Hilbert space for multiple kernel learning, which provides clear insights into the reason that multiple kernel support vector machines (SVM) outperform single kernel SVM. These results can serve as a fundamental guide to account for the superiority of multiple kernel to single kernel learning. Subsequently, the constructed multiple kernel learning algorithms are applied to model a nonlinear blast furnace system only based on its input-output signals. The experimental results not only confirm the superiority of multiple kernel learning algorithms, but also indicate that multiple kernel SVM is a kind of highly competitive data-driven modeling method for the blast furnace system and can provide reliable indication for blast furnace operators to take control actions.

Data-Driven Modeling Based on Volterra Series for Multidimensional Blast Furnace System

The multidimensional blast furnace system is one of the most complex industrial systems and, as such, there are still many unsolved theoretical and experimental difficulties, such as silicon prediction and blast furnace automation. For this reason, this paper is concerned with developing data-driven models based on the Volterra series for this complex system. Three kinds of different low-order Volterra filters are designed to predict the hot metal silicon content collected from a pint-sized blast furnace, in which a sliding window technique is used to update the filter kernels timely. The predictive results indicate that the linear Volterra predictor can describe the evolvement of the studied silicon sequence effectively with the high percentage of hitting the target, very low root mean square error and satisfactory confidence level about the reliability of the future prediction. These advantages and the low computational complexity reveal that the sliding-window linear Volterra filter is full of potential for multidimensional blast furnace system. Also, the lack of the constructed Volterra models is analyzed and the possible direction of future investigation is pointed out.

Symmetric extreme learning machine

Extreme learning machine (ELM) can be considered as a black-box modeling approach that seeks a model representation extracted from the training data. In this paper, a modified ELM algorithm, called symmetric ELM (S-ELM), is proposed by incorporating a priori information of symmetry. S-ELM is realized by transforming the original activation function of hidden neurons into a symmetric one with respect to the input variables of the samples. In theory, S-ELM can approximate N arbitrary distinct samples with zero error. Simulation results show that, in the applications where there exists the prior knowledge of symmetry, S-ELM can obtain better generalization performance, faster learning speed, and more compact network architecture.

CHAOTIC FEATURE OF MARTIN PROCESS IMPOSED ON THE COSINE FUNCTION

By analyzing the phase diagram of Martin process on the cosine function, it is shown that with the change of system parameters the system will eventually converge to a chaotic attractor. The process is repeated and stable focus, period doubling bifurcation occurs during this process. Further computation gives the maximum Lyapunov exponent of the system and meanwhile, the bifurcation diagram is drawn. Thus it is proved from theory that the system exhibits strong chaotic properties.