Affiliations: [a] College of Electrical and Power Engineering, Taiyuan University of Technology, Taiyuan, China
| [b] Hong Kong Aircraft Engineering (America) Company Limited, North Carolina, America
Correspondence:
[*]
Corresponding author. Tian Muqin, College of Electrical and Power Engineering, Taiyuan University of Technology, Taiyuan, China. E-mail: tianmuling@126.com.
Abstract: In order to make the RBF hidden layer centres being established more adaptively and avoid the blindness, this paper proposes a fusion algorithm in order to optimize the parameters of the RBF neural network used in recognizing the state of coal flotation. Firstly, in the optimization algorithm, the improved immune algorithm was used to determine the center position and the number of hidden layer of RBF neural network. Before this, the immune algorithm has been improved in several aspects, such as the initial population selection algorithm and the method for segment selection of affinity thresholds. In addition, the antibody removal mechanism, antibody immune mechanism and antibody concentration regulation principle had also been added in immune algorithm. Secondly, in virtue of combining a fuzzy C-means clustering algorithm, the centers of the hidden layer were optimized accurately. Through the sample verification, the RBF neural network obtained by the fusion algorithm was proved to have been improved significantly in the accuracy of identifying the coal flotation state and has better generalization ability.
Keywords: RBF neural network, centers of hidden layer, froth image of coal flotation, immune algorithm, fuzzy c-means clustering algorithm, state recognition
