A Study on Prediction Models for Coking Coal Quality Evaluation
This paper makes a study on the methods of establishing prediction models for coking coal quality evaluation. First, 162 groups of experimental data of coking coals were collected. Next, such indexes of the 162 groups of data as volatile matter (Vdaf), caking index (G), final shrinkage degree (X), maximum thickness of plastic layer (Y), average maximum reflectance of vitrinite in coal (Rmax) and standard deviation of reflectance of vitrinite in coal (S) were adopted for principal component analysis (PCA), and six principal components were obtained. It is proved that the principal component analysis is valid, and the first three principal components contain most information covered by the six original indexes. On this basis, the authors established some mathematic models to predict mechanical strength of coke tested at room temperature with the three synthetical indexes as independent variables by means of backpropagation neural networks (BP neural networks) and multivariate statistics respectively. Testing results show that the models based on BP neural networks which the authors establish fails to achieve the desired results while the models based on polynomial curve fitting is effective in prediction of mechanical strength of coke tested at room temperature achieves high accuracy, which lays a solid foundation for quick and accurate evaluation of the comprehensive quality of coking coal.
quality of coking coal prediction model principal component analysis
Shuguang Ouyang Zhiping Liu Xueying Zhou Zhongshu Dai
Wuhan University of Science and Technology, China Coking Company, Wuhan Iron and Steel Company, China
国际会议
2006现代科技国际研讨会(The International Workshop on Modern Science and Technology in 2006)
北京
英文
179-184
2006-04-01（万方平台首次上网日期，不代表论文的发表时间）