Coking Coals Analysis And Research On Coke Quality Prediction

The coking coals and its cokes from Guangxi Liuzhou Iron and Steel Group Co.,Ltd were taken as the research object.12 kinds of coking coals are selected for its analysis of industrial,process property and coal petrography microstructure.The strength of cold state,thermal properties and pore structure of the coke are measured and analyzed.The relationship between strength of cold state and thermal properties of coke with different coking coal is investigated;The prediction models of ash,volatile matter and sulfur content of blended coals are established by using the liner correction method,it was investigated that whether the method can solve the problem of deviation from additivity of each coal index to a certain extent;The prediction model was established by using the linear fitting method for the ash and sulfur content of the coke,and the relationship between the ash,sulfur content of the coking coal and its coke was examined.The method of support vector machine based on GA parameter optimization is used to predic the strength of cold state and thermal properties of coke.It was investigated that whether the support vector machine method was suitable for the prediction of coke quality and the influence of different quantity factors on the prediction results.The results show that:(1)Six of the 12 kinds of coking coal belong to low-middling ash and medium volatile matter coking coals.Among the 6 kinds of coking coal,there are 4 coking coals with lower softening temperature(401? ~ 409?),higher plastic temperature range(83?~96?),maximum fluidity(3.02~3.59)and the maximum thickness of the plastic layer(16mm~20mm).The others 2 kinds of coking coal have more than80 caking index.In a comprehensive view,these 6 kinds of coking coal are of relatively good quality,and its coke also have relatively good cold strength and thermal properties.(2)The method of linear correction is used to predict the ash,volatile matter and sulfur content of blended coal.The prediction results are more close to the actual value than the linear weighted method,and the correlation is significant at the confidence level of 95%.(3)The correlation and significance analysis of the ash and sulfur of coke prediction models have been carried out.The results showed that the two models had significant correlation at 95% confidence level,and has a certain practical significance for the coke ash and sulfur content control by coal ash and sulfur.(4)The relative error between the actual values and prediction result which is used support vector machines based on GA parameter optimization of the two factors and five factors were analysed,the average relative error is within 5% of the allowable error range,and it can fully meet the actual production requirements.Thesupport vector machine method can be applied to coke quality prediction and reliable prediction results can be obtained.By comparing the average relative error between the two factors and five factors,it is found that the average relative error of the forecast results under the five-factor condition is less than two factors.It can be seen that in the actual production,the forecast model under the five-factor condition is more suitable for guiding production.(5)Through using the different amounts of influencing factors in the prediction of coke quality,it was found that the volatile matter and caking index of blended coal are the main factors affecting the coke strength,but other factors have a certain influence on the coke quality.When the support vector machine is used to predict the quality of coke,if the three indexes of the coal ash,fineness and the average temperature of the coke oven can be added on the basis of the volatile matter and caking index of the blended coal,the prediction results can be obtained more close to the actual value.