Coke Quality Prediction Model Based on Coking Coal Group Composition and Structural Parameters and Its Coke Forming Mechanism

Taking five kinds of coking coal and 44 sets of coal blending as the research object, the coal cup coking experiment was completed in a 40kg small coke oven environment, and the coal heavy group, dense medium group and sparse coal obtained by separating the whole components of coal were obtained. The mass fraction YHC, YDMC, YLMC and the infrared spectral parameters I3 and I4 reflecting the hydrogen bond association and the length of the aliphatic chain or the degree of branching were the main indicators. The BP neural network analysis method was used to establish the coke quality prediction model. And discussed the characteristics of the model, and analyzed the coke formation mechanism under the new model. The results show that using new coal composition parameters to predict coke quality has certain advantages, and the predicted values (CSR), micro-strength (MSI), coke reactivity (PRI) and post-reaction strength (PSR) are measured and measured. The values are well consistent, and the fitting correlation coefficients for y=x are 0.986, 0.982, 0.956, and 0.926, respectively. The model has a good predictive effect on CR, MSI and PRI. The average deviation of the nine predicted samples is 0.53%, 1.58% and 1.28%, respectively. However, the PSR prediction effect after the reaction is poor, the average deviation At 12.22%. The research results provide a good basis for establishing a new method of coking coal blending.