Abstract:With the development of large-scale blast furnaces and oxygen-enriched injection technologies, the role of coke in the blast furnace has been increasingly important. Establishing a coke quality prediction method with good applicability and high accuracy may be important for iron and steel enterprises. Support vector machine(SVM) is a machine learning method developed from statistics theories. It has many unique advantages in solving small sample problems, nonlinear and high-dimensional recognition problems. The prediction of the coke cold strength and thermal properties was based on parameters optimization of SVM method by the genetic algorithm. By comparing the deviations of the prediction results of the two factors and the five factors, it was known that a closer result can before casted when the SVM method building a model was used, considering the effect of the coal ash, the fineness and the average temperature of coke oven based on the effect of the volatile matter and the bond index of coal blend.