Modeling of coke consumption in sintering process based on GA wavelet neural network

The combustion of coke provides the main energy for sintering process production, but there is lack of coke consumption indicators and effective prediction methods in the sintering process till now, so that it is very hard to optimize the sintering process correctly in order to reduce the coke consumption. Based on the analysis on the mechanism of iron ore powder sintering, a reasonable coke consumption indicator is defined. The factors affecting the consumption of coke are determined through the integrated use of the mechanism analysis and gray correlation analysis methods. As the sintering process´ features such as time delay, time variance and nonlinearity are considered, a GA-WNN predictive model is developed, in which the genetic algorithm is used to optimize the connection weights, expansion factors and translation scaling factors in order to overcome wavelet neural network´s shortcoming-that it is often trapped into local minimum. The GA-WNN model is applied to calculate the coke consumption in a simulation. The result demonstrates that GA-WNN provides an effective way to predict the coke consumption, which serves as a basis to optimize sintering process and reducing the coke consumption.