摘要：AbstractSintering process is the second most energy-consuming process in ironmarking. The main source of energy consumption is the consumption of carbon. It is significant to predict the carbon efficiency to reduce the energy consumption. In the paper, first, the comprehensive coke ratio (CCR) is defined and used as an index to measure the carbon efficiency by analyzing the sintering mechanism. The principal component analysis method is used to find the principal components affecting the CCR. Next, they are divided into different subclasses by adopting the fuzzy C-means clustering algorithm. Then, the least square-support vector machine (LS-SVM) sub-models are established based on the subclasses, and the parameters of the sub-models are obtained using an adaptive particle swarm optimization algorithm. Finally, a FCM-LSSVM model is established by weighting each LS-SVM sub-model with a fuzzy membership function. The simulations using actual production data show that the prediction accuracy of the FCM-LSSVM model is higher than that of a BP neural network model and a single LS-SVM model, and it meets the requirements of actual production.
