Iron ore pellets compressive strength prediction model based on KPCA-RBF

Abstract

To predict the key performance index (compressive strength) of Iron ore pellets, a prediction model based on Kernel principal component analysis (KPCA) and RBF neural network was proposed. This paper determined the input variable through the analysis of the chain grate machine — rotary kiln — ring cold pellet production process, deal with the sample data and simplified the model structure with Kernel principal component analysis algorithm (KPCA) and then established the pellet compression strength prediction model with RBF neural network. Using a global optimization performance of the simulated annealing algorithm to optimize the parameters of the network model, obtain the high precision prediction model of the system. The simulation results show that the proposed model can accurately predict the compressive strength of pellets, can overcome the disadvantage of large lag in the original compressive strength test method. The prediction model laid a foundation for the automatic control of the pellets compressive strength.