RBF Neural Network Soft-Sensor Modeling of Rotary Kiln Pellet Quality Indices Optimized by Biogeography-Based Optimization Algorithm

Abstract

A soft-sensor model is proposed with the aim of predicting the finished pellet quality indicators (chemical composition, physical properties, and metallurgical properties) of the rotary kiln pellet sintering process. The model is based on the radial basis function (RBF) neural network that is optimized by the biogeography-based optimization (BBO) algorithm. Six variables that are associated with the reaction mechanism of the rotary kiln pellet sintering process and that are closely related to the quality indices of the finished pellets—the material thickness of the chain grate, the velocity of chain grate, the temperature of the kiln head, the temperature of the kiln tail, the rotary kiln speed, and the quantity of the fed coal—are selected as inputs to the proposed soft-sensor model, and the finished pellet quality indices form the outputs. Accordingly, a multiple-input-single-output (MISO) RBF neural network (RBFNN) soft-sensor model is established. The structural parameters of the RBFNN model are optimized by the BBO algorithm. The simulation results showed that the model yields better generalization results and has a higher prediction accuracy, and therefore, it is capable of meeting the requirements of real-time control as well as those of an online soft-sensor in the rotary kiln sintering process.