A GA-BP Neural Network Model for Predicting the Temperature of Slabs in the Reheating Furnace

Abstract

In this paper, 48 BP neural networks in series connection are constructed to predict the slab temperature, giving full consideration to the continuous heating process of the slab from inlet to outlet. In order to overcome the defects of BP algorithm, a hybrid training algorithm which combines GA method with BP networks is presented to improve the prediction precision of the model. It firstly optimizes initial weights and thresholds of BP networks with GA, and then trains the optimized networks by using BP algorithm. Simulation results show almost all errors of the test slabs are within the ±5°C limits, which indicates that the GA-BP model is verified to be valid and efficient, and is useful for online prediction and optimal control.