Coating Thickness Modeling and Prediction for Hot-dip Galvanized Steel Strip Based on GA-BP Neural Network

To overcome the large time delay, strong nonlinear characteristics and multiple interference sources, a zinc coating thickness model must be constructed as an essential component for the steel strip hot dip galvanizing system. A BP neural network model was proposed to model and predict the thickness of hot-dip galvanized zinc layer. In the model, the main influences of the coating thickness such as the strip line speed, air knife pressure, air knife to strip distance and air knife height of the hot dip galvanizing system are used as the model input parameters, and the coating thickness as the model output parameter. Simulations shows that BP neural network occasionally fell into local optimum. Then a genetic algorithm was introduced to optimize the BP neural network, and the initialization weights and biases of the BP neural network were optimized in advance. Simulations shows the above-mentioned GA-BP algorithm has improved prediction accuracy and converges faster compared to conventional coating thickness models, and can be used in the close loop zinc layer thickness control system as feedback subsequently.