Application of BP Neural Network in Glass Defect Recognition

In the glass defect recognition system, BP neural network structure is designed by usi-ng the basic principle of BP neural network combined with characteristic parameters. In order to more accurately identify the surface defects of glass, this paper proposes a method of adding mo-mentum factor, introducing steepness factor and adaptive learning efficiency based on the traditi-onal BP neural network algorithm. The simulation results show that all three methods can improve the defect recognition rate. Only the method of introducing the steepness factor makes the optimal error and the expected error the closest, and it’s the best to improve the network convergence.