The Application of Neural Network Based on Particle Swarm Optimization in Pattern Recognition of Flatness Signal

In order to improve the precision of flatness recognition of strip, a new pattern recognizing approach of flatness was proposed. A kind of hybrid optimization strategy was used to train the neural network. That is to say, the weights were initialized by particle swarm optimization (PSO) algorithm, and optimized by back-propagation (BP) algorithm, and then the best result was obtained. Patterns of flatness were recognized by the above proposed algorithm. The membership grades relative to the six basic patterns of common flatness defect were obtained, and the corresponding flatness control strategy would base on them. The simulation of theory data and actual data from Benxi-Steel had been carried out. The results show that the approach can provide us with stronger anti-interference ability, and by the approach, the recognizing accuracy and speed are increased greatly