Flatness Pattern Recognition And Control For The Cold Strip Mill Based On Artificial Intelligence

Along with the development of national economy and the improvement of living standards, the demand of board strips is increasing unceasingly, simultaneously the product quality's request of board strips is also increasing. Flatness pattern recognition and control as relations product quality of board strips, is the key technology and has become the key research problem of steel industry. In recent years, the intelligence theory technology develops unceasingly, because intelligence method in modeling, optimization and control has powerful function, the intelligence research methods of flatness pattern recognition and control technology is an inevitable trend in the development of the modern flatness technology. The research subject of this article chooses the flatness pattern recognition and control for the cold strip mill based on artificial intelligence, which has the theory and the practical significance, studies the present theoretical knowledge of the intelligent recognition and control, carries on the comprehensive analysis to it, makes up for one's deficiency by learning from others' strong points, and carries on further research for the flatness control and pattern recognition using fuzzy neural network.Firstly, as the antijamming ability of the traditional least squares flatness pattern recognition method is poor, its precision is low, the result of the neural network flatness pattern recognition method has a long time network studying, being easy to fall into the local minimum and many other problems. This article fuses the merits of the fuzzy theories and neural network, fitting effectively three adaptive neuro fuzzy inference systems, and proposes a kind of flatness pattern recognition method based on adaptive neuro-fuzzy inference system. The result indicates that this method can overcome the above flaws very well, and can distinguish the common flatness flaw effectively, the recognition speed and the precision can be improved to some extent, and the recognition result is also very close to the actual value of flatness.Secondly, as flatness control system's nonlinearity and the close coupling, as well as the traditional effective function method flatness static state effective matrix's insufficient, getting through the massive production measured data's computation and analysis, this article proposes the dynamic effective matrix method for flatness control and sets up flatness forecast model based on subtractive clustering ANFIS (Adaptive Neuro-Fuzzy Inference System). Through the flatness effective matrix forecast model based on subtractive clustering ANFIS, it obtains the effective matrix on line which changes unceasingly, realizes the effective control of flatness and gives dual attention to the flatness production timeliness and the complexity.Finally, this article carries on the simulation application on the 900HC rolling mill using the dynamic effective matrix method for flatness control, confirms fully the validity of the flatness forecast model and the dynamic effective matrix method for flatness control which proposed through the simulation experiment result.