Research On Artificial Intelligent Flatness Controller Model Based On Immune Ant Colony Algorithm

Flatness control plays an important role in the production process of the plate strip.However, the technology of the flatness control is the key technology and difficulttechnology of the medium strip mills. The emergence of artificial intelligence methods hasopened up a new way to solve the flatness control problem. In this paper, in order to solvethe problem that it is difficult to establish the accurate mathematical model for the flatnesscontroller, the artificial intelligence technology is introduced into the automatic control ofthe plate shape. BP neural network flatness controller model based on the immune antcolony algorithm is established by comparing the ant colony algorithm optimized by theimmune operations and BP neural network.First of all, ant colony algorithm is optimized by the immune operations. Thetraditional ant colony algorithm has some shortages in the process of optimization, forexample, its convergence speed is slow and it is easy to appear the phenomenon ofstagnation. This paper puts forward a new vaccine selection method and vaccinationmethod. And a new algorithm called ant colony algorithm optimized by immuneoperations is proposed. It takes advantage of cross and the mutation operations to keep thediversity of the ant colony, and then the vaccine is inoculated to the immuned antibodies,so that the relatively optimal genes can be survived to the next generation, and so as toimprove the convergence speed of the algorithm, and the stagnation phenomenon isavoided.Secondly, the plate shape control prediction model is established based on theoptimized BP neural network. Every parameter value of the BP neural network isoptimized by the immune ant colony algorithm in this prediction model. It guides the antsselecting the path by the pheromone on the path and the expected transfer degree, and trainthe BP neural network by the values that the ants have selected. And it solves the problemsthat the traditional BP algorithm has because of using the gradient descent method, forexample, its training speed is relatively slow, and it is easy to convergence to the localextremum.Finally, the immune ant colony BP neural network flatness controller model based on dynamic influence coefficient matrix is established. The dynamic influence coefficientmethod is introduced into the flatness control process. The controller model is establishedby the local rolling data and the immune ant colony BP neural network. This modelcontrols the flatness shape by using different dynamic influence coefficient matrix underdifferent rolling state, to achieve the purpose of improving the precision and speed of thecontroller.