Application Research On Some Methods Of Computational Intelligence

Traditional Artificial Intelligence (AI) refers to symbolism which is based on knowledge and the solution is searched by reasoning. While Computational Intelligence (CI) is based on data and the solution is searched by training and connecting. There are some main methods in it: fuzzy system, artificial neural network, genetic algorithms, chaotic systems et al.This paper focuses on the application of 3 main methods in CI: fuzzy technique, artificial neural network and genetic algorithms. By using the complements of artificial neural network and genetic algorithms towards each other, this paper combine these two methods to create a better algorithm.This paper started with a project of industrial monitoring and controlling system which applies fuzzy technique. Intensive researches on knowledge has been done. Fuzzy rules have been designed to work in this Active Lime Product Line Security Control System.This paper introduces the other two CI methods: artificial neural network and genetic algorithms to research on the non-linear problem to develop Active Lime Product Line Quality Control System.Neural network is an effective method to solve non-linear problem. Though BP algorithm neural network is able to approach any continuous function in the condition of suitable structure and value of weight. But the suitable structure and value of weight won't be provided by the network itself. BP algorithm neural network works on the gradient descent which will not converge or converge to local minimum. The genetic algorithm provides an effective method of global convergence of BP network.Though in some cases genetic algorithm is able to converge to an optimal solution, but the standard genetic algorithm is not sure to be converged to a global optimal solution. Thus contractive mapping principle is employed to develop a new genetic algorithm— Contractive Mapping Genetic Algorithm(CMGA).This paper provide a new method employing contractive mapping genetic algorithm to learning of multiplayer feed forward neural network. This new algorithm is called contractive mapping genetic algorithm neural network(CMGANN).In the end, this paper builds an Active Lime Product Line Quality Control model. Operation advices can be given by this model. The result of the experiment shows that, this model is credible and is able to provide real-time advices.