Application Of Rbf Soft Senor Based On Ant Colony Clustering Optimization

The higher performance requirements of control systems for modern industrial process have been promoting the development of soft sensor technique. In modern complicated industrial process, some variables are very hard to be measured or even cannot be measured on-line by existing instruments and sensors. Soft sensor is an effective means of implementing the on-line evaluation of these variables. At present, soft sensor technique has become one of the most important research areas in process control field. This thesis based on the electrolytic copper industry production background, made deep and systemic research of the soft sensor based on improved RBF neural network. The main contributions are described as follows:In the paper, we make an overview on the theories of property and training algorithm of the artificial neural network and the RBF neural network. Introduce some traditional methods of training RBF neural network, then made the conclusion that the training problem of RBF neural network is how to choose the centers and the width of the evaporation coefficient.Ant Colony Optimization (ACO) was put forward in early 1990s, which is a type of intelligent optimization methods. Its predominant distributed pattern of problem solving achieves great success in combinational problems, and brings extensively attentions of related research area. Clustering is an unsupervised approach, which can divide the data automaticly by similarity. Now, the clustering method was used in chooseing the center of RBF neural network widely, but when the data was huge, the training speed will be slower and the training precision will be lower.As the followings, to conquer these problems, we proposed a new training algorithm to obtain the RBF centers, which is a novel training method of radial basis function neural network based on Ant Colony Clustering (ACC). This algorithm based on the characters of ant colony such as:the parallel search optimum and the ability of self-adaptive change information capacity using the volatile coefficient, then combined with clustering to make sure the center of each basis function of RBF. In order to predigest the structure of RBF neural network, we combine the two hidden-layer neural cell by comparing the similarity, meanwhile, we made a simulation of sunspots numbers to prove that this network can predict the results more exactly.At last, we use the soft sensor model based on RBF neural network in an electrolyte copper smelting plant. From the application results in a real electrolyte system, we can found that the simulation and the prediction effect were satisfied. Meanwhile, we made the model rectification based on the electrolyte component's back-ground. Contrasted with the un-rectified model, the rectified model was more suitable for the true industry application.