| There is a close relationship between the flotation process and bubble size. Reagents addition can directly affect the bubble size distribution. In order to guarantee the flotation running reliably and effectively,it is important to study the bubble size.Bubble size characteristics can be used as an indicator of the reagents addition. Therefore, establishing the dynamic relationship model between reagents addition and bubble size distribution is the main content of this article.First of all, to effective froth segmentation algorithms for foam image,which are collected by developed computer vision based monitoring system,so that acquire accurate bubble size data. In order to avoid data lost,which is caused by the mean, median and other simple data processing methods. It is put forward in this paper that probability density function (PDF) is researched as objects,and methods of non-parametric estimation describe the probability density distribution of the bubble size distribution.The designed nonparametric estimation is applied so that the output PDF is formulated in terms of dynamic weights.The dynamic weights is the characterization of bubble size PDF.Secondly, In order to establish the relationship between bubble size distribution and reagents addition, the RBF neural network approach is proposed. RBF neural network has a good global approximation ability and generalization ability of learning,but it has local minimum shortcomings in the approximation process.So using a genetic algorithm optimize the RBF neural network, the center,width and network weights are code again, and get optimized modeling, using genetic algorithm optimize RBF neural network and establish relationship between dynamic weights and reagents addition, so that received the relationship of the bubble size distribution between reagents addition.Results show that the approach has practical significance, can effectively guide the production. |
PDF Full Text Request