Genetic Algorithm Study on the Flow Distribution Characteristics of Pneumatic Conveying in Pipe Network System

摘要

This paper mainly conducts experiments on distribution characteristics of the millet and hollow glass beads which have the same average particle diameter and different densities in horizontal T branch pipe, analyzing related data. At the same time it uses BP network optimized by genetic algorithm to conduct simulation prediction. The results show that the distribution fraction of any material in two branch pipes of the same resistance is basically the same, the fluctuation of the material with less density and larger particles is larger when the superficial gas velocity changes. When two control valves opening are not the same, and A control valve fully open, with the B's reducing, the mass fraction which is assigned to the collection container B has larger overall decline, and the range of it gradually increases. It can be seen that the inflection point of fluid state changing in pipeline is related to density and has nothing to do with particle size. When the superficial gas velocity is smaller than the inflection point, with the decreasing of control valve opening in branch pipe B, the reduction extent of the mass fraction for the materials with less density and larger particles which is assigned to collection container B is larger. The comparison between the genetic algorithm predictive value and the experimental value shows that they are in good agreement. It has a higher prediction accuracy to simulate flow distribution characteristics by using genetic algorithm.