基于动态聚类算法的两相流流型识别方法研究

采集了水平管内气水两相层状流、搅拌流以及弹状流三种流型的差压信号，利用概率密度函数(PDF)方法对信号特征进行了分析，定义了反应PDF变化的四个特征参数，即，PDF波峰个数K1、PDF波峰峰值K2、PDF波峰位置K3以及PDF方差K4，然后利用动态聚类算法对三种流型进行了分类识别。结果表明，应用PDF结合动态聚类算法可以快速、准确的对气水两相流流型进行识别，这种方法是可行的。

A Study of Two-phase Flow-pattern Identification Method Based on a Dynamic Clustering Algorithm

Differential pressure signals of three flow patterns in a horizontal pipe, i.e., vapor/water two phase laminar flow, agitation flow and elastic flow, have been collected. Through the use of a probability density function (PDF) method an analysis was conducted of the signal characteristics. The four eigen parameters reflecting PDF function variation, namely, peak number k 1, maximum peak value k 2, peak position k 3 and variance k 4 , have been defined. Then, by making use of a dynamic clustering algorithm a classification/identification was conducted for the above mentioned three flow patterns. It can be shown that the use of PDF in conjunction with the dynamic clustering algorithm makes it possible to realize a speedy and accurate identification of the vapor/water two phase flow pattern, testifying to the feasibility of the method under discussion.