Flow Pattern Identification for Gas-Liquid Two-Phase Flow in Minichannel Based on a New Optical Array Sensor

Abstract

This work develops a new optical photodiode array sensor for flow pattern identification of gas-liquid two-phase flow in minichannel. The array sensor is developed by the following steps to acquire comprehensive information of the flow and maintain a simple structure: 1) optical signals of the flow are acquired by a matrix array sensor with 60 sensing elements; 2) the signals are preprocessed and divided into 5 clusters by hierarchical clustering; 3) one typical sensing element from each cluster is selected; 4) the selected elements of four typical flow patterns are combined and the final sensor structure with 7 elements is determined. Experiments of flow pattern identification based on support vector machine (SVM) is carried out to verify the feasibility of the developed sensor in a 4.04 mm diameter channel. In addition, three types of feature vectors are built, including mean value, standard deviation and frequency, and their performance on the identification is investigated. The flowrate of liquid and gas phase range from 55~332 mm/s and 8~41468 mm/s, respectively. The investigated flow patterns include bubble flow, slug flow, stratified flow and annular flow. Results verifies the feasibility of the optical sensor. Among the three types of feature vectors, the standard deviation has the best performance, while the frequency has the worst. Meanwhile, the results also demonstrate that the combination of some specific feature vectors is an effective way to improve the identification performance.