Digital Image Processing Based Flow Regime Identification of Gas/Liquid Two - Phase Flow

In most of the industries the two-phase flow pattern is obtained when gas and liquid flow simultaneously in a pipe. These two phase flows are complex, dynamic and are difficult to measure. An approach for identifying the flow pattern using Neural Network and Support vector machine is developed. Flow images are captured using high speed SLR camera and are preprocessed. After preprocessing the images, the textural features such as entropy, homogeneity, contrast, correlation and energy are extracted. The textural features extracted are given as the input to the neural network and support vector machine. Four typical flow regimes such as bubbly flow, slug flow, stratified flow and annular flow are captured from the experimental set-up. The results obtained shows that support vector machine method of classification is very effective with accuracy of 98.03 percent and hence higher recognition is done.