The Identification Method Of Gas-Liquid Two-Phase Flow Regime Based On Digital Image Processing

Gas-liquid two-phase flow widely exists in modern industry process. The two-phase flow and heat transfer character are extremely influenced by the flow regimes, meanwhile the accurate measurement of other parameters and the performance of two-phase flow are influenced too. Therefore, the identification of different flow regimes has long been considered as a signification topic in the parameter measurement of two-phase system. The accurate identification of flow regimes is important for the operation and design of interrelated instruments.High-speed photography method is applied to the flow regime identification. It achieves a visual, non-contact measurement, and does not interfere with the normal operation of the production equipment. Moreover, in the high speed gas-liquid two-phase flow, we can adjust the frequency of snapping to be more clear flow regime images. Therefore, it expands the scope of the application for online flow regime identification. Based on a large amount of experimental data, digital image processing, wavelet analysis, neural network, support vector machine and complexity measures theory are used in flow regime identification. Intelligent flow regime identification method is discussed systematically from the aspects of theory and experiment.Firstly, the gas-liquid two-phase flow regime images are captured by digital high speed video systems in the horizontal tube. Secondly, after flow regime images noises are removed by using image processing technology, the flow regime different eigenvectors are obtained by extracting flow regime images statistical characteristics of the gray histogram, moment invariant characteristics, gray level co-occurrence matrix characteristics, wavelet texture characteristics, and wavelet packet information entropy characteristics. Thirdly, the gas-water two-phase flow regime identification model by utilizing BP (Back Propagation) neutral network, Elman neutral network, probabilistic neutral network and support vector machine are trained by using those eigenvectors as flow regime samples. So, the flow regime intelligent identification is realized. The test results show that the combination of image invariant characteristics and support vector machine is the best model among these models, but the difference of identifying rate is not distinct. Finally, the relation is discussed between complexity measures extracted from gray image time series and flow regime transition in gas-liquid two-phase flow. By analyzing the rules of three complexity measures (Lempel-Ziv complexity, fractal box dimension, and Shannon information entropy) with the changes of gas-liquid two-phase flow parameters, we could get the dynamics structure inversion characters of gas-liquid two-phase flow, and they provided an efficient, supplementary diagnostic tool to reveal the flow regime transition mechanism of gas-liquid two-phase flow and quantitatively identify flow regime. All these provide a new way to identify the gas-liquid two-phase flow regime from the aspects of theory and technology.