Study On Parameter Measurement Of Gas-liquid Two-phase Flow Based On Digital Image Recognition

The study of gas-liquid two-phase flow has important applications in engineering, such as oil, atomic energy, aerospace, dynamical and chemical engineering, etc. The two-phase flow system is a complex, nonlinear and dynamic system. There are interface effect and relative velocity between the two phases, so the measurement of two-phase flow parameters is more difficult. At present, how to identify the flow patterns rapidly by using modern information processing means has become one of hotspot in the research fields of two-phase flow. Image processing techniques have been used extensively in many different measurements of parameters. In this paper, the application of digital image processing in two-phase flow was more investigated. Several new methods were proposed for gas-liquid two-phase flow regimes (patterns) recognition. The main works are listed as follows:(1) In order to identify the flow pattern automatically and accurately, a new identification method was developed. Flow images were captured by a high speed CCD. The characteristics of bubbles such as area, width, height and center coordinates were obtained by using image processing techniques. A fuzzy reasoning method was used to identify the flow pattern. The experimental results show that the method is effective to identify the flow patterns of bubbly, plug, stratified, wavy, slug and annular in a horizontal pipe. The identification accuracy is shown as follows: bubbly flow is 93.3%, plug flow is 85.3%, stratified or wavy flow is 97.3%, slug flow is 98.6%, and annular flow is 92.7%. An estimation of the process time is 22 frame/s.(2)A multi-layer fuzzy neural network is proposed by improving on a traditional neural network to identify the two-phase flow pattern. Levenberg-Marquart optimized algorithm was used to learn the network, and its constringency is rapid. The experimental results show that the fuzzyneural network can accurately identify the flow patterns of slug, plug and bubbly in a horizontal pipe.(3) A new method was developed to measure the sizes of bubbles in vertical Pipeline. A rapid and simple filling algorithm was presented acorrding to bubble features. A new method was developed for bubble velocity by searching the united character. An estimation of the process tune is 19 frame/s.(4) A new image segmentation algorithm for overlapped bubbles is proposed. The algorithm was based on Mathematical Morphology method. First, the bright dot character of bubbles image was extracted, and then the bright dots were thicken. The unconnected objects kept unconnected and the increased pixels should not overstep the boundary of original bubbles during the thicken process. The thicken process was stopped when the bright dots do not increase pixels, and then the segmentation image was obtained. The algorithm can also segment the overlapped cells or particles.