Investigation Of Recognition And Active Control Of The Two-phase Flow Pattern

Gas-Liquid two-phase flow exists widely in all kinds of industrialproduction process, while flow pattern as the most important and basic parameter intwo-phase flow system, can directly influence the measurement and research of otherparameters. Pattern recognition is always a hot research topic in current gas-liquidtwo-phase flow study. Because of the existence of such bad flow pattern as plug flow,serious hazards are posed in the industrial transportation systems, even for securitybreaches. As a result, the active control of the two-phase flow pattern has become oneof the hot areas of research on gas-liquid two-phase flow, and will become a certaindirection on two-phase flow study as well. In this thesis, the recognition and activecontrol of the two-phase flow pattern are investigated. The main research andachievements are showed as follows:1. Gas-liquid two-phase flow in horizontal pipe experiment system was designedand set up. Differential pressure signals adapted for flow pattern analysis wereacquired in experimental process. V-cone flow meter was adopted to producedifferential pressure fluctuation. The signals of three typical gas-liquid two-phaseflow pattern including bubble flow, slug flow, plug flow were acquired for furthersignal processing.2. Time-frequency signal analysis method based on the windowed fractionalFourier transform was used. As a result, the auto-terms of the time-frequencydistribution was gotten without too many cross-terms. The results of time-frequencyanalysis indicate that the energy of bubble flow mainly distributes in high-frequencyband20-30Hz, slug flow mainly distributes in high-frequency band20-30Hz andlow-frequency band0-10Hz concurrently, plug flow distributes in low-frequency0-5Hz.3. The joint time-frequency spectrum of the time-frequency analysis was used asrecognition image. First of all, the feature was extracted by ART method based onregion based shape descriptors proposed in the MPEG-7. Secondly, principal component analysis was used for dimensionality reduction. The last, Gaussianmixture model method was used to build models corresponding to different flowpattern and to achieve the flow pattern identification. The results showed that thismethod was viable and effective.4. The active control scheme was designed, treating porous plate as flowconditioner. To investigate the active control effect, differential pressure signals ofdifferent superficial velocity of liquid and gas and different positions as well wasacquired. Such work as discreteness analysis, joint time-frequency analysis, energyanalysis was done to prove the active control effect of the flow conditioner, laying agood foundation for the further research.