Dynamic Characteristics Of Multiphase Flow Based On Chaotic Attractor Morphology

Multi-phase flow, as a complex nonlinear system, is commonly observed inmany industrial applications. Up to now, the complex dynamics leading to theevolution of different flow patterns are still unknown. In this regard, new theoreticallymethods are strongly required for characterizing the underlying dynamics ofmultiphase flow. The chaotic attractor’s morphological characteristics that can beapplied to identify different flow patterns have been widely used in the field ofmultiphase flow signal analysis. In this thesis, we focus our research on the verticalgas-liquid two phase flow, vertical oil-water two-phase flow and oil-gas-waterthree-phase flow. Based on the conductance fluctuation signals measured fromvertical multi-electrode array (VMEA) sensors, we performed our research on themulti-graph chaotic attractor’s phase space embedding, the comparison of chaoticattractor’s probability distribution and the unstable periodic orbits of three-phase slugflows.The innovative achievements of this thesis are as follows:1. We propose a high dimensional phase space embedding method based on themulti-graph theory. After optimal selecting of the embedding dimension and timedelay, we map the high-dimensional vector point into the two-dimensional radialplane graph, i.e., the high-dimensional vector point is transformed correspondingly toa geometric polygon, and finally we get a new multi-graph chaotic attractor (MGCA).We investigate the trajectory polygon barycenters of three kinds of signals i.e. theperiodic signals, the Gaussion white noise and the chaotic signals. We find that thereexsit obvious difference among trajectory polygon barycenters of these signals. Byanalysing the barycenters trajectory of the vertical gas-water two phase flow VMEAsignals, we find that the slug flow has the simplest complexity, and its barycenterstrajectory reflect the quasi-periodic dynamic characteristic of the gas slug and liquidslug; The barycenters trajectory of the bubble flow and the churn flow are morecomplex than that of the slug flow, the dispersion degree relative to the referencesection of their barycenters trajectory reflect their complexity, and the bubble flow hasmore variable and complex dynamic pattern. We can identify the bubble flow, slugflow and churn flow by the combination of different barycenters trajectory momentquantity. This method is not complicated and has audio-visual geometry meaning,supplying a new approach for the high-dimentional chaotic attractor’s morphologicalanalysis. 2. We present the attractor comparison method to analyse the high water cutvertical upward oil-water two phase flow and the oil-gas-water three phase flow.Comparing with the adaptive optimal kernel time-frequecy representation (AOK TFR)method, the statistic value can reflects the differences between various flow patterns,especially for the oil in water bubble flow and the very fine oil in water bubble flowwhich are very difficult to distinguish. In addition, the statistic value of theoil-gas-water three phase flow can also indicate the transition of the flow patterns withchanging the gas and the oil phase volume fraction. The attractor comparisonapproach observing changes in the attractor as a whole and this makes the methodmore generic than the previously described methods. The stastic value larger than3indicates that a significant change has taken place in the hydrodynamic behavior ofdynamics. As to the flow patterns which is difficult to measure using single sensor ortranditional analysis method of the complex multi-phase flow system, the stastic valuecan indicate the change has occurred on the flow hydrodynamic, and it can be apowerful tool for studying the complex multi-phase flow.3. We use the close return method and the adaptive threshold selection method todetect the unstable periodic orbits from the signals measured from experimentaloil-gas-water three-phase slug flows. The results show that the period of the emulsiontype slug flow is longer than that of the oil in water type slug flow. In addition, wefind that the orbits of oil in water type slug flow are basically composed of one bigloop and one small loop, while the emulsion type slug flow orbits are composed oftwo big smooth loops. In this thesis, we also employ the adaptive optimal kerneltime-frequency representation (AOK TFR) to investigate the flow behaviors of twotypical oil-gas-water three-phase slug flows, and find that the energy of the oil inwater slug flow exhibits a dispersed distribution and its frequency spectrum consistsof various components distributes in a wide range. In contrast, the energy of emulsiontype slug flow distribution is rather concentrated and the high frequency component inits frequency spectrum is much less than that of the oil in water slug flow. Theseresults well agree with the detected UPOs’ structures and further indicate that the fluidmechanism underlying oil in water type slug flow is more complicated than that of theemulsion type slug flow.