Transport of two-phase air-water flows in radial centrifugal pumps

Abstract

This dissertation aims at investigating gas-liquid transport by centrifugal pumps and the related performance degradation due to the gas accumulation in the impeller. Therefore, the focus is set to understand the interactions between the phases and the possible phase segregation. The investigations cover single and two-phase pumping performances, head degradation, performance hysteresis, two-phase flow regimes, flow pattern maps, flow instabilities and pump surging. The dissertation involves also the in fluence of employing closed or semi-open impellers, standard or increased tip clearance gaps, an inducer on pump performance and the two-phase flow regimes. For these purposes, two experimental test-rigs, manufactured from acrylic glass, were employed, i.e. 1) a horizontal diverging channel (static test-rig) and 2) a radial centrifugal pump (dynamic test-rig). The diverging channel was designed to investigate the process of gas accumulation that occurs similarly in the impeller, providing a first step toward understanding the flow behavior. The chapters and the main findings of the dissertation can be summarized as follows. Chapter 1 gives a general introduction on two-phase flows and fundamentals of centrifugal pumps. Chapter 2 presents a review of the previous studies, where the main findings of the literature are summarized, highlighting several gaps and disagreements. This confirmed that the transport of gas-liquid flows is still not fully clear yet, particularly the gas accumulation process. Furthermore, detailed and accurate numerical simulations and flow modeling are still not possible, due to the lack of measurement data and/or boundary conditions. Bridging this gap, is also one of the main objectives of this dissertation, allowing accurate comparisons with CFD simulations and potential model developments. Chapter 3 presents the test-rigs details along with all measurement devices employed. The results of the diverging channel are discussed in Chapter 4. The parameters leading to large gas accumulations were observed and discussed, revealing interesting insights. It was shown that the presence of large recirculation zones, visualized by Particle Image Velocimetry (PIV), can lead to gas bubble trapping and rapid accumulation. Even for very small gas volume fractions (ε = 0:05%), a gas pocket was observed. Quantifying the gas accumulation size by shadowgraphy, it was found that the accumulated gas can be reduced by 1) avoiding large separations in the main liquid flow, 2) ensuring high enough turbulence levels and/or 3) choosing a stratified flow regime after the diverging part. It was also shown that the accumulated gas strongly affects pressure recovery; a significant decay in pressure recovery is observed when more gas is accumulated. Furthermore, sample boundary conditions for the inlet velocity and the bubble size distributions (BSDs) are measured by Laser Doppler Anemometry (LDA) and shadowgraphy, respectively. Chapter 5 discusses the experimental results of the centrifugal pump. Comparing a closed and a similar semi-open impeller, it was shown that the single-phase performance of the semi-open impeller is slightly lower, due to the leakage flow. The difference is considerably larger when the tip clearance gap of the semi-open impeller is increased. When an inducer is installed, only insignificant changes could be seen in the single-phase performance. For two-phase flow, the semi-open impeller with a standard gap can generally resist gas accumulations up to ε = 3%, showing better performance than that of the closed impeller. However, the trend is reversed for ε between 4 % and 6 %, where the accumulation of huge pockets starts in the semi-open impeller. Increasing the semi-open impeller tip clearance gap leads to increased leakage flow, enhancing the gas accumulation resistance. This retards the sudden performance drop and provides more robust two-phase performance up to ε = 7%. Further, installing the inducer with the semi-open impeller resulted positively in improved performance, particularly for part-load ow conditions (up to ε = 7%). Interestingly, the pump exhibited substantially different two-phase pumping performances (hysteresis) for exactly the "same" flow conditions, depending on the history of setting the desired flow parameters. The semi-open impeller with standard gap involved strong hysteresis, especially when the air is reduced from an originally high flow rate. Increasing the tip clearance gap was found able to eliminate the performance hysteresis while installing the inducer could strongly reduced it. No significant hysteresis effects could be observed in the closed impeller. It was also found that the semi-open impeller involves generally lower instabilities and more limited surging conditions compared to the closed impeller. When the inducer is installed, the flow instabilities and the surging region could be reduced. The two-phase flow patterns were recorded by a high-speed camera and associated along with the undesirable phenomena (breakdown, surging, cavitation. . . ) directly to the pump performance curves. The resulting maps confirm the higher gas accumulation resistance of the semi-open impeller, which positively increases by increasing the tip clearance gap. The inducer has only a slight influence on the flow regime map. For ε 8%, a segregated flow regime occurs in the closed impeller, which was never found in the semi-open impeller. In this range, neither increasing the gap nor installing the inducer could significantly improve the performance due to the occurrence of gas-locking. A video library including time-resolved and cyclic recordings was generated in this work. The videos involve recording for the flow regimes of 115 various two-phase flow conditions. Furthermore, selected two-phase flow cases were optically analyzed by time-averaging to obtain the gas accumulation size. The time-averaged images are very useful to be compared with CFD simulations for validation. Using shadowgraphy measurements, the BSDs were also measured, which can be implemented for some two-phase flow models. Chapter 6 presents results of relevant CFD simulations to give more details about the flow behaviour in the channel, across the inducer, and in the pump. Some important flow features could be justified numerically. Additionally, the limitations of several common numerical models are discussed. Finally, conclusions and all findings of the dissertation are summarized in Chapter 7, together with recommendations for possible future work.