This work presents a numerical investigation of the fluid flow in the first stage of a two-stage centrifugal pump with a vaned diffuser. A computational fluid dynamics (CFD) analysis is performed by using the ANSYS-CFX software for a wide range of volumetric flow rates and also for different rotor speeds. The numerical results are validated against measured values of pressure rise through the impeller and diffuser of the first stage with an overall good agreement. Nevertheless, not only the best efficiency point evaluated numerically is overestimated in comparison with the measured two-stage pump values but also the computed hydraulic efficiency of the first stage. Investigation of the flow pattern for different flow rates reveals that the flow becomes badly oriented for part-load conditions. In such cases, significant levels of turbulence and blade orientation effects are observed, mainly in the diffuser. In spite of different flow rates or rotor speeds, the flow pattern is quite similar if the flow dimensionless coefficient is kept constant, showing that classical similarity rules can be applied in this case. By using such rules, it was also possible to derive a single equation for the pump head to represent the whole operational range of the pump.

Summary This paper presents experimental data and a preliminary analysis of the influence of fluid viscosity on centrifugal-pump performance. Two centrifugal pumps, a conventional radial (specific speed Nq=8 rev/min) and a semi axial electrical submersible pump (ESP) (Nq=28 rev/min) were tested with 1-cp water and clear glycerin. Adjusting and controlling the fluid temperature in a closed test loop, it was possible to vary the glycerin viscosity from 67 to 1,020 cp within the range of light and heavy crude oils. The main purpose of these tests, in addition to appraising the influence of viscosity on the pump's overall performance through the measurement of the derating factors for head, flow, and power, was to supply detailed information on the energy-transfer processes taking place in the pump's internal components. To accomplish this, the pressure distribution along the flow path from the pump inlet eye to the discharge section, including detailed pressure difference across impellers and diffusers, was measured. Thus, in addition to measuring the flow rate, the overall pressure difference, the speed, the power and the mean operation temperature for fluids with various viscosities within a full range of operational conditions, detailed data on the energy-transfer processes performed by impellers and diffusers were also taken. Later analyses indicated that, in addition to the physical dimensions, operational conditions, and fluid properties, the pump performance is set by the strong flow interactions that exist between impellers and diffusers. In other words, these succeeding internal blade rows influence each other in terms of the head gain and the viscous dissipation effects. Thus, any generalizing approach dealing with the influence of viscosity on the pump performance must account for those interactions to give a proper measure of the derating factors over an extended range of operational conditions. Unfortunately, this is not true for the procedures available in the open literature. They lack representation and do not deliver proper correction factors for pumps that are not similar to those that generate the correlation database or for pumps working under operational conditions other than at the best-efficiency point (BEP). The data presented herein can be a launching point for a deeper analysis aimed to tackle these limitations.

Summary This paper presents an experimental study about the performance of a centrifugal pump, focusing on the effects of the fluid viscosity and the intake free-gas fraction. The first part evaluates a single-phase performance map proposed by Solano (2009). By use of dimensionless numbers, this map enables us to determine the pump performance when handling viscous fluids as a function of rotational speed at operational points other than the best efficiency point. Additionally, while the performance map of Solano (2009) is built solely on experimental data, this article expands his work by use of polynomial correlations that are based on basic phenomena that enable the prediction of the pump performance under a wider range of conditions. It was observed that these polynomial correlations are able to represent—with good agreement—the head, the brake horsepower, and the efficiency of a specific pump. Most of the work on multiphase flow through electrical submersible pumps has been performed with water as the liquid. Unfortunately, this does not correspond to actual petroleum liquids, which many times present viscosities a hundred times greater than that of water. In view of this, the second part of the present study focuses on the pump performance when handling a highly viscous fluid under two-phase flow. An extension of the application of the dimensionless numbers is proposed to fit two-phase-flow data for the same pump.

No abstract

Quero agradecer em primeiro lugar à Petrobrás por ter oferecido essa oportunidade de fazer o curso de doutorado. Quero agradecer a todas as pessoas, colegas do departamento, que colaboraram para que eu conseguisse realizar este trabalho e em especial: Ao meu orientador, o Prof. França, pelo incentivo, boa vontade e auxílio no atendimento às dúvidas. Ao Prof. Eugênio E. Rosa, pela sua contribuição através de idéias e sugestões e ajuda no encaminhamento das montagens de laboratório. Ao Eng. Marcelo Morandin e ao Técnico de laboratório Alcimar Ferreira, pela execução e auxílio nas montagens de laboratório. Aos colegas Ricardo Mazza e Silvio Gonçalves Dias pelo auxílio prestado na realização da parte experimental dessa tese. Aos colegas da Petrobrás, José Roberto Fagundes Neto e Francisco Alhanati, pelas sugestões que permitiram melhorar esse trabalho.

This work is concerned with the development of an artificial neural network (ANN), capable of classifying two-phase flow patterns, such as discrete bubbles, stratified, slug-flow, intermittent and annular regimes. Experimental operating data from the literature and the physical properties of the fluids were used to define and calculate dimensionless numbers. These numbers constitute the inputs of the neural model. They successfully describe the flow because they account for the competing forces occurring within the multiphase fluid. The training procedure was performed using a Levenberg-Marquardt algorithm. The methodology used to find the best network architecture is described in detail. All the flow regimes were accurately classified presenting only a small deviation. The final goal is to develop an automatic classification tool for multiphase flow patterns aimed at laboratories and field applications.

The submergible centrifugal pump (SCP) is one of the most common artificial lifting techniques employed in the Brazilian offshore scenario. However, free gas in the suction of the SCP is one of the most important limitations in the design of the pumping system. Gas-liquid flow in annular-ducts is found in the petroleum industry associated with gravitational gas separators that are applied with the SCP technique for oil production in directional wells. The main goal of this study is to investigate the behavior, i.e., flow patterns and maps, of gas-liquid flow in an annular duct at several inclinations in a setup with dimensions which are closer to real offshore application. High-speed video recording together with an objective technique based on the time-frequency pressure-signature analysis were applied for the flow pattern characterization. The experimental data were collected in the apparatus consisted of an inclinable test section, 0 to 90 degrees, with inner and outer diameters of 75 mm and 111 mm, respectively, hence the annular channel possessing an 18 mm gap. The total length of the test section was of 10.5 m. Air, water and oil at near atmospheric pressure constituted the gas and liquid phases. The air, water and oil superficial velocities were in the range of 0.02-30 m/s, 0.02-5 m/s and 0.005-0.5 m/s, respectively. Comparisons between data and flow pattern maps available in the literature for annular duct allowed the discrimination of regions and flow situations for which improvements of the transition modelling are required.

No abstract