Simulation of Airlift Pumps for Deep Water Wells A. NENES1, D. ASSIMACOPOULOS1, N. MARKATOS1, and E. MITSOULIS2* 1 Department of Chemical Engineering, National Technical University of Athens, GR-157 80 Athens, Greece 2 Department of Chemical Engineering, University of Ottawa, Ottawa, Ontario K1N 6N5, Canada * Author to whom correspondence should be addressed 2 Abstract A mathematical model for the simulation of water airlift pumps is developed, based on the "interspersed continua" approximation for two-phase flow systems, together with an algorithm that selects the appropriate friction correlation for specific flow regimes. The model presented can either predict the water or air flow rate for a given airlift system. Predictions obtained by the model were compared with a series of experiments performed by the Greek Institute of Geological and Mineral Exploration and were found to be in good agreement. The present predictions are far superior to those obtained by an existing simple model currently in general use. Keywords: airlift pumping, two-phase flow, flow regime prediction, finite-volume method 3 Introduction Airlift pumping was invented by Carl Loscher at the end of the eighteenth century (Giot, 1982). Operation is based on the pumping effect achieved when air is injected into a liquid or a solid- liquid mixture. This type of pumping system has a low efficiency in comparison with other pumping methods. However, simplicity in construction and absence of moving mechanical parts are two very important advantages that make it useful in certain applications, such as pumping corrosive liquids (sandy or salty waters) (Giot, 1982) and viscous liquids (e.g., hydrocarbons in the oil industry) (Giot, 1982; Kato et al., 1975). Airlift pumping is also used in shaft and well drilling (Giot, 1982; Gibson, 1961) (the drillings being lifted by underground water), undersea mining (Giot, 1982; Mero, 1968), and in certain bioreactors and waste-treatment installations, providing excellent aeration of the pumped fluid (Chisti, 1992; Tristam et al., 1992). A typical airlift pump involves a vertical pipe of length L divided into two parts (Figure 1). A suction pipe of length L between the bottom end and the air injection port (points 1 and i), and an e upriser pipe of length L between the air and discharge ports (points i and 2), which is partially u submerged by a length L. s The type of flow in the suction pipe is either one-phase (liquid) or two-phase (solid-liquid) while in the upriser pipe is either two-phase (air-liquid) or three-phase (air-liquid-solid). The upriser pipe can be of constant or varying diameter, increasing from injection to discharge point (tapered systems). The latter are much more efficient when pumping from large depths, because this ensures slug flow along the upriser. Otherwise, i.e., when a fixed diameter system is used, due to gas expansion, the flow changes to annular, which is characterized by poor pumping efficiency (Giot, 1982). 4 Air supplied from a compressor is injected through an external or internal airline (Figure 2). At the beginning of pump operation, an initial drop in water level is observed, depending on the rate of pumping. There is also an additional drop in water level during pump operation, but it is usually very small and for simplicity omitted. Thus, two water levels are defined, one at idling conditions, and one during pump operation. The first level determines the compressor hydraulic overhead, i.e., the pressure in which the compressor must initially supply air for the pump to start operating. The second level affects operation parameters (water outflow, submergence, etc.), and determines the pressure at which the pump must supply air during steady-state conditions. Although external airline systems are more efficient, internal airline pumps are more frequently used because of their versatility and ease in assembly. As the water level inside the well fluctuates or changes, maximum efficiency can always be achieved by changing the airline length inside the upriser. Simulation of the pump is essential for determining the optimum operational conditions. For this purpose, various correlations (Zenz, 1993) and simple mathematical models (Kato et al., 1975; Casey, 1992) have been proposed. This paper presents a new model for the pumping of a liquid (water) and uses a more sophisticated approach to simulate the flow of the two-phase mixture in the upriser part. The model uses the full differential equations describing two-phase flows, that are based on the well-established interspersed continua concept. Finite-volume techniques together with the interphase-slip algorithm (IPSA) (Markatos and Singhal, 1982) are used for solving the system of differential equations. Friction terms in the momentum equations are calculated by correlations appropriate for various types of flow regimes. A flow regime map (Taitel et al., 1980) predicts the flow pattern at any point in the pumping system, using the local flow rate and physical properties of air and water. The model can give predictions, among others, for important design parameters such as the liquid outflow rate for a given type of compressor, 5 and the air flow rate needed to achieve a certain liquid flow rate from the well (i.e., air compressor specifications). Furthermore, the present work includes analysis of real field data collected from experiments on the outflow rate of the airlift pump. The experimental results have been compared with predictions given by the new model and by another one currently in general use. Mathematical modelling Kato et al. (1975) proposed a simple model for airlift simulation, based on the momentum balance along the upriser and the use of a mean air-volume fraction (see appendix for a detailed reference of working equations). The mean air-volume fraction model is valid for both internal and external airline systems, and a simplified version for external airline systems is given by Giot (1982). The model, although simple in use, has two major drawbacks: (i) predictions are acceptable for wells up to 11 meters deep, due to the assumption of a single flow regime along the upriser (slug flow), and (ii) the model is heavily dependent on empirical information (correlations) needed for air-volume fraction and friction drop calculations. In the present work, the simulation of the air-lift pump is carried out through a full hydrodynamic model solved numerically using iterative procedures. The low accuracy of the mean air-volume fraction model mentioned above can be improved by describing the flow along the upriser with a standard set of differential equations, suitable for two-phase flows. Empirical correlations are used only for calculating the friction terms, while the possibility of many flow regimes is also allowed, the type of which is determined by a flow regime map. The differential equations are integrated and solved by using the finite-volume method. Flow in the suction pipe is calculated by simply applying the Bernoulli equation and inserting an additional term for friction in the entry region of the suction pipe. 6 The analysis takes place in two steps, one for the suction pipe and another for the upriser. The pressure at the injection point is calculated independently in these two steps. A physically acceptable solution is obtained when these pressures are equal, and this represents the convergence criterion of the model. A. Flow in the suction pipe The flow in the suction pipe is simulated by applying the Bernoulli equation between points 1 and i (see Figure 1): (cid:1)dP(cid:4) P = P - r gL +(cid:3) (cid:6) L (1) i 1 l e (cid:2) dz(cid:5) e f,l The pressure at point 1 is calculated according to the following formula, provided the water level during operation is known: 2 ( ) M P = P +r g L +L - x l (2) 1 2 l s e 2r A2 l 1 where x is the pipe entry loss factor and is approximately equal to 0.5. The friction pressure drop is calculated by the following relations: (cid:1)dP(cid:4) 4 M2 (cid:3) (cid:6) =- f l (3) (cid:2) dz(cid:5) m D 2r A2 f,l 1 l 1 (cid:7)(cid:9)16/Rel when Rel £ 2000 f =(cid:8) (4) m (cid:10)(cid:9)0.079Rel- 0.25 when Rel >2000 M D Re = l 1 (5) l A m 1 l 7 B. Flow in the upriser The governing equations are derived from application of mass and momentum conservation principles over differential control volumes. The approach is based on the space-sharing interspersed-continua concept (Markatos, 1986), according to which the two phases share the space, and each phase can occupy a certain point in space with a probability expressed by its volume fraction, R. The following assumptions have been made in formulating the equations: • steady-state operating conditions; • compressible gas phase; • no exchange of mass between phases; • exchange of momentum between phases only through interphase friction processes; • isothermal flow for both phases; • one common pressure field for both phases; and • one-dimensional variation of properties within a cylindrical co-ordinate system with the variation axis defined along the upriser. The independent variable is the distance measured from the injection port z (Figure 3). Although a steady-state flow is assumed, the unsteady set of equations (Markatos and Singhal, 1982; Markatos, 1986) is used, as the transient solution of the differential equations increases the stability and convergence of the algorithm. Continuity For the gaseous phase: ¶ R ¶ R (cid:11)¶ r ¶ r (cid:14) g + (R U ) = - g (cid:13) g + U g(cid:16) (6) ¶ t ¶ z g g rg (cid:12) ¶ t g ¶ z (cid:15) For the liquid phase: 8 ¶ R ¶ l + (RU )=0 (7) ¶ t ¶ z l l Momentum conservation For the gaseous phase: ¶ ¶ ¶ P (R r U )+ (R r U2 )= - R + f gl + f gw- R r g (8) ¶ t g g g ¶ z g g g g ¶ z z z g g where f gl and f gw are the gas-liquid and gas-wall friction terms, respectively. For the liquid phase: ¶ ¶ ¶ P (R r U )+ (R r U2 )= - R + flg+ flw- R r g (9) ¶ t l l l ¶ z l l l l ¶ z z z l l where f lg and f lw are the liquid-gas and liquid-wall friction terms, respectively. The interphase friction source terms f lg and f gl always satisfy the following relation: f lg = - f gl (10) The volume fractions at every point must satisfy the constraint (also known as consistency criterion): R +R =1 (11) g l i.e., the space is fully occupied by the two phases. The perfect gas law was used for calculating the air density. 9 The solution of the governing equations is possible after a complete set of boundary conditions has been defined. At the injection port i, the following boundary conditions are prescribed: R =1.0 R =0 l g (12) M U = l l R r A l l 2 At the discharge point 2, the pressure P is known (atmospheric conditions), and a free outflow 2 boundary condition is implied on the remaining four variables R , R, U , U : g l g l ¶Rg = ¶Rl = ¶Ug = ¶Ul =0 (13) ¶z ¶z ¶z ¶z C. Flow regimes and friction correlations Interphase friction is calculated from correlations that differ within each flow regime. In order to select the appropriate relation for each cell, a flow map proposed by Taitel et al. (1980) is employed. This map uses phase velocity, volume fraction, density and pipe position in order to predict the type of flow regime prevailing. During the solution this procedure of the finite-volume equations is repeated for every cell and allows the prediction of different local flow regimes and physical properties along the upriser. A typical map is shown in Figure 4. Change in flow regime results in jumps in the interphase friction factor, which may lead to convergence problems. In order to smooth discontinuities and ensure good numerical behaviour, transition regions between regimes are used (instead of transition lines), in which friction coefficients are calculated as a weighted mean of the correlations used for both regimes. 10 The wall-phase frictional forces integrated over the volume of the computational cell for phase i are calculated from the relation (Markatos and Singhal, 1982): F =(cid:17) f iwdV =0.5f r U2A (14) iw P iw i i iw where V is the volume of the computational cell, f iw is the volumetric friction force term given in P the momentum equation, f is the friction coefficient, r and U are the phase density and velocity, iw i i respectively, and A is the area of contact between wall and phase for the current cell. The iw friction coefficient f is calculated from the Blasius equation (Markatos and Singhal, 1982): iw f =0.079Re- 0.25 (15) iw i where the Reynolds number Re is based on the equivalent diameter of flow D for the given cell. i eq The quantities D and A depend on the flow regime, and the expressions used are given in Table eq iw 1 (Markatos and Singhal, 1982). Interphase friction is calculated by the following linear expression (Markatos and Singhal, 1982): ( ) F =C U - U (16) ip fip g l where C is the interphase-friction coefficient, and is calculated differently for each flow regime. fip In the present work, two expressions were used: - Bubble and slug flow (Cheng et al., 1985): 3 ( )3 C = 110.0r U - U R 1.0- R V (17) fip l g l g g P 8 - Churn and annular flow (Govan et al., 1991):