EXPERIMENTAL AND NUMERICAL MODELLING OF THE PRIMING OPERATION IN SPACECRAFT PROPULSION SYSTEMS Marcos Lema (1)(2), Jorge Pinho(1), Johan Steelant (3), Fernando López-Peña (2) and Patrick Rambaud (1) (1)
von Karman Institute for Fluid Dynamics, 1640 Rhode-St-Genése, Belgium, Email:
[email protected] (2) University of A Coruña, 15000 A Coruña, Spain (3) ESTEC-ESA, 2200 AG Noordwijk, The Netherlands
ABSTRACT The present study investigates experimentally and numerically the fluid hammer phenomenon in a confined environment, representing the priming operation of spacecraft propulsion systems. For the experimental approach, a new facility has been designed at the von Karman Institute, which can reproduce all the physical phenomena taking place in the propellant lines during a priming process. The simplified test procedure allows the recording of the pressure evolution of the liquid front with a fast response pressure transducer. The numerical approach is done with the numerical codes used by the European Space Agency (ESA) to support the propulsion systems design, i.e. EcosimPro/ESPSS and CFD-ACE+, simulating the experimental configuration both with 1D and 3D approaches. Comparing the numerical results with the experimental measurements indicates that the modelling of two-phase, two-component transient flows is a requirement for a proper simulation. 1. INTRODUCTION During the launch of a satellite, the liquid propellant is confined and isolated in the spacecraft tanks. Once the satellite has been ejected from the launch vehicle, the operation of spacecraft propulsion systems starts with the filling of the propellant lines with pressurized liquid propellant. This maneuver is called “priming” and is carried out by the fast opening of a pyrotechnic isolation valve. This may turn out to be critical if the corresponding overpressures are not correctly taken into consideration in the pipe line and sub-system dimensioning. Furthermore, since the propellant lines are initially vacuum pumped or filled with a noncondensable gas (NCG) at low pressure, the classical water hammer is further complicated by the presence of various multiphase phenomena, such as cavitation and boiling front. Finally, the driving pressure gas can be absorbed into the liquid propellant during storage and can desorb during the priming process. Due to the complexity of these multiphase phenomena occurring in spacecraft hardware, there are very few literature references describing experiments with all the
specifications of the above configuration, which are necessary for a proper validation of the physical models implemented in the CFD codes. Nowadays, various numerical codes are able to accurately predict the effects of liquid compressibility when computing a single-phase fluid hammer but they still need to be extended and calibrated for cases with cavitation (single component) and two-phase (two-component) flow. Furthermore, the treatment of a dissolved NCG in the liquid phase needs to be improved. In both cases, the existence of a well documented experimental database is mandatory for the improvement and validation of these physical models. 2. OBJECTIVES The aim of this study is to investigate the fluid hammer phenomenon in a confined environment with experimental and numerical approaches. The experimental results intent to complete and enhance the data already available in the literature, [1] and [2], by using an experimental set-up which can reproduce all the physical phenomena taking place in the propellant lines during priming. The numerical investigation is done with 1D and 3D CFD codes and the results are compared against the experimental data. The final objective is to improve and validate the numerical models. 3. EXPERIMENTAL APPROACH 3.1. Experimental facilities In order to study the priming process in the lab and to create an extensive database for validation purposes, two facilities have been designed and built at the von Karman Institute (VKI). The first facility follows the same conception rules than the propulsion systems in satellites and includes all the elements directly involved in the fluid hammer occurrence. The facility layout is shown in fig. 1, which is intended to be clamped vertically and horizontally on a 1m thick concrete wall. The main components are a pressure vessel, a fast opening valve (FOV), and a given length of the propellant line, ¼” diameter, with different configurations, referred to as “test element” hereafter. The facility also includes a vacuum system to set the
test conditions (test element initially vacuum pumped or filled with a NCG gas at low-pressure). pressure). A measurement module is attached to the end of the test element, which is the impact location of the liquid front. fro This module allows the unsteady measurements of pressure and temperature, with dynamic transducers flush mounted at the end wall and on the side wall of the measurement module. The test vessel, which is a spherical accumulator that can mount an elastic membrane, is equipped with an ultrasonic transducer to measure the speed of sound in the liquid. On the other hand, hand the membrane allows running experiments without absorption of the NCG into the liquid. The results of this test will provide very useful insights sights to understand how the dissolved gas affects the fluid hammer mechanism. Furthermore, since the working liquid is already saturated with air in standard conditions, the fluid needs to be deaerated to run experiments without any dissolved gas. The liquid uid deaeration is done by means of a depressurization process using a second accumulator, called deareation vessel, connected to the vacuum pump. The dissolved gas is removed by keeping the liquid in a low-pressure pressure atmosphere. The experiments are carried out ut with inert liquids as test fluids and nitrogen as driving pressure gas.
in fig. 2. The test element consists of a 2 m straight carbon steel tube, with internal diameter 12, 5 mm, and installed horizontally. The working liquid is water and the driving pressure gas is air.
Figure 2. Simplified implified facility built for experimental procedure validation The construction of the main facility has been recently concluded but it is too early to show results here. For this reason, the experimental measurements presented hereafter have been obtained only with the preliminary facility. 3.2. Experimental measurements of the pressure front evolution Due to the low pressure resistance of the steel tube used in the preliminary facility (P ( max < 10MPa) and the limited capacity of the vacuum pump (Pmin > 10kPa), the pressure range in the tank and in the test element was limited to 0,4MPa and 10kPa respectively, both in absolute scale. Thus, the test matrix was created with three levels of driving pressure in the tank: 0,3MPa, 0,35MPa and 0,4MPa, and with three partial vacuum levels in the pipe: 10kPa, 25kPa and 50kPa. The graph plotted in fig. 3 shows the unsteady pressure evolution recorded at the bottom end for three test conditions, where the initial vacuum level in the test element was kept constant.
Figure 1. Experimental facility layout The design of this facility has been supported with a second set-up, up, also built at the VKI, for the validation of the experimental procedure. re. This facility is described in [3] and it is a simplified version of the one described above, using a simple tank, a ball valve with pneumatic actuator and opening time t = 300 ms, and a measurement module with a single unsteady pressure transducer flushh mounted at the bottom end, as sketched
Figure ure 3. Experimental results varying the initial pressure in the tank and PP=10 kPa
For all the test conditions it was observed that the first pressure peak level is directly related to the pressure difference between the tank and the test element; when the pressure difference between the tank and the test element grows, the liquid front velocity also increases, producing a higher pressure raise. This is consistent with the Joukowsky’s relation (1), where the pressure surge intensity is proportional to the product between the liquid density, the speed of sound in the liquid and the velocity of the liquid front.
∆P = ρ a ∆V
model all aspects of a functional propulsion system. This is achieved by four libraries: Fluid_properties library, Fluid_flow_1d library, Tanks library and Turbo_Machinery library. The Fluid_properties library is a data base with the properties of most of the fluids used for aerospace application. Fluid_flow_1d library allows 1D transient simulation of two-fluid, two-phase systems. Tanks library is used with the Fluid_flow_1d library to simulate the aerospace two-phase, two-fluid propellant tanks. Finally, Fluid_flow_1d library allows the simulations of pumps, turbines and compressors.
(1)
The time needed before the appearance of the first pressure peak after the valve opening is also related to the liquid front velocity. Fig. 4 shows the time delay of the four first peaks obtained by varying the initial pressure in the tank and keeping constant the initial pressure in the pipe, PP=10 kPa. The fluid hammer peak that appears first is the one produced, as expected, with the highest pressure in the tank, which induces the highest front velocity. The time delay between successive pressure peaks is mainly determined by the speed of sound in the liquid. According to the slope of the dotted lines connecting neighboring peaks, we can deduce that the value of the speed of sound in the liquid between peaks is nearly the same for the three test conditions. This behavior shows that the two-phase phenomena affecting the speed of sound, i.e. vaporous and gaseous cavitation, has the same influence on the flow under the three test conditions presented here. We have to point out here that both the pressure in the tank and the vacuum level in the test element are quite conservative, which reduces the appearance of related two-phase phenomena. For instance, the initial pressure in the pipe is alreadu higher that the liquid vapour pressure (Psat=3540 Pa). This assumption will be proved with the final facility, where the pressure in the tank and in the test element can be set to a maximum level of 2 Mpa and 1 kPa, respectively. Finally, according to the results obtained with the simplified facility, it can be concluded that the experimental procedure proposed so far has proved to be valid for the modelling of the priming process. 4. NUMERICAL APPROACH 4.1. EcosimPro/ESPSS The numerical study is carry out with two commercial codes: EcosimPro/ESPSS and CFD-ACE+. EcosimPro is a 1D simulation tool, object-oriented and capable of computing steady state and transient physical processes that can be expressed in terms of differential-algebraic equations or ordinary differential equations. The code employs the European Space Propulsion System Simulation (ESPSS) library [4], which is needed to
Figure 4. Time delay between consecutive peaks. First peak referred to the valve opening. PP=10 kPa. The schematic view of the simplified facility created with EcosimPro is shown in fig. 5. Fig. 6 shows the pressure evolution computed at the bottom end of the test element when the test conditions are Pt=0,4 MPa and the test element is vacuum pumped at Pp=10 kPa. This result is compared in the same graph with the experimental data obtained with the same test conditions. The comparison gives a reasonably good agreement for the first pressure peak, even if the measurements were performed on a preliminary facility. We have to keep in mind that in this facility, some aspects as the fluid structure interaction, valve opening process and gas fraction dissolved in the liquid, were not correctly characterised. The following step will be to improve the validation by using the experimental data obtained with the new facility, where the dissolved gas or the two-phase effects on the speed of sound will be taken into consideration.
Figure 5. Schematic view of the experimental set-up under EcosimPro.
Figure 6. 2D domain used for the CFD computations (the tube length has been shortened)
Figure 6. Experimental and EcosimPro results with test conditions PT=0,4 MPa and PP=10 kPa kP 4.2. CFD-ACE+ CFD-ACE+ is a general 3D computational fluid dynamics and multiphysics solver. It has been updated with a full cavitation module,, which allows to work with a homogeneous mixture of liquid, vapor and NCG (with absorption and desorption), to model nonnon isothermal flows and liquid compressibility pressibility for fluid hammer simulations. In this module, the cavitation is modelled by a dedicated transport equation for the vapour mass fraction, which is solved along with the continuity and momentum equations for the mixture. mixture Furthermore, the vaporization and condensation rates are derived from the Rayleigh-Plesset Plesset equation for bubble growth and collapse [5].. The model version implemented in the cavitation module uses an advanced version [6], where turbulence induced pressure fluctuations are taken into to consideration by raising the phase-change change threshold pressure value, Pvapor. Regarding the liquid compressibility, the module proposes several models for different liquids. There are two models for water compressibility: Rouleau [7] and Elansary [8], onee model for oil by Rouleau, and two models proposed by Onera for ethanol and MMH. The numerical simulations with CFD-ACE+ ACE+ are based upon the simplified facility described above. We have used a 2D domain, which includes the liquid tank, the piping system and nd the 2 m long test element downstream the FOV, as sketched in fig. 6. The tank height is given by the liquid volume filling the tank before each test. The initial pressure conditions set in the domain are the same used in the previous section with EcosimPro, i.e. PT=0,4 MPa upstream the valve and PP=10 kPa in the test element, using water as liquid fluid and air as NCG.
Regarding the initial amount of NCG in the computational domain, it has been assumed a mass fraction of 2,5E-05 05 in the tank and upstream the FOV (red coloured in fig. 6) and a mass fraction of 1 in the test element (blue coloured in the figure). This means that the computational domain downstream the t FOV is initially filled only with air. This configuration assumes that the valve opens instantaneously, since the pressurized fluid is free to move downstream as from the first computational time step. The computations have been performed with the highReynolds turbulence model using standard wall functions and the water model of Elansary for the liquid compressibility. The numerical accuracy has been limited to first order, both in time and space, necessary for the stability of the computations when using u the cavitation module. Fig. 7 shows some of the most representative time steps after the FOV opening, and during the first impact at the forward end. The color scale represents the void fraction, which is the volume fraction of vapour vapo and NCG. Note that the domain height has been magnified 50 times for a better visualization of the test element. Before the FOV opening, the test element is only filled with NCG, as set in the initial conditions. Right after the valve opening, the pressurized liquid starts flowing into the test element, developing a turbulent boundary layer. When the liquid front reaches es the forward end, there is no longer any gas phase present in the domain, since the vapour has condensated and the NCG has been absorbed by the liquid. At this point, the flow velocity drops to zero, and the fluid hammer taking place creates a pressure wave that travels upstream. When the pressure wave reaches the tank, the new pressure conditions, i.e. PT=0,4 MPa,, creates a pressure gradient (P ( T
