Outboard Engine Emissions: Modelling and Simulation of Underwater Propeller Velocity Profile using the CFD Code FLUENT

Transcription

1 6 t Australasian Fluid Mecanics Conference Crown Plaza, Gold Coast, Australia 2-7 December 2007 Outboard Engine Emissions: Modelling and Simulation of Underwater Propeller Velocity Profile using te CFD Code FLUENT J.O. Egerton, M.G. Rasul and R.J. Brown 2 College of Engineering and te Built Environment Faculty of Sciences, Engineering and Healt, Central Queensland University Rockampton, Queensland 4702 AUSTRALIA 2 Scool of Engineering Systems, Faculty of Built Environment and Engineering Queensland University of Tecnology, Brisbane Queensland 4000, AUSTRALIA Abstract Te mixing and dispersion created downstream of te marine propeller is critical to te spread and impact of pollutants introduced into te water from outboard motors. Suc propulsion systems vent exaust gases under water were a complex mass transfer of missions occurs to water. Tis paper presents te modelling and simulation of te propeller velocity profiles and initial plume spread created by marine propellers using te computational fluid dynamics (CFD) software code FLUENT. Te model is verified wit experimental data measured using a Laser Doppler Anemometer (LDA) in te controlled environment of a laboratory flume cannel. A working solution as been developed by employing te sliding mes metod and inducing a rotating flow field. Te model as fair agreement wit experimental results owever te study as exibited potential for model refinement and improvement. Considerably more work is needed to obtain an overall understanding of te flow field and gain an accurate description of te velocity profile downstream from te propeller. Introduction Te outboard motor is rougly a century old and as a marine power unit, its global production is approximately 700,000 units per annum wit annual growt of approximately 7% per annum. Tecnology breaktrougs ave increased its viability and application scope, continuing its growt and popularity over te larger, more cumbersome and expensive inboard engine. Altoug tere as been very little researc done on te dispersing action of te propeller, numerous studies to caracterize te emissions of outboard motors as well as te proportion of suc emissions transferred to te water during motor use ave been carried out at te Queensland University of Tecnology (QUT), Australia. Most modern outboard motors exaust te combustion products below te water wit an intention of reducing noise levels and marginal gains due to exaust jet propulsion. Due to boat forward motion tis metod of exaust release forms an under water line source plume. Initial plume spread and mixing is due to propeller turbulence and te wake created by te boat, wile later spread is due to ocean currents and turbulence. Tis paper will focus on te initial plume spread and mixing due to propeller motion. In Queensland alone, boats and sips release approximately 4.5 ML of oil into marine environment eac year, wile tis figure is about.0 million tonnes world wide [,2]. On top of tese figures are te substantial quantities of unburnt fuel and combustion by-products tat are released and dispersed by vessel propellers. Contamination to te marine environment by ydrocarbons as become a major concern wit many governments putting laws in place to stop te dumping of oily waste and contaminated ballast. Boating can ave a number of adverse effects on marine ecosystems attributable to noise, propeller contact, wake effects and in particular engine emissions. A lot of canges ave come about in regards to marine pollution over te past decade wit te single most effective improvement to outboard motors being te introduction of direct fuel injection (DFI) [3]. In two-stroke engines, [] Direct fuel injection engines emit 75% to 95% less ozoneforming exaust tan conventional marine engines do for te same orsepower. Four-stroke engines emit even less. Equipping engines wit catalytic converters also elped to reduce exaust emissions considerably wile increasing fuel efficiency. Oter advantages of tese two design updates include no-smoke starting, reduced noise levels, reduced operating costs, enanced trottle response and removing te need to pre-mix fuel for two-stroke motors. Wit very little researc being conducted on te dispersing effects of propellers, literature and information was scarce and as suc, understanding te fundaments of bot propeller action and turbulent flow was te key to te success of tis study. It is oped tat by understanding te dispersion caracteristics of te outboard plume a greater relationsip can be found wit te cemical reactions between dissolved/suspended combustion products and water borne reactants. Te initial aim of tis study was to develop a dispersion model using CFD and verify results wit experimental data. However, due to resource acquisition issues te main objective was redefined to modelling te propeller velocity profile using CFD and verification of te model wit experimental data measured using Laser Doppler Anemometer (LDA) by a team at QUT at te controlled environment of a laboratory flume cannel. Te experimental procedure for data collection and te process of modelling and simulation are briefly described. Te results discussed. Experimental Metods and Results An experimental study of te Jet of a Boat Propeller was conducted by Loberto and Brown using te closed loop flume 777

2 of te Department of Mecanical Engineering at Kyoto University, Japan [4]. Te specifications of te flume were lengt 2 metres, widt 0.4 metres and eigt 0.2 metres. Te experimental set-up comprised a two-blade propeller, powered by a variable speed electric motor using a flexible cable transmission, eld in place by a wing saped frame were te long axis coincided wit direction of flow. Te propeller diameter, D was 20 mm (tip-to-tip). Velocity measurements were conducted wit a 2D Dantec forward scattering Laser Doppler Anemometer. instabilities leading to te jet breakdown observed at 3000 rpm. Tis needs furter investigation. At te downstream end, te flume water level was controlled by a sarp-crested weir and maintained te dept ( + 2 ) at a constant 0.5 meters as sown in Figure. Measurements were performed at several longitudinal locations ranging from near-field, x/d = 2 to far te field, x/d =50, were x is te distance downstream of te propeller and D is te propeller diameter (tip-to-tip). Te data gatered from [4] is used to verify te results of CFD simulation in tis study. Figure 2: Mean velocity field, U / U m at 3000 rpm wit profiles from Eq. (). Te raw velocity data are given in Table. Tese velocity data are referred to a downstream location were x/d is equal to te ratio of propeller (20 mm) by distance in te x-axis, i.e. te distance downstream for x/d = 2.5 is equal to 50 mm. Figure : Scematic diagram of experiment set-up Te mean velocity field was recorded in te longitudinal and tangential axes from x/d=2.5 to x/d=50 for two propeller speeds, 500 and 3000 rpm. Downstream of te propeller, te evolving jet can be represented by a Gaussian profile once it is establised [2]. Typical results of mean velocity profiles are sown in Figure 2 for 3000 rpm. Te data ave been presented in a normalised format to empasise jet flow field evolution. Te data are compared wit te Gaussian equation given below, 2 r r' () U = U m exp 2 σ as used be Brown and Bilger [5] for a study of reactive plumes in grid turbulence were U, r, r' and σ are te m maximum jet velocity (m/s), radial distance from centreline ( m), radial offset of curve centreline from x = 0 (m) and standard deviation of Gaussian Profile. Overbars represent mean for velocity. Gaussian curves were fitted wit least square criteria to te data, using a steepest descent, unconstrained multivariable curve fitting procedure. Equation () was cosen because it comprised fundamental parameters tat are clear descriptors of te jet sape. At 3000 rpm, te jet velocity data exibited some scatter in relation to te Gaussian profile. In particular at te fartest downstream position (x/d = 50), te velocity data exibited a breakdown and irregular profile wereas a slower jet (500 rpm, results not sown) maintained Gaussian profile. Te likely causes for jet breakdown at fartest position could be jet s interaction wit wall wic may ave induced some Table : Experimentally measured velocity at different x/d Velocity x/d = 2.5 Velocity x/d = 5 Exp mean y position Exp mean vel y position vel (mm) (m/s) (mm) (m/s) Velocity x/d = 0 Velocity x/d = 20 Exp mean y position Exp mean vel y position vel (mm) (m/s) (mm) (m/s)

3 Figure 3 sows measured velocity profiles in m/s for different downstream distances in mm. Gambit Details Gambit was used to create geometry and grid (mes). Te tank volume was made as a single rectangle volume. A 24 mm diameter, 6 mm ig cylinder was created to ouse te propeller and to be te rotating volume inside te tank. Once te cylindrical rotating volume ad been situated in te flume tank (00 mm downstream from te tank inlet) were te propeller will be located, te volume was split wit te tank making sure tat te connected option was unselected. Tis creates two unconnected volumes tat combine to make te flume tank. Te propeller geometry was ten centred in te rotating volume and subtracted, leaving a null region wic was defined as a wall. Figure 5 sows te propeller ouse in te rotating volume x/d = 2.5 x/d = 5 x/d = 0 x/d = 20 Velocity (m/s) Y- position (mm) Figure 3: Experimental velocity profile for a range of downstream distances. Experimental measurements for x/d = 2.5 and x/d = 5 did not ave axisymetric profiles and ad scatter, wic could indicate tat te flow as not yet establised. Anoter factor to be taken into consideration is tat te measurements for te two curves, x/d = 2.5 and x/d = 5 were taken at regions of ig velocity and near te initial flow field wic is not stabilised, unlike te remaining two profiles tat lay furter downstream. Negative velocities near te flume bottom could be due to te discarge current induced by te propeller forces te water forward and causes some to be reversed and recirculated. Modelling and Simulation Process Te purpose of modelling of te fluid flow was to obtain an overall understanding of te flow field and gain an accurate description of te velocity profile downstream from te propeller. Te processes and steps used for creating CFD model is sown in Figure 4 [6]. Figure 4: Fundamental metod for creating a CFD model as given by te Fluent Manual [6] Figure 5: Gambit Propeller ouse witin te cylindrical rotating volume Following te creation of geometry, te discretization of te domain was done wic involved breaking te domain into a set of discrete sub-domains, or computational cells, or control volumes and is referred to as a mes. After creating te propeller saft te blades were made and aligned to an assumed pitc, in tis case, 35 degrees. It must be noted tat various blade pitces were analysed including 45 and 20 degrees, to see te affect on output velocities. Bot te boundary and continuum type parameters can be set in fluent, owever te difficulty is increased if all aspects of te geometry suc as edges, faces and volumes are not labelled. Hence te autor advises using te simple Gambit grapical user interface (GUI) to save time Fluent Details Tere are four steps tat sould be given consideration wen planning to solve a problem in Fluent and tese are, defining te model goals, coosing a computational model, coosing a pysical model and te determination of te solution procedure. Te break down of setting up Fluent for tis study was as follows:. Import te grid 2. Scale te grid 3. Ceck te grid 4. Define units 5. Selection of te solver 6. Coose te basic equations to be solved 7. Specify material properties 8. Define operating Conditions 9. Specify boundary conditions 0. Define Grid Interfaces. Set discretization 2. Cange residual monitor 3. Initialize 4. Iterate 5. Examine Results 6. Save Results, and 7. Review. 779

4 Fluent as te capacity to evaluate moving zones wit a powerful set of features. Te sliding mes model is te model of coice wen a more accurate simulation is needed and assumes te flow field is unsteady [6]. However, as wit most coices in Fluent, te option tat provides te greatest accuracy is associated wt increased computational demand and te Sliding Mes Model is no exception. Te motion of te propeller was realistically modelled due to te surrounding grid moving as well and simulating te interaction wit te stationary tank. Te set of conservation equations are solved in an iterative process for eac movement step and it is during tis state of quasi-steady calculations tat information is relayed between te interface linking rotating and stationary regions. Some of te issues faced during creating CFD model were te inability to utilise te user defined function (UDF) feature using C programming was required, generating an accurate yet low element mes to minimise computational time yet yield ig accuracy, simulation times and correct geometry generation. Te position of te propeller and rotating volume is 00mm in front of te flume inlet Flume inlet assumed to ave a uniform velocity of 0.04m/s Floating water surface made into a wall for ease of computation Boundary layers left out as te core flow profile is te major item for analysis Single inlet and single outlet for flume Figure 6 displays te establisment of te velocity profile over a period of time. Boundary Conditions Te boundary conditions used in te model are given in Table 2. Table 2: Te boundary conditions tat were used in te CFD modelling Boundary Condition Walls Default Tank Velocity Inlet Velocity: 0.04m/s (experimental conditions) Turbulence: Intensity and Hydraulic Diameter option. Intensity 5% Hydraulic Diameter 0.288m 2 Tank Outlet Outflow set to Tank Volume Set as liquid (water) and stationary Rotating Volume Set as liquid (water) and rotation in te x-axis at 3000rpm (anticlockwise) For fully developed flow te turbulence intensity at te core can be estimated as [6]: = 0.6 Red 8 I (2) were Red is te Reynolds Number based on te ydraulic diameter. For a duct wit dimensions of lengt, a, and widt, b, te ydraulic diameter can be calculated by: d ab = (3) a + b Reynolds Number for water (at a temperature of 20 degrees C) was calculated by, ρvd (4) Re = μ Figure 6: Te flow field establisment over a one minute period Te velocity contours establised after minute and 2 seconds are sown in Figures 7 and 8 for auto velocity setting and maximum velocity settings respectively. Tese contours display flow fields, as expected of a propeller. Setting te maximum display velocity to.5m/s (Figure 8), it can be observed tat approximately 400 mm downstream from te propeller te flow dispersion as spread enoug to reac te flume top and bottom. Results and Discussion Te following assumptions were made to refine and simplify te simulation of te flow distribution: Blade geometry was taken as a flat plate set to a specified angle (35 degrees) Tip-to-tip propeller blade diameter of 20mm Two blade propeller Propeller saft ignored, only ub included in model Figure 7: Te velocity contours created in Fluent in te XY plane (auto velocity setting) 780

5 At 75 mm in te y-axis for x/d = 2.5, te velocity is.08 m/s in te experimental data, wile te maximum registered velocity for FLUENT was 0.69 m/s. Tis as been sown in Figure 9, troug te grapical comparison of experimental and FLUENT acieved velocities. Figure 8: Te velocity contours created in Fluent in te XY plane (.5m/s Max Velocity setting) Velocity (m/s) Measured Predicted Y-position (mm) Figure 9: Experimental velocity against FLUENT velocity values Te inaccuracy is approximately 35%. However, due consideration must be made for te limitations employed by te CFD model, most of wic ave been outlined in te assumptions at te start of tis section. Neverteless, it is believed tat te core issue for lower velocities is due to inaccurate propeller geometry, as numerous tests trying different solving parameters wit results aving minimal to no variance. Furter assumptions can be made about accuracy due to propeller cavitation. Originally tere were conflicting velocities between 75 and 05 mm in te y-axis, for flow at x/d = 2.5 and x/d = 5 and ad sligt discrepancies wen in contrast to te more establised velocity profiles of x/d = 0. In Figure 3 te LDA conflicting measurements in te near stream, x/d = 2.5, can clearly be recognised in comparison to FLUENT acieved results. Tese discrepancies could be due to air in te flow stream causing incorrect LDA measurements as a result of cavitation. Modelling cavitation was beyond te scope of tis study and as not been considered. Investigations were also carried out on te affect of propeller angle alterations to quantify te effect on te velocity exiting te propeller in te x-axis direction. However, te result was only a cange of m/s in te near flow field. Mixing caracteristics were observed wen te plates were angled at 45 degrees, compared to tose at 35 and 20. In order to make te velocity profiles in FLUENT matc te experimental range required increasing te rotation speed to 4500rpm, up from 3000rpm. Conclusions and Recommendations Overall, tis study is considered a single piece of a muc larger puzzle. By creating a flow profile model from experimental data, te foundations ave been set for future studies to elaborate on and enance wat as already been acieved. Quantifying te exaust product pat will elp to understand te relationsip between combustion waste and its dilution into bot water and sediment. Te benefits of a greater understanding can be passed down to improve environmental awareness wic is increasingly important for a century tat is looking to te future wit sustainable tinking. Investigation into more realistic marine blade geometry to provide iger trust and increase te velocities is currently underway. Suc a propeller will require a muc different approac to its development ten a flat blade, wit te creation of curved faces tat need to be stitced into a single volume. It is predicted tat suitable propeller geometry can produce te swirl tat causes te negative velocities displayed in experimental results. It is oped tat te simulated results using refined blade geometry will sow better agreement wit te existing experimental data. Inevitably due to time and resource constraints tis study as become merely a stepping stone, laying te foundations for more detail progression to continue. Acknowledgement We acknowledge te use of te Komori Laboratory at Kyoto University and te assistance of A/Prof K Nagata and Dr Y Ito in te conduct of te velocity experiments. References [] Oregon State Marine Board, 2006, Partnersip Encourages New-Tecnology Outboards, PWC Engines, Outboard Marine Engines, 25//2006, ttp:// [2] Alberston, M.L., dai, Y.B. and Jensen, R.A, 948, Diffusion of Submerged Jets, in ASCE Transactions (Fort Collins), [3] Curran, S., 2006, Marine Pollution, Recreational Boating, 5//2006, ttp:// [4] Loberto, A. & Brown, R.J., 2004, An Experimental Study of te Jet of a Boat Propeller, 5 t Australasian Fluid Mecanics Conference, Te University of Sydney, Sydney, Australia. [5] Brown, R.J. and Bilger, R.W., 996, An experimental study of a reactive plume in grid turbulence, Journal of Fluid Mecanics, [6] Fluent, 2005, Fluent 6.2 User s Guide, Reaction Design Inc., Labanon, USA. 78