Sie sind auf Seite 1von 31

Intake Manifold Design using Computational Fluid Dynamics

Matthew A. Porter University of New South Wales at the Australian Defence Force Academy

Formula SAE is a relatively unique motorsport formula in that power is limited not only by size of engine but also by means of a restrictor placed in the intake system. This restrictor is the ultimate limitation in the power production of these vehicles. In order to provide the best air delivery to these engines an investigation was conducted into the internal geometry of a Formula-SAE intake manifold. An investigation into the effects internal geometry had on flow through an intake manifold was done utilising the Computational Fluid Dynamics package Fluent. The determination of strengths and weaknesses of different manifold designs was achieved through a comparison of flow characteristics through individual stages of the manifold, and finally through the manifold as a whole. With an improved understanding of the internal flow characteristics within restricted intake manifolds an improved design was produced for use on the 2009 ACME Racing vehicle.

CONTENTS
Contents ....................................................................................................................................................... 1 Nomenclature .............................................................................................................................................. 2 I. Introduction....................................................................................................................................... 3 A. Background .................................................................................................................................. 3 1. F-SAE competition .................................................................................................................. 3 2. 3. B. C. ACME Racing Involvement ..................................................................................................... 3 Competition Rules.................................................................................................................... 3

Aim ............................................................................................................................................... 4 Scope ............................................................................................................................................ 4 II. Foundation Knowledge ..................................................................................................................... 4 A. Introduction .................................................................................................................................. 4 B. Nomenclature of Intakes............................................................................................................... 4 C. Wave Theory ................................................................................................................................ 5 D. Fluid Flow Through ducts and Pipes ............................................................................................ 6 1. Pressure Losses in Pipes .......................................................................................................... 6 2. Velocity profiles ....................................................................................................................... 6 E. Summary ...................................................................................................................................... 7 III. Assumptions and set-up .................................................................................................................... 7 A. Introduction .................................................................................................................................. 7 B. Assumptions ................................................................................................................................. 7 C. Turbulence Modelling .................................................................................................................. 8 1. Spalart-Allmaras ...................................................................................................................... 8 2. 3. 4. D. 1. K-epsilon .................................................................................................................................. 8 K-omega ................................................................................................................................... 9 RSM ......................................................................................................................................... 9 Choosing a turbulence model ....................................................................................................... 9 Simulations of each turbulence model ................................................................................... 10 1

2.

Results .................................................................................................................................... 10

E. Summary .................................................................................................................................... 12 IV. Simulations ..................................................................................................................................... 13 A. Introduction ................................................................................................................................ 13 B. Restrictor .................................................................................................................................... 13 1. Model ..................................................................................................................................... 13 2. 3. 4. C. 1. 2. 3. D. V. Validation ............................................................................................................................... 14 Varying Diffuser Angle ......................................................................................................... 18 Varying Throat Radius ........................................................................................................... 19 Entire manifold ........................................................................................................................... 20 Model ..................................................................................................................................... 20 WS03 ..................................................................................................................................... 22 WS04 ..................................................................................................................................... 24

Summary .................................................................................................................................... 27 Design of a new Manifold .............................................................................................................. 27 A. Introduction ................................................................................................................................ 27 B. Impacts of simulations ................................................................................................................ 27 1. Restrictor ................................................................................................................................ 27 2. C. 3. 4. 5. Plenum and Intake Runners ................................................................................................... 27 new features ................................................................................................................................ 28 Restrictor ................................................................................................................................ 28 Plenum ................................................................................................................................... 28 Intake Runners ....................................................................................................................... 28

D. Summary .................................................................................................................................... 29 VI. Conclusion and Recommendations ................................................................................................. 29 A. Introduction ................................................................................................................................ 29 B. Conclusions ................................................................................................................................ 29 C. Recommendations ...................................................................................................................... 30 D. Summary .................................................................................................................................... 31 References ................................................................................................................................................. 31

NOMENCLATURE
ACME CAD CATIA CFD Dyno F-SAE MAP RPM RSM SAE SLS WS0# Aeronautical Civil and Mechanical Engineering Computer Aided Design Computer Aided Three Dimensional Interactive Application Computational Fluid Dynamics Dynamometer Formula Society of Automotive Engineers Manifold Air Pressure Revolutions Per Minute Reynolds Stress Model Society of Automotive Engineers Selective Laser Sintering Warren Smith followed by Vehicle Number

I.
A. BACKGROUND

INTRODUCTION

F-SAE competition SAE International, formerly the Society of Automotive Engineers, has been organising a competition for students since 1978. The competition has been designed to allow University students to compete against each other in a motor racing environment while applying many of the skills required of the automotive industry, not just engineering disciplines. The competition is based around the hypothetical scenario that the team has been approached by a manufacturing company to produce a prototype formula style race car. The vehicle is to be marketed at a nonprofessional weekend autocross racer. The prototype will be evaluated for its potential as a production item through a series of competitions which analyse the thoroughness of design, the marketing considerations as well as the overall performance of the vehicle. (SAE International, 2008) The competition has been developed to allow students to improve their overall engineering skills by providing a real word project to which they can apply text book theories. In order to provide some extant of a level playing field, the organisers have imposed a set of rules on the competition. These rules are not designed to impinge on a teams creativity in producing their product, in fact the rules are very free in this regard, they do however put in place several requirements aimed at improving safety and reducing risk levels to all involved with the competition. Rules such as these are in place in almost every form of motor sport. The wide variety of vehicles produced at the competition each year illustrates exactly the freedom for innovation afforded by the competition. ACME Racing Involvement In 2004 the University of New South Wales at the Australian Defence Force Academy (UNSW@ADFA) made its first entrance into the Australasian F-SAE competition. Undergraduate students produced the majority of the vehicle and several of the components of the car were final year thesis projects for the students. The vehicle, being the first produced by the team, took the name WS01. It was powered by a standard Suzuki GSXR 600 motorbike engine. The vehicle had many custom made parts to suit the application and this was all housed in a specially designed steel space frame chassis. The majority of these components were manufactured at the School of Aeronautical Civil and Mechanical Engineering (ACME). 2005 brought the second iteration of the ACME Racing teams entry to F-SAE Australasia, WS02. This vehicle had many improvements over WS01 as it represented 12 months of testing and evaluation of WS01. Some of the areas for specific improvement were the intake, chassis, engine tuning and the entire suspension system. The 12 months of testing and evaluation, combined with the experience of a competition allowed paved the way for these significant improvements. 2006 was the third iteration in WS03, it incorporated many of the changes originally slated for 2005 that were not ready in time, as well as several new developments. These developments included a new differential housing, improvements to un-sprung mass as well as incorporating more work on the engine and intake. In incorporating these upgrades there was a loss of focus in build quality which adversely affected the performance of this vehicle at the 2006 competition. In 2007, WS04 was produced and was quite a radical change from WS03. It incorporated a fully stressed engine and a spur gear final drive system. These changes were made after considerable research done as a part of final year thesis projects. The team learnt mistakes of 2006 and significantly improved build quality of the final project. In pre-competition testing there were several failures relating to the gear final drive. In a last ditch effort to get to the competition there was some re-engineering conducted on the final drive and the system was not tested prior to attending the competition. At the competition the final drive failed during the auto-cross event. During the small amount of testing time of the vehicle it was noted that power output was lower than had been obtained in previous years. The team was unable to determine whether this was related to a lack of tuning, the new final drive or the modifications to the design of the intake. In 2008 the team decided not to attend the competition as it was scheduled during a time of military training. As a consequence of this only a small number of the team would be able to attend and significantly hinder the performance of the team. Subsequently the team has shifted their attention towards 2009. This has allowed extended development of concepts which were originally intended for 2008, which will now be incorporated in 2009. 3. Competition Rules The F-SAE rules contain a comprehensive set of guidelines and limitations for the intake system used on any car competing in the competition. These rules generally relate to the location of the intake, the manner in 3 2.

1.

which the throttle may be actuated, the sequencing of the significant parts of the system, forced induction and the intake restrictor. It is this last mentioned section which is of specific relevance to this thesis. Below is an extract from the 2008 F-SAE rules, full rules regarding the intake system can be found at annex A. (SAE International, 2008) 3.5.4.3 Intake System Restrictor In order to limit the power capability from the engine, a single circular restrictor must be placed in the intake system between the throttle and the engine and all engine airflow must pass through the restrictor. Any device that has the ability to throttle the engine downstream of the restrictor is prohibited. The maximum restrictor diameters are: - Gasoline fueled cars - 20.0 mm (0.7874 inch) B. AIM This thesis has three main aims. The first objective is to gain understanding of the flow characteristics inside an intake manifold which is fitted with an intake restrictor. The second objective is to produce a qualitative comparison between the intakes used on WS03 and WS04 The final objective is to use the knowledge gained on the flow characteristics of these intake manifolds to produce an improved intake package for use on ACME Racings 2009 vehicle

C. SCOPE The first aim of this thesis is to gain an understanding of the flow characteristics in a manifold fitted with an intake restrictor. The first step in this process will be to consider the restrictor alone and simulate the air flow through the restrictor using the Computation Fluid Dynamics (CFD) package, FLUENT. In conducting these simulations, two geometric parameters will be altered to determine if there is any effect on the flow through the restrictor. The next process will be to extend the flow into the full manifold model. The second aim of this thesis is to produce a qualitative comparison between the intake manifolds of WS03 and WS04. The knowledge gained by achieving the first aim will greatly aid in the analysis of these two previous designs. The analysis of the two manifolds will again be conducted using the CFD package FLUENT.

II.

FOUNDATION KNOWLEDGE

A. INTRODUCTION In order to more fully understand the work to be conducted by this project some background information is required. This information will allow a more thorough discussion of intakes and much of the work which has been done previously on intake manifolds. The two main areas which will be focussed upon will be general terminology of intakes and a background into pressure wave theory. Pressure waves are an area in which a great deal of research has been done. The pressure based phenomena are not only confined to the automotive industry but also appear in musical instruments and acoustics. The tuning of intake manifolds to harness these pressure waves has occurred since the 1970s. While these pressure waves are not an area of investigation in this project, the knowledge of them and the ability to harness them in design are of great importance. B. NOMENCLATURE OF INTAKES Almost all intake systems consist of a set of basic common components. These components are air filter, throttle body, inlet pipe, plenum, intake runners, fuel injectors, air temperature sensor, manifold pressure sensor and mounting flange. Typically these components are arranged in a common way, although there are some variations for certain high performance applications where there are no restrictions on the intake. The simplest way to describe an intake manifold is to consider the path taken by air from when it enters the intake system, to the point it enters the engine.

Figure II-1: A cross section of a basic intake manifold Air flows through the air filter which provides clean air to the engine, the sole task of the air filter is to remove any particulate matter from the air to reduce the potential for wear and or damage to engine components. The air the passes through a section of piping, the length of this section of pipe is almost entirely dictated by geometric constraints around the engine. From here the air passes through the throttle body, the role of a throttle body is to control the amount of air which is allowed to pass into the engine. Typically this is achieved by use of a butterfly valve. A butterfly valve is a flat typically round or oval shaped plate which rotates through 90 degrees from fully closed to fully open. Having been throttled to the appropriate flow rate for engine operation, the air now flows through the inlet pipe, or in the case of F-SAE an inlet restrictor, into an air reservoir. This inlet restrictor is of specific interest as it reduces the capacity of the intake manifold to flow air and so will be a significant focus of this project. The air reservoir is known as the plenum and acts as a reservoir of air for the engine to feed from and also a mass damper to help reduce any oscillations or unsteady air flows. The size and shape of the plenum can greatly affect the overall engine performance and also will be an area of investigation of this report. From the plenum, the air must complete its path to the engine. The air passes from the plenum to the cylinders via the intake runners. The design of these runners has a very profound effect on the power and torque characteristics of an internal combustion engine. The length of these runners is of most importance. Within the intake runners occur oscillating pressure waves which can increase the volumetric efficiency of the engine. By careful selection of runner lengths an engine can be tuned to a specific RPM band. A great deal of investigation has been done on so called tuned lengths for intake runners, both within the ACME Racing team and by other researchers, consequently this area will not be an area of investigation within this report. C. WAVE THEORY In order to understand the pressure waves which occur in an intake manifold it is easiest to consider the application in pipe organs. The overarching principle in a pipe organ is the way the pressure waves inside the pipes reflect back along the pipe based upon whether they encounter an open or closed end of the pipe. To briefly explain what occurs and for future reference, two main types of waves form inside the pipe. These waves are known as rarefaction and compression waves. A rarefaction wave is a wave of less than atmospheric pressure and a compression wave is one greater than atmospheric pressure. When a wave reaches an open end of a pipe a wave of opposite form is reflected back down the pipe, if the wave reaches a closed end, a wave of the same form is reflected back along the pipe. (Smith & Morrison, 2002) In an intake manifold there are two significant events in the intake stroke of the engine, these are the opening and closing of the intake valve. When the intake valve closes, a compression wave forms, whereas when the intake valve opens a rarefaction wave is formed. These waves reflect up and down the intake runner, in the same fashion as in a pipe organ. As these waves are created and propagate they interact with each other in

a similar fashion to any other sounds waves, they sum together to form either a wave of higher amplitude, or even possibly diminish to a wave of zero amplitude. In the instance where a rarefaction wave is created upon valve opening it travels along the intake runner to the plenum where a compression wave is reflected back to the valve, when this compression wave hits the intake valve it propagates into the cylinder and increases the pressure in the cylinder. D. FLUID FLOW THROUGH DUCTS AND PIPES An intake manifold is ostensibly a network of pipes and ducts which feed air into the engine to feed the combustion process. As such it is open to analysis and optimisation as any network of pipes and ducts may be. One well documented and theorised section of pipe flows involves a head loss, or pressure loss due to certain geometries within the flow, specifically for bends, valves, entrance and re-entrance flows. Another well researched characteristic of pipe flow is velocity profiles for both turbulent and laminar flows. Pressure Losses in Pipes Pressure losses in pipes are split into two categories, major and minor. Major losses occur due to the physical length of the pipe and the viscous losses associated with the friction between the wall and the fluid. Minor losses occur due to variations in geometry through the piping such as bends, elbows, valves, entrances and re-entrances. (Munson, Young, & Okiishi, 2006) The terms major and minor do not refer to the relative sizes of the losses necessarily, but in typical piping systems involving many long straight sections with few bends and valves the major losses are more substantial than the minor. In the case of an intake manifold however, the minor losses are far more significant, and typically dominate the pressure losses experienced. Several text books quote pressure loss coefficients for various geometries whether they be entrances, reentrances, bends or valves. While these particular values are important in an analysis of a pipe system their values are not important specifically for the design of a new intake, but their relative size is. 2. Velocity profiles Velocity profiles are of importance to the work being carried out as they are one of the tools utilised in validating the results of various simulations. The velocity profile within a pipe is a well studied phenomena and has been noted to depend on a number of factors. The predominant effect on the final shape is the nondimensional quantity Reynolds number, more specifically whether it is above or below the transition from laminar to turbulent flow. This factor ultimately determines whether the profile is parabolic (laminar flow) or much flatter (turbulent flow). The velocity profiles can be seen in figure II-2 below. 1.

Figure II-2: velocity profiles within a pipe (Chaurette, 2008) Further to these velocity profiles is exactly when they form. Typically flow enters a pipe with uniform velocity and as such these profiles develop over some distance. When a velocity profile no longer varies the further we go down the pipe is said to be fully developed, and the distance taken to get to this point is called the entrance length. Experimental data in conjunction with theoretical calculations have determined two equations for approximating the entrance length, one for laminar and one for turbulent flow, they are: (Munson, Young, & Okiishi, 2006)

for laminar flow

for turbulent flow For 104 < Re < 105 typical values for range between 20 and 30. Typical pipe applications involve pipe lengths that are many times greater than these values. However in the instance of an intake manifold we see values well short of this, the restrictors to be considered later have a typical of 10-15.

E. SUMMARY This chapter has examined many facets of background knowledge required to perform analysis of previous intake manifolds of the ACME Racing team, and to produce an improved design for future use by the team. The first section discussed the components of a basic intake manifold and their individual function. The second section provided a brief background about wave theory and how it can be harnessed to improve the performance of an internal combustion engine. The third section examined some of the background into viscous fluid flow through pipes and ducts. The intake manifold is a system of pipes and ducts and knowledge of the strengths and weaknesses of certain geometries and characteristics will be critical in performing accurate analysis and improving on past designs.

III.

ASSUMPTIONS AND SET-UP

A. INTRODUCTION CFD allows the simulation of fluid flows through or over models of any size and or shape, furthermore it allows an in depth look at the occurring inside a model with great ease. In the case of an intake manifold it makes a clear choice for examining the flow occurring inside the manifold itself. Another significant advantage of CFD is it allows the comparison of different models without actually having to spend any resources constructing the models themselves. This will allow the author to compare several variations in intake geometry in both diffuser angle and throat angle with great ease.

Figure III-1: axis-symmetric representation of the restrictor B. ASSUMPTIONS FLUENT can handle several different parameters and inputs in performing its simulations. Many of these inputs and modelling assumptions will have a significant effect on both the final solution achieved, and the time taken to reach that solution. In order to make some assumptions to simplify the problem and thus the overall computational time the first step was to consider the flow characteristics. Property Flow velocity Flow temperature Total Pressure Reynolds Number Value In excess of Mach 1 Up to 50C Atmospheric (in testing 94.4 kPa) Re > 30000

Table III-1: Flow characteristics through the restrictor The first simulation setting we considered was the properties of the material. The first assumption affecting the material we need to make is that the flow will be compressible: With flow speeds over mach 1, there are clearly going to be density variations in the flow. Consequently the fluid flowing through the models, in this case air, will be modelled as an ideal gas. With temperatures remaining low and relatively constant as the flow progresses through the restrictor or manifold, specific heat and viscosity will be set to constant. 7

The next consideration was to whether the simulation would be steady or unsteady, in considering the engine at redline of 11 000 RPM this puts an intake occurrence in each cylinder over 80 times per second. If we move to considering the pressure waves travelling up and down the intake runners, which have several reflections during each intake occurrence, then to accurately model the transient effects would require modelling several dozen time steps for each intake. This would require incredible computing power to produce a single simulation. Consequently the simulations will be conducted considering steady state. The next consideration to make was regarding turbulence in the flow. When Reynolds did his experiments concerning fluid flow in a pipe he found that the transition from laminar to turbulent flow occurred for 2100 < Re < 4000 (White, 2003). In the case we are considering, Reynolds number would exceed 30000; clearly some level of turbulence modelling would be required to capture this. C. TURBULENCE MODELLING The viscous turbulence modelling feature within FLUENT provides the user the ability to model turbulence making use of 4 different turbulence models, these are: - SpalartAllmaras - K-epsilon - K-omega - Reynolds Stress Model These models, all in simplistic effect, produce a time averaged equation to simplify the governing equations of turbulence, which if considered in full are of such high frequency and small scale that it would be too computationally intensive to run even the simplest of simulations. (Fluent Inc, 2005) In order to determine which model was most appropriate for this particular case of internal ducted flow it was necessary to consider the backgrounds and merits of each model Spalart-Allmaras The Spalart-Allmaras model is a 1-equation turbulence model which solves a transport equation for kinematic turbulent viscosity. This is a relatively new turbulence model and has applications to the aerospace industry, specifically those involving wall-bounded flows. (Fluent Inc, 2005) A validation study on the model conducted by Paciorri, Deconinck and Degrez from the Von Karman Institute in Belgium concluded that the Spalart-Allmaras model provided excellent agreement with experimental data for most models tested. For those models where agreement was not as good it still produced excellent correlation for pressure distribution and heat transfer but under estimated the size of separation regions. (Paciorri, Dieudonn, Degrez, Charbonnier, & Deconinck, 1997) A critical survey on numerical methods by Knight and Degrez investigating the prediction capabilities of various turbulence models relating to shock wave/boundary layer interactions concluded that the SpalartAllmaras model produced very accurate results when compared with experimental data. (Knight & Degrez, 1997) K-epsilon The k- turbulence model is a 2-equation turbulence model which independently calculates turbulent viscosity and a length scale. The two equations relate to kinetic energy of the turbulence k, and the rate of dissipation . The model has been widely used by industry and has become almost a standard by virtue of its economy of computational efficiency, accuracy and robustness for a wide range of turbulent flow applications. (Fluent Inc, 2005) A validation study for a k- model was conducted by Poroseva and Iaccarino in which they concluded that the k- model produced good agreement with experimental data, but that the k- model would often produce higher peaks in velocity than were obtained experimentally 2. 1.

Figure III-2: Velocity profile in a combustion chamber. Notation: Circles is the experimental data, dashed line is the standard k-e model, solid line is an improved k- model and the dash double-dot is a 4-equationn turbulence model As can be seen, the velocity profile by all three turbulence models produced a higher peak than was obtained experimentally and these peaks were generally sharper than what was obtained experimentally. (Poroseva & Iaccarino, 2007) The study mentioned in the section on the Spalart-Allmaras model on shock wave/boundary layer interaction indicated that while the k- model produced agreement with the trends of experimental data, the results were less accurate. (Knight & Degrez, 1997) K-omega The k- model is another 2-equation model similar to the k- model, it models the kinetic energy of the turbulence, k and the specific dissipation rate . The specific dissipation rate can be considered a ratio of to k. (Fluent Inc, 2005) Several journal articles have eluded to the sensitivity of the k- model on the upstream and or freestream values of turbulence variables, particularly . (Kok, 2000) and (Bredberg, Peng, & Davidson, 2002) While work has been conducted to reduce this dependence the update model has yet to be implemented into the version of FLUENT being utilised. In the case being simulated we only have an approximation of the turbulence of the flow entering the restrictor and this may indicate a potential weakness of this model. It will however still be included for comparison. RSM The Reynolds Stress Model is a 5-equation model in 2 dimensions and 7-equation in 3 dimensions. It calculates the individual Reynolds stresses utilising differential transport equations. The equations are derived directly from the momentum equation, the equations are used to close the unknowns of the full momentum equation. The added complexity of this model and the 5 or 7 equations that need to be solved significantly increase the processing power required to conduct simulations. Improvements to the algorithm have significantly improved the performance of this model and computational time is approximately 50% higher per iteration than the 2-equation models. (Fluent Inc, 2005) A study into Reynolds Stress modelling involving shockwave boundary layer interactions by Vallet of the Pierre and Marie Curie University compared the performance of several Reynolds-stress models and also considers a k- model. The study concludes that the RSMs could reproduce, quite accurately, the experimentally determined values for the flow, while the k- model failed. (Vallet, 2008) D. CHOOSING A TURBULENCE MODEL The findings of various literature reviews all have indicated that the previously mentioned turbulence models are, given the right set-up, capable of simulating the flow we wish to examine. Each model had identified strengths and shortcomings for various simulations and it was clear that no one model was able to be utilised reliably given any case. As a consequence there would have to be further investigations into each model in order to pick one as more suited than others. It was decided to conduct simulations applying each of the potential turbulence models and comparing the results obtained to each other, and to expected characteristics for pipe flow. 4. 3.

1.

Simulations of each turbulence model Four simulations were carried out, all on the same restrictor model, all with the same pressure drop across the restrictor of 20kPa. This pressure drop was larger than experienced inside the inlet manifold but was believed to be a good value as it would definitely incite choked flow. This would cause a fully developed shockwave so that the full effects of any shockwave interactions might be captured. On each model the parameters with no relation to the viscous model were kept constant to ensure accuracy of comparison.

Figure III-3: 2D axis-symmetric grid used to compare turbulence models Good levels of convergence were produced for simulations using the Spalart-Allmaras, K-epsilon and Komega models, with all variables converging below 1e-04. However the simulation would not converge making use of the Reynolds Stress Model. Previous simulations used a courant number of 5, and even reducing the courant number to 0.05 would not remove the divergence. This was a disappointment as initial research indicated that this model was the most accurate and would be a likely baseline for comparison of other models against. The exact reason for this was not ascertained as the author felt that problem solving this particular issue would be not be practicable considering the potential time taken versus the likely benefits. Results With the failure to produce useful results from the Reynolds-stress model, there were only three models left for comparison. The Spalart-Allmaras, k-, and k- models were all applied to the same grid model, as seen in figure III-3. The software package was allowed to perform grid refinement to improve resolution and accuracy in the vicinity of the shockwave which formed in the restrictor. The models were compared on the basis of three parameters. The first factor considered was total number of iterations to convergence, and time per iteration. Overall computational time would be a factor in this project as there were limited opportunities to perform the simulations due to other demands on software licenses from other students and the number of simulations which needed to be run. Turbulence Model Number of Iteration time % Iterations difference Spalart-Allmaras 3593 57 mins +0% k- k- 5602 2hrs 01mins +112% 2.

6229 2hrs 12mins +131% Table III-2: Computational impact of the turbulence model selected This comparison showed that not only were the 2-equation models more intensive on a per iteration basis, but further that these more complicated models took longer to converge to a final solution. Clearly this would represent significant increases in computational time, particularly when extended to simulating the whole manifold which would potentially contain an order of magnitude more grid points. The next factor to be considered was a visual inspection of the flow simulations provided by FLUENT. The work reviewed in considering strengths and weaknesses of each turbulence model gave insight into the characteristics of the flow being simulated. The work by Knight and Degrez revealed that shockwave boundary layer interactions caused regions of separated flow downstream of the shock, so we should be expecting this in our simulations. The Spalart-Allmaras model produced significant regions of separated, recirculating flow extending 5-7mm from the wall. The k- model produced no separation, only a rapidly growing turbulent boundary layer downstream of the shock. The k- model produced a small region of separated recirculating flow extending at most 2.5mm from the wall. Knight and Degrez showed that the Spalart-Allmaras model produced the best correlation for shock-wave boundary layer interactions. Work conducted by Paciorri, Dieudonn, Degrez, Charbonnier, & Deconinck showed that cases where the Spalart-Allmaras model was

10

slightly lacking in agreement it tended to undersize regions of separation. These factors lead the author to feel that the k- and k- models were likely doing an inferior job of simulating the flow given the results obtained. The final consideration was to examine the velocity profile of the flow exiting the restrictor and make a comparison across each individual model, and to the expected shape of turbulent flow within a pipe. Figure III-4 shows the expected velocity distribution in a pipe for fully developed flow. However we are considering typical values lower than those for fully developed flows and as such would expect to see a central region of uniform velocity.

Figure III-4: effect of laminar and turbulent flow on velocity profile in a pipe The velocity profiles from each simulation were extracted from FLUENT and a comparison was made across all three to determine which model was producing results in line with expected data. The results were taken at the outlet of the restrictor; keeping in mind the minimum diameter of the flow is 20mm and the length of the restrictor is 250mm, giving an flow. =12.5, which is well below the expected 20-30 for fully developed

Figure III-5: Exit Velocity profile for the Spalart-Allmaras turbulence model

11

Figure III-6: Exit Velocity profile for the k-epsilon turbulence model

Figure III-7: Exit Velocity profile for the k-omega turbulence model Figures 3-5, 3-6 and 3-7 show that which turbulence model is being used greatly effects the results being produced. The k- model in figure III-6 appears to be simulating a fully developed flow, and the profile is analogous to a laminar velocity profile, more so than a turbulent one. The k- model shown in figure III-7 by contrast looks much more like a turbulent velocity profile, though once again it looks much closer to a fully developed profile. Finally if we examine the profile shown in 3-5 which was generated by the Spalart-Allmaras model we see a profile much more in line with the results expected. The simulation has a region of uniform flow in the centre, characteristic of flow which is not yet fully developed and we see a moderate region of recirculating flow towards the wall. E. SUMMARY Selecting an appropriate turbulence model is critical in producing accurate simulations using computational methods. The methods offered by FLUENT all have their strengths and weaknesses and a thorough evaluation of each was necessary to determine which was most suited to the application. Each consideration made, whether in reference to works carried out by other scholars or examining the simulated results for each model lead to the same conclusion that the Spalart-Allmaras model was most suited for the conditions simulated in the intake manifold.

12

IV.

SIMULATIONS

A. INTRODUCTION The restrictor is a very significant part of the intake system being modelled. The restrictor is the ultimate restriction on the amount of air which can flow into the intake system, and thus, the amount of power produced by the engine. Consequently this segment of the intake manifold is the logical place to commence simulations. A thorough understanding of the flow through this section will allow the author and the ACME Racing team to improve the design as much as possible, giving the best possible air flow into the plenum. B. RESTRICTOR Model The model used for the simulations was an axis-symmetric representation of the restrictor itself. FLUENT has the capacity to simulate a full 3-dimensional flow based upon a 2-dimensional section rotated around an axis. While technically a 2-dimensional simulation it provides an accurate representation of what occurs in 3-dimensions. The axis-symmetric model has an incredible advantage that it significantly increases the resolution of the grid. Below is a comparison between the 3D and 2D models, for the restrictor. 1.

Figure IV-1: 3D restrictor model 30 000 grid points.

Figure IV-2: 3D restrictor model - plane of symmetry.

13

Figure IV-3: 2D Axis-symmetric model 20 000 grid points. Comparing figures 4-2 and 4-3 show the exact difference between the resolutions of the two grids. The 3dimensional model has over 30 000 grid points. When a slice is taken through the centre of the model, it is clear that the grid is quite coarse, and each cell is plainly visible in the figure. In contrast the axis symmetric grid has only 20 000 grid points, however due to this being only in 2-dimensions we see that each cell is much smaller, and it is nearly impossible to make out each individual cell in the model. This clearly illustrates that the axissymmetric modelling capability will greatly enhance accuracy of the final solution by making use of a considerably more accurate mesh. Furthermore, the axis symmetric nature will reduce the computational power required to perform the simulations as only 2 dimensions of flow need be considered. 2. Validation Having chosen the model to conduct the simulations with, it was time to perform the simulations on the restrictor. However, with any numerical method, it is necessary to validate the results obtained through simulation. So before performing several simulations, and potentially finding them useless, a single simulation was conducted in order to validate the method and model to be utilised. There were no experimental results to validate the simulations against, but these were not necessary. Two factors were considered in validating the flow, firstly did the flow look as expected. The flow through a nozzle such as the restrictor is a well studied phenomenon, both theoretically and practically. We expect to see Mach 1 flow at the throat of the restrictor and a normal shock occurring somewhere along the diffuser section of the restrictor.

Figure IV-4: Contours of Mach number in the initial simulation. Figure IV-4 provides a visualisation of the flow through the restrictor; we see a flow speed of Mach 1 at the throat in line with expectations. We were expecting to see a normal shock form in the diffuser, however this expectation comes from the isentropic theory of flow through such a nozzle. The viscous effects of the flow explain why the shock takes on a curved appearance. On closer inspection of the shockwave we see that the shockwave is not taking on a sharp appearance as expected, figure IV-5 shows a zoomed in section of the flow, specifically around the shockwave. A shockwave is a near instantaneous drop in velocity from super-sonic to sub-sonic flow, the simulation conducted, and the visualisation in figure IV-5 indicates this transition is occurring over the space of 2-3mm. This suggested that the resolution around the shockwave was not sufficient to provide an accurate simulation.

14

Figure IV-5: Zoomed view of the shock wave in the initial simulation FLUENT provides the capability to dynamically adapt the grid around an area of interest. This area of interest is characterised by a sharp gradient of pressure or velocity, and we let the program refine the grid based upon areas of high gradient. The program was allowed to refine the grid by up to a factor of 4, that is 1 cell could be refined into as many as 16 smaller cells. This method obviously has the set back that it could get out of hand in the number of cells created by the program, consequently a limit of 150 000 cells was placed on this adaptation. Another simulation was run, this time allowing the grid adaption to occur and the results were examined for validation.

Figure IV-6: Contours of Mach number utilising grid refinement

15

Figure IV-7: Zoomed view of the shockwave simulated using grid refinement In figure IV-7 we can see a significantly sharper shockwave, much more in line with expectations. In comparing the flow from figure IV-5 and figure IV-7, we see the flow pattern is very similar between the two simulations. If we compare the graphs of velocity along the axis of symmetry we can see a numerical difference in how much more abruptly the flow goes from supersonic to subsonic.

Figure IV-8: Capturing the shockwave without grid refinement

16

Figure IV-9: Capturing the shockwave with grid refinement Figure IV-8 and figure IV-9 show the velocity along the centre axis of the restrictor, of interest is the area around the 60mm mark, where there is a sudden drop in the Mach number, this is the shock wave. In figure IV-8 we notice there is only one grid point in the middle of the shockwave, whereas figure IV-9 has five grid points. On initial inspection one might think this is a less accurate result, however in consider the size of the grid points between the refined and unrefined case we know that each grid point is one quarter the size in the refined case. Consequently, in the unrefined simulation the one data point between the drop off correlates to seven data points for a refined case. The next way to validate the flow being simulated is to consider the isentropic flow through such a nozzle. The derivation of the isentropic flow through a nozzle occurs in most fluid dynamics text books that cover compressible flows. The derivation is repeated in annex B. The final result of the derivation is the relationship for mass flow rate.

Where, is the mass flow rate is the total pressure is the total temperature is the cross sectional area at the throat ratio of specific heat is the universal gas constant This relationship now gives a maximum theoretically possible flow through the nozzle for isentropic, adiabatic, inviscid flow. The result of applying the values for the boundary condition used in the simulation will give an upper limit and a value to compare simulation results against. Model No Refinement Grid Refinement Isentropic Theory Mass Flow Rate 0.07358 kg/s 0.07338 kg/s 0.07413 kg/s Table IV-1: Comparison of mass flow rates through the restrictor using different models The difference in flow between the isentropic flow and the simulated flow is 1.02% and the simulated air mass flow rate is lower than the isentropic case, which is the maximum theoretically possible given the best possible situation. This number gives an upper limit to the amount of air that can be pulled through the restrictor. The next test was to consider how much air the engine is using in real world examples. To check this, the first step was to perform testing on a dynamometer to determine when the engine would choke. From here an approximation would be made as to how much air the engine was breathing in. A simple MATLAB code was 17

written to read the dyno results and produce a power vs. RPM graph for easy determining of when the engine choked and can be found at annex C. The output of the dyno testing can be seen below in figure IV-10.

Figure IV-10: Determining the choke point of the Engine Having determined a choking point of 9500 RPM an approximation could be made regarding how much air the engine would be breathing. This value would be then compared to simulated results. In order to obtain an approximation the pressure inside the manifold needed to be known, this was measured using a manifold air pressure (MAP) sensor. From here a MATLAB code was developed to conduct the calculations, the code can be viewed at annex D. The approximation based upon the dyno results was 0.0718 kg/s, which is 2.2% below the simulated results. During the run, there were fluctuations in the manifold pressure that made getting an exact manifold pressure difficult. Furthermore the MAP sensor was not properly calibrated, for a barometric pressure of 94.4 kPa, the MAP sensor gave a reading of 98 kPa. Given the variations that could be produced by these effects the 2.2% difference in results is very encouraging, and leads to the conclusion that the simulation is likely to be quite accurate. The result gives confidence to continue with the simulation of the restrictors and varying the geometry of the restrictor. Varying Diffuser Angle The first geometry we considered to vary was the diffuser angle, there are clearly compromises to be made with the variation of this geometry. As the angle increases, the likelihood of separation increases, however as the angle decreases the pipe becomes longer. Practical results show that pressure loss is a minimum for a diffuser angle of 4 degrees, these results however were obtained utilising water as the fluid flowing through the pipes. (Munson, Young, & Okiishi, 2006) In the case being simulated the fluid is air, and the presence of a shock is causing a significant region of separation which indicates that the value ideal diffuser angle is likely to be larger than this figure of 4 degrees. Work by Jawad, Hoste and Johnson of Lawrence Technological University suggested that 6 degrees was the optimum angle as it allowed the greatest volumetric flow rates, (Jawad, Hoste, & Johnson, 2001) however mass flow rate is the parameter which is of importance and not volumetric flow rate. Consequently values from 5 to 7 degrees, in increments of 0.5 degrees, were examined. Each simulation was conducted to the same order of convergence, all residuals were below 10-5 for convergence. This figure was selected as it represented less than 0.1% variation in values of mass flow rate, and was not overly onerous on computational resources to perform. It was important to ensure each simulation reached this level of convergence to allow a valid comparison between results. Pressure difference 5 degree 5.5 degree 6 degree 0.04 atm 0.06 333 0.06 022 0.05 836 Mass Flow Rate (kg/s) 0.07 0.1a atm tm 0.07 0.07 320 353 0.07 0.07 113 342 0.07 0.07 036 334 18 0.15 atm 0.07 344 0.07 339 0.07 336 0.2a tm 0.07 337 0.07 331 0.07 333 3.

6.5 degree 7 degree

0.05 0.06 0.07 0.07 744 875 329 333 0.05 0.06 0.07 0.07 459 701 329 327 Table IV-2: Effect on mass flow rate of varying diffuser angle

0.07 334 0.07 327

Figure IV-11: Mass flow rate through various restrictor designs varying diffuser angle 4. Varying Throat Radius The next geometry to be considered was the radius between the divergent and convergent nozzles making up the restrictor. The lower limit for this is clearly a 0mm radius, that is a sharp point, however this would cause a region of separation immediately downstream of this point, there is no upper limit, but as the radius increases, the overall length of the restrictor increases as does the length of the throat itself, this is clearly an undesirable situation so a radius of up to 50mm was considered in the simulations. Again the simulations were run until all residuals were below 10-5.

Pressure difference 10mm 20mm 30mm 50mm

0.0 4atm 0.0 5817 0.0 5836 0.0 5787 0.0 5419

0.0 7atm 0.0 6963 0.0 7036 0.0 706 0.0 6777

Mass Flow Rate 0.1 atm 0.0 7293 0.0 7334 0.0 7347 0.0 7361

0.1 5atm 0.0 7301 0.0 7336 0.0 7346 0.0 7348

0.2 atm 0.0 732 0.0 7333 0.0 7344 0.0 7336

Table IV-3: Effect on mass flow rate of varying throat radius

19

Figure IV-12: Mass flow rate through various restrictor designs varying throat radius C. ENTIRE MANIFOLD Model There is a substantial jump between simulating the restrictor alone and the entire manifold, in both complexity of the model and computational time required. As there is this added complexity it was desirable to simplify the simulation if this was achievable. In simulating the restrictor a two dimensional model was utilised by virtue of the axis-symmetric nature of the restrictor. The manifold however is not an axis symmetric-part and this feature cannot be used. FLUENT does provide a 2-dimensinal simulation feature that allows you to make use of a 2-dimensional model and considers it as a slice of an infinitely long prism. Clearly this could be useful if its accuracy could be validated. The first step taken was to go back and consider the restrictor model, granted the geometry is very simple, but it is a place to start and any glaring discrepancies would immediately rule out the method. A simulation was run making use of the 2-dimensional option and the results examined. The first test was a look at the velocity contours throughout the model. 1.

Figure IV-13: 2-D plane simulation

20

Figure IV-14: 2-D axis-symmetric simulation Figures IV-13 and IV-14 show many similar characteristics in the flow, the main difference being the shock wave which forms, its forming further downstream and consequently there is a large drop across the shock as the flow has accelerated more as it has expanded. A more subtle difference is the region of separation (the dark blue area downstream of the shock), which is smaller in the case of the 2-D plane simulation. The next step was to consider velocity profiles at the exits; these were taken directly from the FLUENT simulations.

Figure IV-15: 2-D Plane simulation outlet velocity profile

Figure IV-16: 2-D axis-symmetric simulation outlet velocity profile The velocity profiles continue to show a similar trend between the simulations, again they are not an identical match, with the 2-d plane simulation under simulating the region of separated flow near the wall with comparison to the axis-symmetric simulation. The average flow speed across the outlets were, 77m/s for the 2-D 21

plane simulation and 64m/s for the axis-symmetric model, this discrepancy seems to be large (20% difference) however if we consider the size of the separated region and that it is smaller on the 2-D plane simulation it becomes apparent that this is the major cause of the discrepancy and that overall there is a reasonable correlation between the velocity profiles of particular encouragement is the maximum velocities are very similar. Further consideration of other flow properties, including total pressures, and densities showed variations of less than 1% between the two simulations. Before continuing with further validations it was necessary to consider what the goals of simulating the manifold were. It was not to provide a quantitative analysis of the manifolds, but rather a qualitative analysis produce a comparison of each design. As such it was decided that the agreement between the two simulation methods was sufficient to apply to whole manifold simulation. WS03 The intake manifold employed on WS03 was designed by OFFCDT Laine Pearce in 2006 as a part of his engineering thesis. It was a compromised design when it became apparent that the variable geometry intake he had designed was not financially viable for the team. The design featured tuned intake runners to optimise torque for 9000RPM. However the focus of the work done by Laine Pearce was towards tuned intake runner lengths and although he recognised the important role of other geometries, there was no analysis conducted to determine the efficiency of his design. (Pearce, 2006) Below in figure IV-17 is a picture of the manifold design for WS03. 2.

Figure IV-17: WS03 Intake Manifold As mentioned in the previous section a 2-D model would be used to simulate the manifold. A section cut was made of the manifold and imported into the software package Gambit for grid generation. In order to properly mesh the model it was necessary to make use of a triangular grid, this is less desirable than a square grid as it creates many more grid points for the same resolution of mesh. However this was a necessity in order to mesh the complex shape of the manifold.

22

Figure IV-18: Grid model for the WS03 Intake Manifold. As can be seen in figure IV-18 this manifold has a plane of symmetry running through the axis of the restrictor, consequently only two cases needed to be considered, those being flow into the first and second cylinders. As discussed in the previous section of this chapter, a 2-D simulation such as the ones conducted on the entire manifold could not produce quantitative results, however are useful in examining the characteristics of the flow through each manifold, and a comparison can be made between different cases for each manifold, and between manifolds. Simulations were conducted on the above manifold for flow to each of cylinder 1 and 2, a pressure drop of 0.2atm was applied across the manifold. The simulations were conducted using a similar set-up to the restrictor simulations, the Spalart-Allmaras model was used to model the turbulence and the grid refinement function was enabled in order to capture the regions of significant flow with greater resolution. Residuals were relaxed to 10-4 as it was too computationally exhaustive to achieve convergence to 10-5, however this was not detrimental to the accuracy as the flow rates through the manifold were considerably higher due to the fact that mass flow rate is representative of a 1m section of an infinitely long model.

Figure IV-19: Flow through the WS03 manifold to cylinder 1

23

Figure IV-20: Flow through the WS03 manifold to cylinder 2 The resulting flow fields in figure IV-19 and IV-20 reveal the importance internal geometry has on the flow performance through the manifold. The extra bend required for the flow to travel into cylinder 1 significantly reduces the overall flow through the system. An examination of the scale of the flow field demonstrates that the flow achieves choked conditions for flow to cylinder 2, yet is substantially below that for flow to cylinder 1 this is due to the more significant bending of the flow required to flow into cylinder 1. A closer investigation of figure IV-19 reveals a significant region of separation on the left hand edge of cylinder 1. The separation is a phenomenon known as vena contracta and is caused by the fact that the flow cannot turn sharply enough to remain attached to the wall of the intake runner. Correcting this flaw in design is relatively easy through a modification to the entrance to the intake runners and will improve the characteristics of the flow propagating into the combustion chamber.

3.

WS04 The intake manifold employed on WS04 was designed by OFFCDT Michael Olsen in 2007. The design was a modification of the previous years design by Lane Pearce. Most of the design modifications were inspired by a desire to keep the packaging of the rear of the car, neat yet functional. While there were considerations on the effects of the new geometry, the advice offered to Michael Olsen in his design process were based on opinion only and no analysis was conducted to determine if these opinions were correct.

24

Figure IV-21: WS04 Intake Manifold The set-up for conducting the simulations on the model of the WS04 intake manifold was the same as for the set-up of the WS03 manifold, making use of the Spalart-Allmaras turbulence model combined with grid adaptation and a pressure drop of 0.2atm across the manifold. Once again simulations were run until all residuals were below 10-4.

Figure IV-22: Flow through the WS04 manifold to cylinder 1

25

Figure IV-23: Flow through the WS04 manifold to cylinder 2

Figure IV-24: Flow through the WS04 manifold to cylinder 3

Figure IV-25: Flow through the WS04 manifold to cylinder 4 The in the manifold employed on WS04 has a substantial bend in the piping downstream of the restrictor. This 180 degree bend is quite sharp and practical testing reveals this layout carries with it a substantial head 26

loss. The exact cause of this pressure loss can be clearly seen in the outside of this bend where we see a significant region of separation in the flow. As a consequence the potential to get air into the individual cylinders is already significantly diminished as a result of this pressure loss. If we examine the flow as it enters each intake runner, we again see that the flow has separated from the wall of the intake runner forming a vena contracta. Progressing to comparison between the flows to each of the four cylinders we notice that the separation is more severe close to the plenum entrance, where the flow needs to make the sharpest turn into the runner. The region of enlarged separation has implications upstream of the runners. If the flow is traced back to the throat of the restrictor, a comparison reveals that the more severe this separation the less air is flowing into the manifold. A comparison between WS03 and WS04 and the flow speeds achievable demonstrates that the worst case scenario for the WS03 manifold correlates to the best case of the WS04 manifold. In the limited dyno testing conducted on WS04 in November 2007 it was noted that the power output of the vehicle was noticeably lower than either WS02 or WS03. The cause of this deficit was not determined nor was tuning of the engine completed. This discrepancy identified by these simulations would go some distance to explaining the lack of power produced. D. SUMMARY This chapter has provided an in depth look at the way in which geometry of certain components of the intake manifold can affect the overall performance of the entire system. Beginning simply at first it was determined that both parameters varied, both the throat radius and the diffuser angle had a noticeable effect on the flow through the restrictor. It was found that the best geometry was a 5 degree diffuser angle in combination with a throat radius between 20mm and 30mm. The intent of the restrictor is to place a power restriction on the output of the engine, as a consequence maximising the flow through this section of the restrictor gives the potential for yielding the highest power output of the engine. Simulation of the entire manifolds of both WS03 and WS04 reveals more about the effects of the internal geometry. The first significant factor noticed is that bends in the flow have a marked effect on the amount of air flowing through the restrictor. Further inspection of results revealed that the entrance to the intake runners was also an important piece of geometry. The fine tuning of this section of the intake to reduce the area of separation of flow as it enters the runner is of high importance in reducing the losses of the intake manifold.

V.

DESIGN OF A NEW MANIFOLD

A. INTRODUCTION The investigations conducting using CFD in the earlier sections of this project have returned valuable results for the design of a new intake manifold for use by the ACME Racing team. Specifically the impacts on where the intake runners are being fed from and the effect of geometry on air flow through the restrictor. These outcomes provided a baseline for the overall intake manifold. CATIA provided a 3 dimensional computer aided design environment to produce the next generation intake manifold for ACME Racing. B. IMPACTS OF SIMULATIONS 1. Restrictor The first impact of simulation was the outcome that the restrictor needed to be longer with a more acute diffuser angle downstream of the throat. As determined by the outcomes of the simulations conducted in chapter IV a diffuser angle of 5 degrees produced the highest mass flow rates across all flow conditions considered. For this reason, the new restrictor design will incorporate a diffuser angle of 5 degrees. The other results of simulation on the restrictor showed that varying the radius at the throat also affected the mass flow rate. It was found that a radius between 20mm and 30mm provided the highest flow rate, consequently the new design features a 25mm throat radius. 2. Plenum and Intake Runners The simulations conducted in chapter IV provided insight into the flow occurring inside the plenum and around the entrances to the intake runners. A comparison of flows indicated that flow into the intake runners could be significantly enhanced in all cylinders by taking advantage of certain layouts. Firstly it was demonstrated that if the plenum was fed from the centre, rather than from one edge flow was significantly enhanced in all cylinders. Secondly it was shown that if the flow had to turn too much after a centre feed, like in the case of the cylinders at the extremities of the plenum, that the flow rate was significantly diminished. As a consequence the new design incorporates a distribution of the intake runners such that they all are equidistant

27

from the centreline of the plenum, and that distance is of a similar magnitude to the separation inside the centre feed manifold simulated. C. NEW FEATURES 3. Restrictor In previous intakes the restrictor had been a separate item to the plenum in order to keep size of individual parts down, which helps in manufacturing of the manifold, and also to allow easier finishing of the inner surface of the manifold assembly. The laser sintering process can leave quite a rough surface. Given the layout of this new design there was no requirement to have the restrictor detachable from the plenum as the restrictor and top half of the plenum could be manufactured as one piece and this one piece would still allow ample access for the post manufacture finishing. Although if at a later time it was deemed necessary, this would be a simple modification to the design. The new design can be seen below in figure V-1.

Figure V-1: Integrated restrictor and top half of plenum 4. Plenum With the intake runners taking a significant modification to their arrangement inside the manifold, there was a clear need to modify the plenum design. A bulbous shape was chosen as it allowed for the best route for the air to take to each individual runner without sacrificing extra plenum volume. The latter point is of important consideration as increasing plenum side has a detrimental effect on throttle response. The new design for the top half of the plenum can be seen above in figure V-1. Intake Runners The intake ports to the cylinder head of the engine are in an inline configuration, whereas the desired configuration inside the plenum is one where the runners are equidistant from the centreline. In order to achieve this, the runners would need to incorporate a bend so as to transform the circular arrangement into an inline one. The desired runner length had been previously determined by the ACME Racing team to be 200mm, this would give the characteristic power and torque curves desired. This length was obtained in the runners using an iterative process, the horizontal displacement was fixed and the angle of the bend adjusted, along with the vertical height of each runner to maintain the desired length. Of further consideration was ensuring sufficient space to allow installation of fuel injectors and fuel rail onto the manifold. Further, there are improvements to the entry to each intake runner, the trumpet shape is more pronounced in an effort to reduce separation on entry into the runner. This can be seen in figure V-2 below and compared to the cross section shown in figure IV-18. 5.

28

Figure V-2: Bottom half of plenum and intake runners

D. SUMMARY This chapter has outlined the design of a new intake manifold for the ACME Racing team. Each design influence has come either as a direct requirement of the team, or as a result of the research and simulations conducted in previous chapters of this thesis. The design takes advantage of a straight line passage for the air through the restrictor and into the plenum, which was shown in chapter IV to pose significant advantages. The intake runner arrangement shows an equidistant arrangement which is closer to the centreline of the manifold which was again shown to have advantages in chapter IV.

VI.

CONCLUSION AND RECOMMENDATIONS

A. INTRODUCTION This thesis formed part of the ACME Racing teams desire to produce a new car for the 2009 competition. The aims of this thesis were to gain an improved understanding of the flow through a restricted intake manifold, to produce a qualitative analysis between the intakes employed on WS03 and WS04 and finally to design a new intake manifold for use on the ACME Racings 2009 F-SAE entrant vehicle. The research performed was based upon, in part an extension of the work conducted by OFFCDT Laine Pearce in addition to knowledge acquired by the author through his involvement with ACME Racing. The ultimate motivation behind improvements to the intake system is to ensure that the team is gaining every performance advantage it can from the engine. The motivation behind this project is not only to deliver this advantage, whether in part or in full, but to enhance the teams understanding and thus increase the ability of the team to think critically and produce an effective component for future vehicles. B. CONCLUSIONS The simulations conducted in this thesis provided the bulk of the work load of the project, however they also returned the most benefits in terms of knowledge gained into the workings of an intake manifold. Simulation of the restrictor itself showed that the geometric characteristics of the restrictor, both diffuser angle and throat radius, had a noticeable effect on the maximum flow rates through the restrictor. It was determined that the best flow rate came from having a 5 degree diffuser angle and a 20mm-30mm throat radius. The difference in flow rates for some flow conditions could be as high as 16% just by varying the diffuser angle 2 degrees. These new ideal parameters gave definitive targets to be used in the design phase. When simulations moved to the entire manifold they again revealed considerable potential for improved flow rates. Simulation of 29

an end feeding manifold, as is employed on WS04 revealed that there is a considerable difference in the flow rate to each intake runner. The simulations also revealed how detrimental a significant bend in the intake can be, a comparison between manifolds used on WS03 and WS04 suggests that the flow losses could be as high as 15%. Simulations of the manifold used on WS03 shows that even if we consider a superior placement of the plenum feed, to the centre, there is still a significant loss occurring inside the manifold. The simulations revealed that in a log style manifold fed from the centre, flow to the outer most cylinders was as much as 16% lower than flow to the central cylinders. This deficit was due to a combination of the flow having to turn in a short distance and the separation occurring at the entrance to the intake runners. These results gave goals to achieve in the new design. With a strong set of results from the simulation phase it was time to implement them in the design of a new manifold. As mentioned, the ideal geometry for the restrictor had been determined and this was an easy task to feed them into CATIA to create a 3D model. The most difficult aspect of the new design was the intake runners. The desired configuration inside the manifold was a circular pattern for the runner entrances, however it was a constraint of the engine that the runners form an inline configuration where they joined to the cylinder head. For this reason bends were incorporated into the intake runners to allow this transformation. The plenum was designed around the intake runners and to incorporate the restrictor and had a bulbous shape. This new design incorporated all the design improvements suggested as a result of the simulations performed.

Figure VI-1: Complete intake manifold C. RECOMMENDATIONS There are finally some recommendations to be made for the continuing development of the ACME Racing F-SAE vehicle. This thesis aimed to fill some holes in the teams knowledge of intakes particularly relating to internal geometries. The work conducted in this thesis in combination with the work done by OFFCDT Laine Pearce provides a strong foundation for the continued development of the intake system for a FSAE vehicle. The author would like to see an amalgam of the two works to produce the best possible intake for a F-SAE vehicle. In terms of continued research the following recommendations are made:

30

1.

2.

Investigate the performance advantages available through the use of a turbocharger or supercharger. While this would not increase the maximum power due to the restrictor, it would increase the range at which max power is available. Complete CFD analysis of the intake manifold by producing a 3D model of the intake system which also simulates the pressure wave phenomena produced by the opening and closing of intake valves.

D. SUMMARY Although the intake will not be constructed at the time of the thesis submission, the aims of this thesis have been met. The simulations carried out revealed opportunities for significant improvements and these opportunities were carried through into design goals and implemented in the final design. It is hoped, with proper tuning, that the final design will deliver a measureable gain in the power and torque characteristics of the F-SAE Vehicle. This thesis has provided some valuable knowledge in the flow characteristics inside a restricted intake manifold. This knowledge along with the data gained by OCDT Lain Pearce will provide a stronger basis for the team to design future intake manifolds. It is hoped that this new design will help the ACME Racing team with their performance in 2009 and beyond.

Bredberg, J., Peng, S., & Davidson, L. (2002). An improved k- turbulence model applied to recirculating flows. International Journal of Heat and Fluid Flow vol23 , 731-743. Chaurette, J. (2008). Pump and Pump system glossary. Retrieved October 14, 2008, from Light My Pump: http://www.lightmypump.com/images/laminar-profile2.jpg Fluent Inc. (2005, April 1). Fluent 6.2 User's Guide. Retrieved September 22, 2008, from Old Dominion University: http://www.engres.odu.edu/Applications/fluent6.2/help/html/ug/main_pre.htm Knight, D., & Degrez, G. (1997). Shock wave/boundary layer interactions in high-Mach-number flows; a critical survey of current CFD prediction capabilities. Hypersonic experimental and computational capability, improvement and validation vol2 , 1-35. Kok, J. (2000). Resolving the Dependence on Freestream Values for the k- Turbulence Model. AIAA Journal Vol 38 No 7 , 1292-1295. Munson, B. R., Young, D. F., & Okiishi, T. H. (2006). Fundamentals of Fluid Mechanics. Hoboken: Jown Wiley & Sons. Paciorri, R., Dieudonn, W., Degrez, G., Charbonnier, J.-M., & Deconinck, H. (1997). Validation of the Spalart-Allmaras Turbulence Model for Application in Hypersonic Flows. AGARD AR-319 Vol 2 , 1-35. Poroseva, S., & Iaccarino, G. (2007). arXiv:physics/0701112 - Validation of a new k- model withthe pressure diffusion effects in separated flow. Cornell University. SAE International. (2008). Formula SAE Rules & Important Documents. Retrieved Apr 22, 2008, from SAE International: http://students.sae.org/competitions/formulaseries/rules/rules.pdf SAE International. (2008). SAE Collegiate Design Series: Formula SAE: About. Retrieved September 4, 2008, from SAE International: http://students.sae.org/competitions/formulaseries/about.htm Smith, P., & Morrison, J. (2002). Scientific Design of Exhaust and Intake Systems. Cambridge: Bentley Publishers. Vallet, I. (2008). Reynolds-Stress modelling of a M=2.25 shock-wave/turbulent boundary-layer interaction. International Journal for Numerical Methods in Fluids vol 56 issue 5 , 525-555. White, F. M. (2003). Fluid Mechanics fifth edition. New York: McGraw-Hill.

REFERENCES

31

Das könnte Ihnen auch gefallen