Aerodynamic investigations of ventilated brake disks

D. Parish and D. G. MacManus*

Department of Aerospace Sciences, Cranfield University, Bedfordshire, UK

*Corresponding author. Dept. of Aerospace Sciences, School of Engineering, Cranfield

University, Bedfordshire, MK43 0AL, UK.

Abstract:

The heat dissipation and performance of a ventilated brake disk strongly depends

on the aerodynamic characteristics of the flow through the rotor passages. The aim of

this investigation was to provide an improved understanding of ventilated brake rotor

flow phenomena, with a view to improving heat dissipation, as well as providing a

measurement dataset for validation of computational fluid dynamics (CFD) methods.

The flow fields at the exit of four different brake rotor geometries, rotated in free air,

were measured using a five-hole pressure probe and a hot-wire anemometry system.

The principal measurements were taken using two-component hot-wire techniques and

were used to determine mean and unsteady flow characteristics at the exit of the brake

rotors. Using phase-locked data processing, it was possible to reveal the spatial and

temporal flow variation within individual rotor passages. The effects of disk geometry

and rotational speed on the mean flow, passage turbulence intensity and massflow

were determined. The rotor exit jet and wake flow were clearly observed as

characterized by the passage geometry as well as definite regions of high and low

turbulence. The aerodynamic flow characteristics were found to be reasonably

independent of rotational speed, but highly dependent upon rotor geometry.

Keywords: Brake disk, ventilated, aerodynamics, heat transfer, unsteady flow,

turbulence.

1
Nomenclature

b Brake rotor passage width, m

m Mass flow, kg/s

m* Non - dimensiona l mass flow, m U 2 A2

r Radial distance, m

y Circumferential distance, m

z Axial distance, m
2
A Brake rotor area, m

B Brake rotor total width, m

BPF Blade passing (frequency), Hz

C Relative velocity, m/s

D Brake rotor diameter, m

NB Number of blades

Red Reynolds number, U2D2/

Tu Turbulence intensity, %

U Blade speed, m/s

V Absolute velocity, m/s

Absolute flow angle, degrees

Relative flow angle, degrees

Flow pitch angle, degrees

3
r Density, kg/ m
2
n Viscosity, m /s

w Rotational speed, rad/s

Subscripts

1 Brake rotor inlet

2 Brake rotor outlet

r radial direction

2
1 INTRODUCTION

Automotive brakes are used to convert kinetic energy into thermal energy through

friction between the brake pads and the rotor faces. If the brake disk temperatures

become too high then the structural integrity will be compromised and the thermal load

may lead to deformation, judder and increased wear. Although conventionally the

cooling performance of the disk has been of less importance than other engineering

aspects such as structural and manufacturing issues, improved heat dissipation will

enable lighter, more reliable designs. A common method of enhancing the disk cooling

is by using a ventilated brake rotor which improves the convective cooling by using air

passages separating the braking surfaces. The rotor behaves as a centrifugal fan,

drawing cool air from the inboard side, passing through the rotor passages and

exhausting at the outer diameter. Ventilated brake rotor heat dissipation depends

strongly on the aerodynamic behaviour of the flow throu gh the rotor passages. All

three mechanisms of heat dissipation contribute to the disk cooling, but it is anticipated

that convection has a large impact for most ventilated rotor configurations. Conduction

is frequently kept to a minimum in the design process to protect other components and

the radiation terms only play a significant role at high temperatures [1].

The aerothermodynamics of a ventilated brake rotor is complex and highly

dependent on the geometry of both the brake rotor and its surrounding environment.

Although the aerodynamics of conventional centrifugal impellers has been extensively

studied from a pumping performance and efficiency point of view [2, 3, 4, 5, 6], there

has been very little work done on understanding the unusual aerodynamic design

objectives for an automotive brake rotor where the dominant requirement is for cooling

performance and thermal management. In addition, the brake rotor does not have the

familiar turbomachinery inlet or outlet surrounding geometries and the operating point

encompasses a broad range of rotating speeds. Clearly it is a field where the

established turbomachinery knowledge base has a role to play but requires re -

evaluation for the current applications.

3
 Previous work has addressed both aerodynamic [1,7,8,9] and heat transfer

[1,10,11], aspects for an assortment of ventilated and solid disks. Most previous

workers measured the rotor exit airflow using pressure probes to measure the velocity

profile and to calculate the rotor massflow. Barigozzi [7] is one of the few previous

workers who used hot -wire anemometry (HWA) to examine the unsteady rotor exit flow

field for two disk geometries featuring backward curved vanes and a pedestal

arrangement. He showed that the pedestal configuration increased both the non -

dimensional massflow as well as the reported turbulence intensity. Johnson [9]

examined the flow field both surrounding the brake disc as well as inside the radial

rotor passages using a two -component PIV system and, as anticipated, revealed large

areas of separated flows within the passages. Other workers have shown that both

local geometry modifications to the rotor inlet [8] and cross-drilled holes [11] can have

a beneficial effect on the cooling performance. Computational studies have also been

performed to examine the entire wheel arch area [12, 13] as well as the brake rotor

itself [1,15,16]. Voller [18] performed CFD analyses which showed good agreement

between the experimental and predicted heat transfer coefficients for a range of

rotational speeds and disk temperatures. Some workers have attempted to rank the

relative importance of the mode of heat transfer and Limpert [11] suggests that internal

convective cooling of ventilated rotors can contribute as much as 50 60% of total

cooling at high rotor speeds. Conduction is limited by design to control the temperature

of surrounding components and the cross-sectional area of the rotor is designed to

prevent excess heat conduction to the wheel and hub. While attempts have been

made to improve rotor-cooling performance, it is clear that the heat transfer

phenomena and aerodynamic flow characteristics of the basic rotor geometry are not

well understood. This is illustrated by the wide variety of rotor design improvements

suggested by some of the work reported above.

The aim of this investigation is to provide an improved understanding of ventilated

brake rotor flow features as well as to establish a CFD validation database. The flow at

the exit of four different brake rotor geometries was measured using a five-hole

4
pressure probe and a hot-wire anemometer to determine mean and unsteady flow

characteristics. The measurement of the unsteady velocity field using the hot-wire

enabled phase-locked ensemble averaging to provide passage-to-passage velocity

and turbulence intensity distributions at the exit of the brake rotors. This facilitates the

identification of detailed flow features and comparisons between the different rotor

designs and geometries. While it is appreciated that a brake rotor operating in free air

is far from the true configuration, it was intended that this investigation would provide

the first step in fully understanding the broader issues of brake rotor aerodynamics and

cooling as well as providing a useful CFD validation database. In all of these

experiments, no braking or thermal load was applied to the brake disk. Measurements

were taken at a variety of radial distances from the rotor exit plane and at three

rotational speeds (750, 1500 and 2250 rpm). The Reynolds number, based on the
5 6
outside diameter and rotational speed, Red, ranged from 3.610 to 1.110 .

2 EXPERIMENTAL APPARATUS

2.1 Test Pieces

The brake rotors were rotated using a mechanical lathe at up to 2250rpm and

without forced convection in still, ambient air at room pressure and temperature. Four

brake rotor geometries were tested as summarized in Table 1. Rotors 1, 2 and 4

featured backward curved vanes with a variety of diameters, passage numbers,

passage widths and aspect ratios. The blade geometries are circular arc with squared -

off leading and trailing edges. Rotor 3 is a different style of configuration which

features internal pedestals and partial vanes. Full details of the geometry cannot be

provided due to commercial confidentiality reasons.

2.2 Measurement Systems

The miniature five -hole pressure probe had an outside diameter of 1.5mm and was

used to determine the mean flow properties such as yaw and pitch angle as well as

mean velocity and massflow. It was primarily used as a preliminary measurement tool

and also provided interesting comparisons with the hot-wire measurements. A constant

5
temperature hot -wire anemometry system (Dantec Streamline 90N10) was used to

measure magnitude and direction of the velocity vector in the yaw plane, Figure 1. The

Streamware system processes the raw voltage into velocity and decomposes the

measured velocities into probe co-ordinates, using the manufacturers directional

calibration coefficients. A Dantec 55P61 probe was used which has a tungsten wire of

5mm diameter with a low thermal inertia. Following the five-hole pressure probe

surveys, the hot-wire was orientated so that the primary flow direction was set at

approximately 45 to both sensors. Although the velocity components were

simultaneously measured, the wires are off-set by about 1mm so that the

measurements are not co-incident in space. An optical probe was used to provide a

once-per-revolution which enabled phase locked data to be acquired.

To enable the results taken in the fixed frame of reference to be viewed

appropriately in the rotating frame of reference, the measurement system must deliver

sufficient data resolution per rotor vane passing. The hot-wire amplifier filter and gain

were optimised to give an overall frequency response of 40kHz when tested in a

representative airflow. Digital recording of the analogue signal provided by the hot-wire

anemometer was performed using a National Instruments A T-MIO -16E-10 analogue to

digital (ATD) converter, with a total sample rate of up to 100kS/s. A total of four input

channels on the ATD card were occupied by the once-per-revolution optical sensor,

the anemometer temperature probe and a separate channel for each of the two hot -

wire signals. The sampling rate was set to 20kS/s per channel which established the

measurement spatial resolution in the rotor frame of reference. The highest expected

blade passing frequency, and corresponding lowest resolution, was for the 48 bladed

rotor at 2250 rpm which equates to 1800Hz. Table 2 outlines the achieved points per

rotor passage for each disk and rotational speed. With the sampling rate fixed at

20kHz, the Nyquist criterion dictates that the maximum expected resolved frequency is

10kHz. Although the achieved sampling rate was modest, the spatial resolution in the

rotating frame of reference was still sufficient to discern the principal flow features.

6
 2.3 Hot-wire calibration and uncertainty

The hot-wire calibration process requires the establishment of a relationship

between the measured voltages and the flow velocity and direction. For this test the

manufacturers recommended directional calibration coefficients were deemed

sufficient. This could be improved in further investigations by performing a full bespoke

directional calibration for each individual probe. The velocity calibration was carried out

using a dedicated calibration jet which delivered filtered, compressed air through a

settling chamber to provide a low turbulence, flat velocity profile jet. The hot -wire probe

was positioned one jet diameter downstream of the nozzle exit plane and a calibration

was performed over the expected velocity range. During the experiments the

temperature was simultaneously recorded and a correction applied if the temperature

difference between calibration and experiment was greater than 5C.

As with all measurement techniques the final error depends on the chain of systems

each of which will have its own systematic and random type errors. It is difficult to

provide a comprehensive error assessment although this type of hot-wire

measurement technique is well established. A brief summary of the main error sources

is provided for completeness. The calibration jet velocity was determined from a digital

manometer which was estimated to be accurate within 2% over the range of interest.

The hot-wire probe was positioned with the wires at 45 to the calibration jet by

measurement. This was estimated to have an error margin of 1. Curve fitting of the

resulting calibration data yielded results to within 0.7%. Jorgensen [17] suggests that

commercially available anemometers have low drift, low noise and good repeatability,

which does not add significantly to the experimental uncertainty. Probe initial

positioning relative to the brake disk was estimated to have an uncertainty of 0.2mm.

Traversing was carried out using the lathe digital measuring system, which had an

accuracy of 0.01mm. Temperature variations were corrected by using the

anemometers built-in temperature correction facility and are not thought to have

contributed significantly to the uncertainty.

7
3 MEASUREMENTS

3.1 Introduction

Given that the brake rotor acts as an impeller, there is much to be gleaned from the

broad knowledge base of turbomachinery aerodynamics when examining the current

flow fields. A conventional impeller flow field is very complex and has been explored

both experimentally and computationally by a variety of workers [ 2, 4, 19, 20, 21]. The

flow is usually highly three -dimensional and unsteady with regions of separated flow at

some operating points. In addition, it is susceptible to secondary flow patterns due to

Coriolis and gyroscopic forces as well as the flows driven by radial equilibrium and

cross-passage pressure gradient terms [3]. One of the dominant features for an

impeller flow field is the jet and wake flow pattern as well as the possibility of large flow

separations leading to significant flow blockage and non-uniform mass distributions.

Pressure measurements were taken using a five-hole probe along axial traverses at

a range of radial positions at the rotor exit (Figure 1) for all four rotors. Hot-wire axial

traverses were performed at a fixed circumferential location and at the closest radial

position, 2mm from the exit plane, for Rotors 2, 3 and 4. The measurement domain

extended beyond the plane of the rotor by about 6mm on both the inboard and

outboard sides and data was taken at 1.0mm axial intervals which typically gave 30

points across the domain. The hot-wire measurements were taken over 100

revolutions and ensemble averaged based on the once-per-revolution (OPR) signal

from the optical probe. The measurements provided mean and unsteady data for the

radial and circumferential velocities in the absolute frame of reference. Relative

velocity and whirl angle were subsequently determined by taking the rotational velocity

into account.

3.2 Spectral analysis

The frequency power spectrum was calculated for each of the measurement sets,

based on the rotor relative velocity. The highest blade passing frequency was 1800Hz

(48 blades at 2250rpm), although there was no significant measured frequency content

8
above 5000Hz for all of the rotors. This gave confidence that the measurement system

had the capability to measure primary details of the flow field. For all rotors, the

fundamental blade passing frequency (BPF) was clearly observed and for some

configurations up to the sixth harmonic of the BPF was also measured (Figure 2). Just

as the fundamental BPF reflects the passing of a sinusoidal jet-wake perturbation, then

the higher harmonics represent higher spatial frequency within the rotating blade

passage. Figure 2 shows a high harmonic content because it has the lowest number of

passages which means that the rotor exit flow field has a greater spacing between the

wakes, is less of a simple sinusoid in shape and therefore contains higher spatial

frequency. For most of the rotors, the exit flow field is dominated by the jet-wake flow

pattern and, as expected, the spectral power decreases strongly for these higher

harmonics. Figure 2 also shows that Rotor 4 exhibits a sub-harmonic unsteadiness at

BPF (300Hz). This is due to the fixing arrangement of the impeller where the inlet to

every second passage is partially obscured by a bolt head. Similar effects were

observed on the other rotors where every fourth passage was partially obstructed and

the spectral frequency showed a peak at BPF.

3.3 Circumferentially Averaged Results

Circumferential averages for the velocities and flow angles were calculated from the

data measured over 100 revolutions which highlighted some basic features of the

aerodynamics. Overall, the rotor measurements (both five-hole probe and hot-wire)

showed that, the exit flow angles, both relative and absolute, are predominately

insensitive to rotational speed. In addition, the absolute and relative velocities primarily

vary linearly with rotational speed and the velocity and massflow axial distributions do

not significantly change across the operating range of 750 to 2250 rpm. This highlights

the point that over most of the measured rotational speed range, the aerodynamic flow

features of the rotor are unchanging. For example, Figure 3(a)-(d), shows the absolute

angle (a), relative angle angle (b),non-dimensional relative velocity and radial velocity

for Rotor 2 for rotational speeds of 750, 1500 and 2250 rpm. There are only minor

changes in the flow between 1500 and 2250 rpm and the largest differences occur at

9
the lowest speed. This is evident in the relative flow angle and relative velocity, Figure

3(b)-(c). It also shows that at the lowest speed the brake rotor has a slightly higher

non-dimensional velocity and provides turns the flow through a greater angle in the

relative frame of reference i.e. lower exit relative flow angle b2. The radial velocity is a

direct measure of the rotor massflow and shows the distribution across the passage

span, Figure 3(d). There is a small decrease in the non-dimensional radial velocity at

the lowest rotational speed which is consistent with the increased relative velocity

exiting at a lower relative flow angle. The massflow peak is consistently located

towards the outboard side of the disk. This is probably due to a separation at the inlet

where the flow has to turn 90 degrees from axial to radial across a sharp cornered

geometry. The maximum non-dimensional radial velocity (and massflow) for this case

is achieved at 1500 rpm although the axial distribution does not change significantly.

Figure 1 includes an illustrative schematic of the mid-span velocity triangles at the rotor

exit.

Comparisons of the circumferentially averaged exit flows for the different Rotors (2,

3 and 4) are presented in Figure 4(a)-(e). The measurements were taken 2mm from

the rotor exit plane at 1500rpm. Overall, the trends and levels are reasonably similar

for the different designs which is not too surprising given the common dimensions of

diameter, passage height and blade exit angle (Table 1). Th e main difference in

geometry is the different number of passages (Rotor 2 and Rotor 4) as well as the part-

span vanes and pedestals of Rotor 3. One of the most notable features is the improved

uniformity in the inboard region (z/b = 0.5-1.0) for Rotor 3 in terms of the relative angle

and non-dimensional velocity (Figure 4(b), (d)). This may be due to the more

complicated and irregular geometry of Rotor 3 preventing the formation of a large

separation region and the pedestals encouraging mixing of the flow and redistribution

of the massflow. Figure 4(e) compares the radial velocity distributions for each of the

rotors and highlights the more uniform massflow distribution delivered by Rotor 3 -

particularly in the inboard region which is expected to improve the cooling performance

on the inboard face.

10
 3.4 Rotor massflow

An important parameter when considering the cooling performance of a brake disc

is the mass flow rate which the impeller draws through the rotor passages. The rotor

average massflow was calculated from the radial velocity component and was non -

dimensionalised based on the exit rotation speed, U2, a reference density and a total

geometric passage exit area (m* = m U2 A2 ). Overall, the non-dimensional massflow

flow varies only slightly with rotational speed indicating that the rotor aerodynamics do

not radically change at the different operating points (Figure 5). This is in agreement

with previous workers who showed a linear trend of massflow with rotational speed

[1,8,10]. However, there is a modest deviation in Rotor 2 which shows a maximum

non-dimensional massflow at 1500rpm followed by a very small decrease at 2250rpm.

Another significant aspect is the comparison between the different rotor designs which

show that, relative to Rotor 2, Rotors 3 and 4 deliver up to 25% and 45% more

massflow, respectively. All three rotors have essentially the same diameters, passage

width and vane thickness (Table 1) and the only significant changes are in the number

and design of the passages. Comparing Rotors 2 and 4 shows the effect of reducing

the number of rotor blades from 48 (Rotor 2) to 24 (Rotor 4). The large change in

massflow could be due to a range of effects but the primary influences are expected to

be due to the change in passage aspect ratio and the reduction in inlet geometric

blockage. The increased passage area means that for the same blade length the

boundary layers and wake regions are, to a first order, proportionally smaller. In

addition, the reduced number of leading edges at the inlet reduces the geometric

blockage as both rotors have the same approximate vane thickness. Although

increasing the massflow increases the potential cooling available to the rotor, it does

not necessarily mean that the rotor cooling is improved as its benefits are offset by the

reduction in surface area, and the overall perform ance will also depend on the detailed

massflow distribution within the passage, turbulence intensity as well as other

mechanical constraints. Although Rotor 3 has similar overall dimensions to Rotors 2

and 4, the geometry of the vanes is very different with nominally 36 passages which

11
include part-span turning vanes as well as mid-passage pedestals. Nevertheless,

Rotor 3 delivers more massflow than Rotor 2 and at 1500rpm is only 5% below the

Rotor 4 peak massflow. On this basis it appears that Rotor 2 is over bladed which is

restricting the massflow and that the complex geometry of Rotor 3 has a beneficial

effect on both the massflow level and axial distribution. This will be referred to later on

when the effect of the Rotor 3 geometry on the turbulence char acteristics will also be

examined.

3.5 Comparison between hot-wire and five -hole probe measurements.

The pneumatic pressure probe used in this, and most previous experiments, has a

very low frequency response. The probe measures quasi-steady changes in total

pressure and was calibrated for yaw, pitch and magnitude in a reference known flow

and velocity was determined by assuming that the exit flow was at atmospheric static

pressure. As discussed in 2.3, the hot-wire velocity magnitude calibration was

performed using a calibration jet and the manufacturers geometric transformation to

resolve into the required velocity components. Overall, the measurement systems were

calibrated completely independently of each other.

Figure 6(a) compares the absolute exit yaw angle measured by the five -hole probe

with the circumferentially average hot-wire results for Rotor 2 at 1500rpm. The results

are in very good agreement with the largest difference of about 3. However, a

comparison between the measured absolute velocity shows significant differences

between the pressure probe and the hot -wire (Figure 6(b)), where on average the

magnitude of the hot -wire results are about 75% of the pressure based measurements.

This could be due to a wide range of issues such as calibration differences or

processing errors. However, there are a variety of aspects worth considering which

could have resulted in an artificially high mean velocity for the pressure based data.

The miniature 5-hole probe had a diameter of 1.5mm relative to the passage height of

19.5mm and although the jet is unbounded, there may be a local blockage effect which

would result in an increased velocity reading from the pressure probe. Also, although

12
the pressure probe was only 2mm from the disc edge, it was assumed that the flow

was at atmospheric static pressure. Another contributing factor may be due to the

highly unsteady nature of the flow field due to both instantaneous turbulence as well as

deterministic wake-jet passing events. The five-hole probe has a very low frequency

response and it is known [22] that high levels of flow unsteadiness can result in

artificially high readings of total pressure and consequently velocity. This argument is

supported by the better agreement which is obtained in the inboard and outboard

regions where both the 'turbulence' and deterministic unsteadiness is very low. Also,

the good agreement in the yaw angle results is due to only small angle deviations

between the jet and the wake as will be presented later on from the hot-wire

measurements. Many previous workers have used five-hole probes as a way of

measuring the rotor massflow but these results indicate that such methods may be

unreliable and prone to over-estimating the massflow.

3.6 Frame of reference

An ensemble averaged flow field was constructed from 100 revolutions which

enabled the average variation of absolute velocity and whirl angle in each separate

blade passage to be examined. For example, Figure 7(a) shows the regular, periodic

variation of absolute velocity over six individual blade passages for Rotor 2 at

2250rpm. A better understanding of the individual passage aerodynamics is obtained

by transforming the results into the relative frame of reference. Figure 7(b) is the

corresponding relative velocity distribution which clearly shows the jet and wake flow

structure associated with each rotor passage. It is interesting to note that due to the

high rotational speed relative to the flow speeds, the regions of high absolute velocity

correspond to the rotating wake regions where there are low relative velocities. This is

illustrated in Figure 8 which shows a schematic of the mid-passage velocity triangles

for both jet and wake regions. The transformation into the relative frame of reference

also accentuates the passage flow structure. Figure 7(c) highlights the modest

changes in relative flow angle between the mainstream and wake regions. However,

13
the combination of changing relative angle and relative velocity results in substantial

changes in the absolute flow angle (Figure 8).

3.7 Ensemble averaged passage results

Processing of the unsteady measurements into the relative frame of reference

enables interpretation of the ensemble averaged rotor flow field and comparisons to be

made between the different brake disk designs. Figure 9 shows the relative velocity

distributions for Rotors 2, 3 and 4 along with schematics of the rotor geometries for

Rotor 2 and 4. For all rotors, the basic jet and wake structure is clearly identifiable

along with some significant differences between the rotors. From Figure 9(a)-(c) it is

apparent that, as expected, the high velocity jet or mainstream flow is closer to the

pressure side (PS) of the rotor passage and that the low-velocity wake region is

towards the suction side (SS). In addition, the overall flow is skewed towards the

outboard side of the passages for all the rotors. This was previously noted in 3.4 and

is probably due to a separation at the inlet where the flow negotiates a 90 degree turn

from axial to radial. For Rotor 2 (Figure 9(a)) the wakes and jets are curved away from

the direction of rotation (as presented the disk rotation is from right to left ). This is

probably due to the secondary flows and asymmetric boundary layer growth as

mentioned in 3.1. The results also show the additional blockage in terms of lower

peak velocity for the passages where the inlet is obstructed by the fixing bolt. This

confirms the spectral plots presented in 3.2. This also applies to Rotor 3 and Rotor 4

where the main effect is due to the balancing clip as shown in the geometry

schematic,Figure 9(b)-(c). Comparing Figure 9(a) and (c) highlights the effect of

changing the number of rotor vanes from 48 to 24. The mainstream jet portion of the

flow in Figure 9(c) is much larger than Figure 9(a) and although the wakes have about

the same velocity deficit, they encompass a much smaller proportion of the flow field.

This reflects the increased massflow rate for Rotor 4 as presented in Figure 5.

The flow field for Rotor 3 (Figure 9(b)) is characteristically different from the other

two configurations. The wake regions are proportionally bigger and exhibit a different

14
shape. This is due to the presence of part-span pedestals which are generating

additional wake flows near both endwalls. Also in evidence is the region where the

mainstream jet narrows as it passes through the mid-height channel created by the

partial pedestal on each endwall. This accounts for the change in shape of wake,

which now occupies a more significant portion of the flow in the important endwall

regions. As with the more conventional Rotor 2, Figure 9(a), the effect of the blockage

due to the fixing bolt heads is seen in passages 1 and 4 (y/b 2 and 7).

Overall, there are only modest changes in the relative yaw angle and for the datum

Rotor 2, at the mid-span location, the angle varied from 28 close to the wake region

and approximately 24 in the jet region. This reflects the underturning (high b2)

experienced by the low-momentum regions. The largest gradients of yaw angle are

between the mid-passage regions and the endwall flows where the relative exit angles

are very low. In these regions the relative velocity is reasonably high but there is very

little radial throughflow and the relative flow angles are therefore low. The relative whirl

angles in the jet regions are reasonably close to the idealized value with very little slip

between the flow and the rotor metal exit angle. The radial velocity contours for each of

the rotors, Figure 10(a)-(c), reiterate the primary observations such as the skewed

massflow distribution towards the outboard side and the localised blockage effects of

the mounting bolts and balancing clips. This is particularly evident in Figure 10(b)

which shows the massflow reduction due to the bolt in every fourth passage (y/b 2

and 7).

3.8 Turbulence characteristics

It is well established that the use of flow turbulators (pins, pedestals, ribs etc) can

significantly enhance the heat transfer for internal passages (23). This type of

technology has been particularly applied to gas turbines where the average surface

heat transfer coefficients can be increased by up to a factor of 2.8. For turbine

applications this typically increased the required pumping power by up to a factor of 10

15
(23,24). Clearly, this type of technology could play an important role in enhancing the

brake disk cooling performance and some designs already include pedestal

geometries. Although this experiment does not measure the heat transfer coefficients

or the detailed internal flow regimes, an examination of the unsteady exit flow field may

prove instructive. From the 100 revolutions of recorded unsteady data the ensemble

averaged relative velocities, C', corresponding to each rotor circumferential location

was determined. This averaged periodic flow field was subtracted from each revolution

dataset to provide a flow unsteadiness term, C'', at each location for the individual rotor

passages. The root-mean-square of C'' was used to calculate a turbulence intensity

term, Tu, as follows:

C' ' 2
Tu % = 100
C'

Even though referred to as turbulence intensity, this term is more accurately

described as a non-deterministic unsteadiness and quantifies the flow variations which

are not at blade passing frequency. Although the sample size of 100 revolutions is

modest for determining turbulence data, the results were consistent and repeatable

which gave some confidence in the presented measurements. To resolve the unsteady

data into the rotating frame of reference, it is vital that the velocities were measured

simultaneously although for this two-component hot-wire, this does mean that there is

a spatial offset between the measured components due to the wire construction of

1mm. Nevertheless, the results obtained are instructive and provide useful data for

comparing the characteristics of the different rotor designs. The term turbulence is

used here as a matter of convenience although it is sometimes interchanged with the

term unsteadiness.

Figure 11(a)-(c) shows the measured turbulence intensity, Tu, distributions for

Rotor 2, 3 and 4. When compared with Figure 9(a)-(c), it is clear that the regions of

high turbulence coincide with the low relative velocity wake regions. The mainstream

16
jet regions of the passage flows exhibit much lower turbulence levels. This is as

expected and has been observed in both conventional turbomachinery flows [25,26] as

well as other brake disk investigations [7]. The wake regions predominately com prise

of the vane and endwall turbulent boundary layers which are augmented by vortex

shedding, wake oscillations and unsteadiness of the separation points. Previous

studies have also revealed a very unsteady flow field at the impeller exit [9]. In this

example, Figure 11(a) shows that the wake flow on Rotor 2 has a peak turbulence

intensity of about 20% whereas the mainstream jet has a very low level of less than

2%. In addition, it also shows that the fi xing bolt head has increased the passage

mainstream turbulence distribution (e.g. y/b 2 and 7). The effect of reducing the

number of blades from 48 (Rotor 2) to 24 (Rotor 4) has no significant impact on the

general turbulence characteristics, Figure 11(a) and (c). The wake regions have a

slightly higher peak turbulence levels (24%) and the mainstream flows are slightly

lower (1-2%). The higher peak turbulence level relative to Rotor 2 may be due to

increased endwall driven secondary flows within the rotor passage as the number of

vanes has been reduced from 48 to 24. Similar to the effect of the fixing bolt head on

Rotor 2, the balancing clip on Rotor 4 (see Figure 9) increases the average

mainstream turbulence level for the locally affected passages (y/b 6 -8) to about 5-

8%. These effects are as expected and simply highlight the additional wakes and

disturbances introduced into the flow.

Although Rotors 2 and 4 are reasonably similar, Rotor 3 is characteristically

different, Figure 11(b). It was specifically designed with the intention of improving the

brake cooling performance by increasing the turbulence through the introduction of

part-span vanes and pedestals. Figure 11(b) shows that although the peak turbulence

levels are similar to Rotors 2 and 4, the extent and distribution of the high turbulence

regions have increased significantly. This is similar to the changes in the mean relative

flow field presented in Figure 9. An additional important change is the increase in

turbulence levels in the regions closest to the passage endwalls where the heat

transfer is required the most. Comparing Figure 11(a) and (b), shows that for Rotor 3,

17
the endwall turbulence levels are higher - particularly on the inboard face (z/b = 1)

where there is about a 50% increase. Figure 12 shows the distribution of the

circumferential averages of the turbulence intensity for Rotors 2, 3 and 4. The traverse

extends beyond the disk width and the outer faces of the endwalls correspond to z/b =

0 and 1, respectively. The plot clearly shows the increased turbulence intensity of

Rotor 3, reaching on average 4% turbulence points greater than Rotor 2. The

distribution highlights the average increase in the turbulence intensity for Rotor 3 at the

inboard passage wall (z/b=1), indicating the potential for improved heat transfer. The

distribution of Rotor 3 also shows a double peak, which is likely to be due to the low

turbulence jet being forced into the centre of the passage by the part-span pedestals.

The passage-t o-passage distributions of flow and turbulence through the datum

brake rotor passages is far from uniform, which will lead to high and low temperature

regions during vehicle braking. This is undesirable as thermal stresses are created

within the material. The turbulence of the flow through Rotor 3 was shown to increase

relative to the standard bladed rotors, indicating that heat transfer is likely to be greater

in this case. This is an important result, as the rotor geometry increases overall

turbulence intensity, without significantly reducing the mass flow or relative velocity.

The surface area of the internal passage is also increased due to the pedestal

arrangement, which suggests that this brake rotor may provide increased heat

dissipation. Overall, Rotor 3 is delivering a massflow and flow field characteristics

which are expected to improve the convective cooling of the brake disk. This is

supported by unreported tests which have shown that, under a known braking load,

Rotor 3 performs better than the other designs and has lower peak temperatures.

4 CONCLUSIONS

The unsteady flow fields of a variety of ventilated brake rotor disks were measured

using a two-component hot-wire anemometer to determine the mean and unsteady

velocity and whirl angle distributions at the rotor exit plane of a brake disk impeller. The

measurements revealed the detailed flow structures, and enabled the exit flow fields of

18
individual blade passages to be examined. The time-averaged measurements show

that the rotor aerodynamic behaviour, as reflected in the exit angle and non -

dimensional velocities, is not significantly affected by rotational speed. In addition, the

rotor massflow is shown to vary approximately linearly with rotational speed. For these

particular brake discs it was found that a reduction in the number of vane passages

increased the delivered massflow. Comparisons between hot -wire measurements and

pressure probe data, show good agreement in terms of flow angle but demonstrate the

inability of the pressure probe to capture the wake regions and the consequent

overestimation in flow rate.

The passage-to-passage wake-jet flow is found to be a characteristic signature of

the exit flow field with large unsteadiness in the wake region compared to the

mainstream flow. These flow patterns are affected by local geometric features which

influence the massflow and turbulence distributions. Although, the number of vane

passages has a primary effect on the massflow, it has only a modest effect on the

turbulence characteristics for each passage. However, the addition of pedestals and

part-span turning vanes has a dramatic effect on the flow by changing the turbulence

distributions. Although the peak levels are unaffected, the mean turbulence increases

by approximately 50% relative to the datum geometry. These local geometric features

also result in more uniform massflow and turbulence distributions along with a marked

improvement towards the inboard endwall. In addition, these enhancements are

achieved without any significant deterioration in the total massflow.

5 ACKNOWLEDGEMENTS

The authors would like to acknowledge the help of the workshop and technician

staff at the School of Engineering, Cranfield University.

19
REFERENCES

1. Sisson, A. E., Thermal analysis of vented brake rotors, SAE Paper

780352, 1978.

2. Cumpsty, N.A., Compressor aerodynamics, Longman Group, Harlow,

1989.

3. Lakshminarayana, B., Fluid dynamics and heat transfer of turbomachinery,

Wiley and Sons, New York, 1996.

4. Japikse, D., Introduction to turbomachinery, Concepts ETI, Vermont USA,

1997.

5. Wilson, D., Korakianitis, T., The design of high-efficiency turbomachinery

and gas turbines, 2nd Edition, Prentice Hall, 1998.

6. Whitfield, A., Baines, N.C., Design of radial turbomachines , Longman

Scientific and Technical, New York, 1990.

7. Barigozzi, G., Cossali, G. E., Perdichizzi, A., Boden, A., Pacchiana, P.,

Experimental investigation of the mean and turbulent flow characteristics

at the exit of automotive vented brake discs, SAE Paper 2002-01-2590,

2002

8. Hudson, M. D. and Ruhl, R. L. , Ventilated Brake Rotor Air Flow

Investigation, SAE Paper 971033, 1997.

9. Johnson, D., Sperandei, B., Gilbert, R., Analysis of the flow through a

vented automotive brake rotor, Journal of Fluids Engineering, Vol. 125, pp.

979 -986, November 2003.

10. Limpert, R., Cooling Analysis of Disc Brake Rotors, SAE Paper 751014,

1975.

11. Limpert, R., Thermal Performance of Automotive Disc Brakes, SAE Paper

750873, 1975.

20
12. Shen, F. S., Mukutmoni, D., Thorington, K. and Whaite, J., Computational

flow analysis of brake cooling, SAE Paper 971039, 1997.

13. Krsemann, R. and Schmidt, G., Analysis and optimisation of disk brake

cooling via computational fluid dynamics, SAE Paper 950791, 1995.

14. Daudi, A. R., 72 curved fins and air director idea increases airflow through

brake rotors, SAE Paper 1999-01-0140, 1999.

15. Zhang, J. J., A high aerodynamic performance brake rotor design method

for improved brake cooling, SAE Paper 973016, 1997.

16. Bleier, F. P., Fan handbook selection, application and design, McGraw-

Hill, Maidenhead, 1997.

17. Jorgensen, F. E., How to measure turbulence with hot-wire anemometers

a practical guide, Dantec Dynamics, 2002.

18. Voller, G., Tirovic, M., Morris, R., and Gibbens, P., Analysis of automotive

disc brake cooling characteristics, Proc. Instn. Mech. Engrs.,Part D:

Journal of Automobile Engineering, Vol 217, pp. 657-666, 2003.

19. Eck, B., Fans design and operation of centrifugal, axial-flow and cross-

flow fans , Pergamon Press, Oxford, 1973.

20. Krain, H., Swirling impeller flow, Journal of Turbomachinery, Vol. 110,

1988.

21. Hathaway, M., Chriss, R., Wood, J., and Strazisar, A., Experimental and

computational investigation of the NASA low-speed centrifugal compressor

flow field, Journal of Turbomachinery, Vol. 115, pp. 527-542, 1993.

22. Goldstein, R.J., Fluid Mechanics Measurements, Hemisphere Publishing,

1983.

23. Han, J., Dutta, S., Ekkad, S., Gas turbine heat transfer and cooling

technology, Taylor and Francis, New York, 2000.

21
24. Bunker, R., Wetzel, T., Rigby, D., Heat transfer in a complex trailing edge

passage for a high pressure turbine blade - Part 1: experimental

measurements, GT-2002 -30212, International gas Turbine Conference

and Exposition, Amsterdam, 2002.

25. Akhras, A., El Hajem, M., Morel, R., Champagne, J., The internal flow

inve stigation of a centrifugal pump, Proc. 10th International Symposium on

Application of Laser Techniques to Fluid Mechanics, Lisbon, 2000.

26. Ubaldi, M., Zunino, P., Ghiglione, A., Detailed flow measurements within

the impeller and the vaneless diffuser of a centrifugal turbomachine,

Experimental Thermal and Fluid Science, Vol. 17, pp.147-155, 1998.

22
 Rotor 1 Rotor 2 Rotor 3 Rotor 4
Outer Diameter (m) 0.33 0.37 0.366 0.37
Inner Diameter (m) 0.222 0.274 0.274 0.274
Total Thickness (m), B 0.028 0.032 0.032 0.032
Passage Height (m), b 0.016 0.0195 0.019 0.019
Area at inlet (m 2), A 1 0.0112 0.0168 0.0164 0.0164
Area at outlet (m 2), A2 0.0166 0.0227 0.0218 0.0221
Number of Blades, NB 48 48 36 24

Table 1 Rotor geometry specifications.

Rotational speed (rev/min) 2250 1500 750 1500 1500

Rotor # 2 2 2 3 4
Number of blades 48 48 48 36 24
BPF, Hz 1800 1200 600 900 600
Sample points
per blade pitch 11 17 33 22 33
Table 2 Estimated spatial resolution in the rotating frame of reference. Sample rate of 20kHz.

Axial traverse r, Cr
locations

z, Cz

V abs

Airflow
Cradial

Outboard Inboard

Pitch Angle, f f
Crel

Figure 1 Definition of measurement coordinate system and schematic of exit velocity triangles

23
 B.P
2xB.P

3xB.P 6xB.P
5xB.P
4xB.P

Fixings
Figure 2 Power spectral density at rotor exit. Rotor 4. 2mm

from exit plane. Mid-passage axial location, 1500 rpm.

50
750rpm 750rpm
(a) 45 1500rpm (c) 1500rpm
2250rpm 2250rpm
40 0.8

35

30
0.75
25

20

15 0.7

10

5
0.65
0
-0.4 -0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4
z/b z/b

Outboard Rotor Inboard

Face passage Face
30
(b) 750rpm (d)
1500rpm
2250rpm
25

20

15

10

0
-0.4 -0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4
z/b

Figure 3 Circumferentially averaged (a) absolute yaw angle (b) relative yaw angle (c) relative velocity (d)

radial velocity. Rotor 2 at 750, 1500 and 2250 rpm. Traverse taken 2mm from the rotor exit.
24
 60 30
Rotor#2 Rotor#2
(a) Rotor#3 (b) Rotor#3
Rotor#4 Rotor#4
50 25

40 20

30 15

20 10

10 5

0 0
-0.4 -0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4
z/b z/b

0.95
0.6
Rotor#2
(c) Rotor#2
(d) Rotor#3
0.55 Rotor#3 0.9
Rotor#4 Rotor#4

0.5
0.85
0.45
0.8
0.4
0.75
0.35

0.3 0.7

0.25
0.65
0.2
0.6
0.15
0.55
0.1 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4
-0.4 -0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4
z/b z/b

(e)

Figure 4 Circumferential averages (a) absolute angle (b) relative angle (c) absolute

velocity (d) relative velocity (e) radial velocity for Rotors 2, 3 and 4 at 1500RPM.

25
 0.25

0.20

0.15

0.10

Rotor 2
0.05 Rotor 3
Rotor 4
0.00
500 1000 1500 rpm 2000 2500

Figure 5 Non -dimensional massflow as a function of rotational speed.

50
(a) Pressure Probe
45 Hot-wire

40

35

30

25

20

15

10

0
-0.4 -0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4
z/b

(b)
0.7
Pressure Probe
0.65 Hot-wire

0.6

0.55

0.5

0.45

0.4

0.35

0.3

0.25

0.2
-0.4 -0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4
z/b

Figure 6 Comparison between hot-wire and five -hole pressure probe measurements for (a)

absolute whirl angle and (b) absolute velocity. Rotor 2, 1500rpm, 2mm from exit plane.

26
 V2/U2

C2/U2

b deg
25

0 20

15
0.5
10
1
5
2 3 4 5 6 7 8
y/b

Figure 7 Passage-to-passage variation of (a) absolute velocity (b) relative velocity and (c) relative whirl

angle. Rotor 2, 2250rpm at 2mm radial position.

U (blade speed)

Vabs Cradial

Crel

a2(wake) b2

Wake
Jet

Rotation

Figure 8Mid-span velocity triangles for Rotor 2 at 2250rpm showing the jet and wake regions.

27
 S e c on d ar y Fl o ws Rotation
Outboard
P as s ag e Wa ll s

(a) Rotor 2
SS PS

Va n e s

Inboard

Curvature of wake
S ma ll er J et du e to W ak e towards J et t ow ar d s S m a l l e r J e t d ue to
caused by secondary
mounting bolts s uc ti on s i de pr es s ur e si de m ou nt in g bo l ts C 2 /U 2
flows
Outboard

I n b o ar d Rotation

S m al l er j e t d u e t o C-s h a p e d w a k e , d u e t o S m al l er jet due to J et for c ed be tw een par ti al

m o unt i n g bo l ts par ti al wi dth pedes tal s m o un t i n g b o l t s w i dth pedes tal s C2 /U2
Jet

(b) Rotor 3 O ut bo ar d

Inboard Rotation
La r g er w a k e r e g i o n
c om pa r ed w i th r o t or #2

(c) Rotor 4
Vanes Rotation
Outboard

Inboard

C 2 /U 2
Larger wake due to Jet di stu rbances Larger wake due to
mounting bolt from balancing c lips mounting bolt

Outboard

Inboard
Rotation
L a r g e v e lo c i t y g ra d i e n t
towards outboard w a l l

Figure 9 Relative velocity distributions for Rotors 2, 3 and 4. Measurements taken at 2mm from the exit

plane with a rotational speed of 1500rpm (a) Rotor 2 (b) Rotor 3 (c) Rotor 4. Rotor 3 geometry cannot be

shown due to confidentiality agreements.

28
(a) Rotor 2 Cr/U2

0.35
0 0.3
0.25
0.5 0.2
0.15
1 0.1
0.05
2 3 4 5 6 7 8
y/b

Cr/U2
(b) Rotor 3 0.4

0.3
0

0.5 0.2

1
0.1
2 3 4 5 6 7 8 9 10
y/b

Cr/U2
0.4
(c) Rotor 4
0.3
0
0.5 0.2
1
0 2 4 6 8 10 12 14 0.1
y/b

Figure 10 Radial velocity distributions for Rotors 2, 3 and 4. Measurements taken at 2mm from the

exit plane with a rotational speed of 1500rpm (a) Rotor 2 (b) Rotor 3 (c) Rotor 4.

29
 Tu %
(a) Rotor 2
20

0 15

0.5 10

1
5

2 3 4 5 6 7 8
y/b

Tu %
(b) Rotor 3 20

0 15

0.5
10
1

2 3 4 5 6 7 8 9 10 5
y/b

Tu %
20
(c) Rotor 4
15
0
0.5 10
1
0 2 4 6 8 10 12 14 5
y/b

Figure 11 Turbulence intensity distributions for Rotors 2, 3 and 4. Measurements taken at 2mm

from the exit plane with a rotational speed of 1500rpm (a) Rotor 2 (b) Rotor 3 (c) Rotor 4.

30
 Outboard Rotor Inboard
Face passage Face
14
Rotor#2
Rotor#3
13
Rotor#4

12

11

10

6
z/b
5
-0.4 -0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4
z/b

Figure 12 Circumferentially averaged turbulence intensity.

Rotors 2, 3 and 4. 2mm from the exit plane. 1500rpm.

31