See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/222782238

Turboshaft engine air particle separation

Article  in  Progress in Aerospace Sciences · July 2010

DOI: 10.1016/j.paerosci.2010.02.001

CITATIONS READS
42 5,048

2 authors, including:

Antonio Filippone
The University of Manchester
129 PUBLICATIONS   1,029 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Multi-degree-of-freedom morphing airfoil featuring highly adaptive aerodynamic performance and innovative actuation system View project

Methods and Experiments for Novel Rotorcraft (MENtOR) View project

All content following this page was uploaded by Antonio Filippone on 08 April 2018.

The user has requested enhancement of the downloaded file.

 Turboshaft Engine Air Particle Separation

Antonio Filippone 1 & Nicholas Bojdo

The University of Manchester

Manchester M60 1QD
United Kingdom

Abstract

This contribution addresses air particle separation technologies (EAPS) used

for helicopter turboshaft engines that operate around unprepared airfields.
Three key technologies are discussed. These are: vortex tube separators,
inertial particle separators and inlet barrier filters. Some of these technolo-
gies are retrofitted or designed around specified airframes; other ones are an
integral part of the turboshaft inlet design. For the vortex tubes, a com-
plete review of the current intellectual property is reviewed, and a first-order
theoretical model is presented to corroborate the performance of this tech-
nology. The inertial particle separators are designed as part of the engine
inlets. This technology is more amenable to computational fluid dynamic
modelling, coupled with particle dynamics and rebound effects. The inlet
barrier filters are part of a large category of filtration systems. A first-order
filter theory is derived to predict filter performance as a function of key pa-
rameters and useful operation time. The integration of the various EAPS
with the helicopter system are discussed.
Keywords: Helicopters, Gas Turbine Engines, Turboshaft, Particle

1
School of MACE, George Begg Building. Corresponding Author. Email:
a.filippone@manchester.ac.uk

Preprint submitted to Progress in Aerospace Sciences December 17, 2009

Dynamics, Filters.

Contents

1 Introduction 7

2 Particle Separation 13

3 Vortex Tube Separators 16

3.1 Scavenge Flow . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.2 Helical Vane Design . . . . . . . . . . . . . . . . . . . . . . . 20
3.3 Vortex Tube Arrangement . . . . . . . . . . . . . . . . . . . . 21
3.4 Vortex Tubes Theory . . . . . . . . . . . . . . . . . . . . . . . 23
3.5 Equilibrium Radius . . . . . . . . . . . . . . . . . . . . . . . . 25
3.6 Particle Distribution . . . . . . . . . . . . . . . . . . . . . . . 28
3.7 Pressure Effects . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.8 Summary of VTS Technology . . . . . . . . . . . . . . . . . . 35

4 Inertial Particle Separators 36

4.1 General Features . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.2 Centre Body Design . . . . . . . . . . . . . . . . . . . . . . . 41
4.3 Scavenge System Design . . . . . . . . . . . . . . . . . . . . . 43
4.4 Hump and Splitter Design . . . . . . . . . . . . . . . . . . . . 47
4.5 Recent Novel Improvements . . . . . . . . . . . . . . . . . . . 50
4.6 IPS Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.7 Summary of IPS Technology . . . . . . . . . . . . . . . . . . . 55

5 Inlet Barrier Filters 56

5.1 Filter Element Design . . . . . . . . . . . . . . . . . . . . . . 60

2
 5.2 Effect on Performance . . . . . . . . . . . . . . . . . . . . . . 62
5.3 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.4 Barrier Filter Theory . . . . . . . . . . . . . . . . . . . . . . . 65
5.5 Summary of IBF Technology . . . . . . . . . . . . . . . . . . 74

6 Turboshaft Simulation 78

7 Conclusions 81

3
Nomenclature
CDF = Cumulative Density Function
Erf = Error function
Erfc = Complementary error function
FOD = Foreign Object Damage
IBF = Inlet Barrier Filter
IPS = Inertial Particle Separator
PDF = Probability Density Function
VTS = Vortex Tube Separator
A4 = cross-sectional area at section denoted “4”
B = buoyancy force on particle
c = particles concentration
Cd = particles drag coefficient
do = collector diameter
dp = particle diameter
D = aerodynamic drag
dvts = diameter of vortex tube (Eq. 15)
Dh = hydraulic diameter
f = fall factor (Eq. 40)
fo = friction coefficient (Eq. 31)
Fc = centrifugal force on particle
k = permeability
Lv = vortex length
p = pressure
pα = 1 − αc /2 (Eq. 19)
ṁ = mass flow rate
ṁ1 = mass flow rate per unit area

4
q∗ = quantile of log-normal distribution (Eq. 19)
Q̇ = volume flow rate
r = radius of curvature
R = total filter “resistance” (Eq. 43)
Re = Reynolds number
Ro = friction factor (Eq. 30)
t = time
U = air speed
Ur = radial velocity component
Ut = tangential velocity component
Ux = axial velocity component
V = volume
x1 , x2 = confidence interval (particle diameters)

α = cake resistance
αc = complement of confidence level
ζ = ratio between tangential and axial velocity
η = separation efficiency
κ = pressure drop coefficient
κ1 = orifice efficiency factor
µ = dynamic viscosity
µ = log-normal mean particle size (Eq. 16)
µ1 = mean particle diameter
ρ = density; log-normal mean value in PDF
σ = log-normal standard deviation (Eq. 16)
σ1 = standard deviation

5
[.]c = relative to filter cake
[.]f = relative to filter
[.]p = relative to particulate
[.]s = relative to separated or scavenged flow

6
1. Introduction

Helicopter operation to and from unprepared airfields has demonstrated

the vulnerability of turboshaft engines to the ingestion of foreign objects, in
particular dust and sand. For this reason, helicopter engines are equipped
with advanced separator systems to protect their rotating components from
erosion and damage. To understand the importance of particle separators
on modern helicopters, it is helpful to quote some performance figures. A
medium-size helicopter operating from and to dusty bases can ingest up to 1
kg of particulate per minute. Even if the average grain is not too large, this
ingestion can damage the engine beyond repair in a few hours of operation.
In practice, this means that turboshaft engines certified to operate in clear
environments for at least 300 to 400 hours between servicing, fail after about
30 to 40 hours of operation, with a loss of 90% of the normal service time.
The particle ingestion problem is particularly damaging during brown-
out, which generally refers to loss of visibility in near-ground operations, due
to a cloud of dust raised by the rotors. In such a condition, an unprotected
engine can be exposed to dust concentrations of 2.5 grams per cubic metre of
ingested air. Such high concentrations of dust cause serious problems for the
engine, including compressor blade erosion and combustor wall glazing. But
the threats to an engine’s performance are not limited to sand ingestion: in
marine environments, an engine may be vulnerable to corrosion and flame-
out as a consequence of saltwater consumption; in areas of vegetation such as
grassy fields, foliage may clog the air intake passage, causing flow distortion
and high pressure losses; and in most operations foreign objects such as
rocks, birds and chunks of ice may devastate a compressor blade, causing a
profusion of problems for the engine.

7
 Under these demanding conditions, the use of inlet filters or other means
of air-borne particle separation is therefore essential, if at least to provide
a physical barrier to such threats. However, the use of particle separators
does not quite solve the problem; in fact, it creates considerable side effects;
these include the added weight and drag, the power requirements to operate
the new systems, the need for constant inspection, reliability issues and the
inevitable costs that arise from installing such devices, including ground
logistics.
The use of filters is by no means limited to helicopters. Nearly every
piece of machinery operating in dirty environments has some sort of air
filter. As a consequence, much of the technology developed for helicopter
applications has relied on technical applications elsewhere. Some separators
are more widely referred to as “Engine Air Particle Separators” (EAPS) in
the technical literature.
Particle separators belong to one of three categories: 1.) Vortex Tube
Separators (VTS), that rely on centrifugal forces created by cyclone-like sys-
tems; 2.) Inertial Particle Separators (IPS), that rely on rapid change in
curvature of the inlet geometry; and 3.) Inlet Barrier Filters (IBF), that
rely on a mesh in front of the inlets. Separators in the second category are
available by default on some modern turboshaft engines, and some manu-
facturers indicate that they can be coupled with the other two categories
of filters. For this reason, they can be considered a category apart, whilst
vortex tubes and inlet barrier filters are more like retrofit technologies. The
decision to implement an IPS system is made much earlier in the aircraft
design process, dictated by the requirements of the manufacturer, whereas
VTS and IPS technologies tend to be implemented on request from the oper-
ator. IPS devices are therefore designed by the engine manufacturer, whilst

8
VTS and IBF technologies are outsourced by the airframe manufacturer
to private companies. A summary of known applications of VTS and IBF
technologies is shown in Table 1.
The first IPS system was installed in the General Electric T700 tur-
boshaft engine for the UH-60 Blackhawk, and was an integral part of the
engine design. Later, the same engine was used on the AH-64 Apache as
a default option. However, more recently it has been recognised that the
use of IPS systems can lead to unwanted pressure drops at the inlet even
in cases when a separator is not needed. Thus, a drawback of IPS systems,
designed as part of the engine, is their lack of flexibility.
VTS systems have enjoyed relative success when retro-fitted to existing
aircraft. The “EAPS” system devised by Pall Aerospace was first supplied to
the RAF for operation on the CH-47 Chinook in 1990, during the Gulf War.
Operating in desert environments at average flying times of 145 hours, 10%
of which were during brown-out, there were no engine rejections as a result
of erosion. By contrast, the US Army’s fleet of CH-47 helicopters which
were not fitted with engine protection, suffered an engine rejection rate of
20 to 40 engines per 1,000 flight hours due to erosion [1]. The success of the
VTS system has led to expansion of the technology into the civilian sector
for rotorcraft of all size, with many manufacturers implementing the VTS
system at the design stage.
In contrast to the previous two technologies, IBF systems are relatively
young. The advancements in filtration material technology have allowed
barrier filters to provide particle separation solutions with a competitive
pressure loss. Furthermore, many operators are now replacing the current
EAPS system with IBF, sacrificing increased maintenance time for higher,
dependable separation efficiency. For example the Bell 429 will offer as an

9
option a barrier filter, where in the past Bell may have favoured the use of
VTS protection.
The effectiveness of any of the technologies listed depends on the inlet
velocity, which is in fact a function of the duct’s area. For turboshaft en-
gines, this area is limited by several design and operational factors, including
engine-airframe integration, engine’s desired life-time, maintenance on the
ground, and other constraints.
In the analysis of a turboshaft engine, whose inlet is fitted with a par-
ticle separator or a filter, the key performance parameters are the rate of
separation (or efficiency of separation), the inlet pressure loss and the loss
of inlet mass flow rate. The residual particles can still cause damage to
parts of the engine, but indirect damage may also occur as a consequence of
contaminated lubricant, fouling of controls, clogging of small passages and
other factors. The loss of inlet performance causes a loss of propulsive effi-
ciency, e.g. a reduced shaft power, and most likely an increase in combustor
temperature. The problem is compounded by the power requirements on
the separator itself (centrifuges, impellers, compressor air bleed hydraulic
systems, etc.), although some separators are not powered.
Reliability is an additional issue. For example, the original EAPS fitted
on the Boeing’s V-22 Rolls-Royce AE-1107C engines caused failures that led
to catastrophic damage to three aircraft and were a major source of wear
and tear. Clearly, the location at which the separator is placed influences
the extent of such damage, and thus becomes an important factor in the
design process. In a retrofit, designers may be constrained by a fixed engine
location, but in new rotorcraft these risks can be overcome if the separator
is placed such that the airframe bears the worst damage from FOD objects.
The re-ingestion of propulsion by-products is another risk that can be

10
alleviated by intake location. This becomes important when considering the
integration of weaponry on military rotorcraft. Engine efficiency is reduced
in atmospheres of raised temperature due to an elevated turbine inlet tem-
perature, and the same holds true in a situation of Engine Gas Re-ingestion
(EGR). Therefore, a further consideration is installing the device at a loca-
tion on the airframe that will not be in the vicinity of the exhaust gas flow
field.
In spite of the increased interest in inlet particle separation, the techni-
cal literature on this subject is rather limited, with much of the information
derived from international patents, manufacturer’s specifications (often un-
substantiated by technical data), operating flight manuals (where data are
required for certification) and unofficial documents. To clarify the matter,
there are no less than 100 international patents on engine particle separa-
tors, some of which demonstrate a small advancement of the art without
showing any technical performance. The only publication with a wide tech-
nical scope is an AGARD Lecture Series [2], a document now out of print
and difficult to retrieve. Seddon [3] in his book on intake aerodynamics de-
votes about one page to helicopter engine inlets. However, filtration is an
engineering branch on its own, with applications on all types of machinery,
with a considerable publication record.
Separated particles need to be continuously removed, in order to avoid
accumulation in some parts of the engine. For this reason, VTS and IPS
methods that do not include scavenge flows are not attractive solutions.
When the separated particles are scavenged, they must be directed away
from the inlet, eventually also with the aid of the rotor downwash, to avoid
particle re-ingestion.
The scope of this paper is to review the various technologies available

11
for particle separation, and to provide a first-order engineering analysis of
their effects on the turboshaft performance, to corroborate or complement
the engineering knowledge of the different separation technologies. In our
analysis we exclude the case of Foreign Object Damage (FOD), which gen-
erally consist of large items, such as bolts, that cause instantaneous damage
and require an unscheduled engine removal. An example of FOD experimen-
tal and numerical analysis has been published by Calcaterra [4]. Laboratory
experiments rely on specific foreign objects shapes, speeds and incidence an-
gles. Numerical simulations are carried out with structural dynamics codes,
based on the finite-element method.
As mentioned, one of the consequences of particle ingestion is the damage
to the compressor blades. The reduction of blade erosion and its theoretical
basis has been investigated by van der Walt & Nurick [5, 6, 7]. These authors
concluded that the reduction in engine power due to erosion is proportional
to the mass of particles ingested by the engines. However, these papers
provide models that rely on a fixed separation efficiency, and do not consider
the erosion or damage on the EAPS itself. The EAPS bears the brunt of
the damage, and it is not unusual for parts of the separators themselves
to be ingested by the engine. In many cases, blade erosion can be partially
overcome by applying erosion-resistant coatings, as done on the Russian Mi-
8 and the American CH-53, powered by the General Electric T64 turboshaft
engine [8].
As a starting point, it is useful to employ a flow chart to visualise the
various types of protection and associated considerations, and the resulting
components of analysis. Depicted in Figure 1 is a breakdown of the engine
constituents, highlighting the intake as the area of interest. Within the scope
of intake appendages, the focus is on intake protection, examining the three

12
main categories mentioned earlier.

2. Particle Separation

Assume a mass concentration c ≃ 10−4 . For a Turbomeca Arriel 2B

engine having an inlet mass flow rate of 2.1 kg/s, there is an ingestion of
about 0.013 kg/min per engine. For the larger General Electric T64-GE-100,
with a nominal mass flow rate of 120 kg/s, the ingestion goes up to about
0.7 kg/min per engine.
One of the fundamental problems is to decide which particles to filter.
Sufficiently small particles, once lifted from the ground, stay aloft for a
considerable time; larger particles would tend to fall under the effect of
their own weight. Thus, although the ingestion of larger particles can be
more damaging, their occurrence is less likely. The terminal descent speed
of a spherical particle in Stokes flow and in absence of buoyant effects is

W ρp d2
U= = (1)
3πdµ 18µ
Accordingly, a 10 µm particle has a terminal velocity U ≃ 0.05 m/min,
whilst a 100 µm descends at about 4 m/min, a value still low enough for the
particle to be picked up by the engine. The Mil-E-5007C test dust consists
of particles greater than 70 µm, but all the particles below 200 µm make up
about 50% of the total mass. Thus, particles in the size range of 10 to 200
µm must be certainly separated.
Unfortunately, these statements are true only for particles that are forced
only once. The helicopter environment causes an intermittent forcing of the
particles, some of which are on the ground, and others are aloft in a cloud
of dust. The combination of rotor downwash, helicopter movement and

13
INTAKE
Compressor Combustor Turbines Exhaust

Intake Low
Protection Anti Icing FOD Mesh Observable
Attachment

Vortex Inertial Inlet

Tube Particle Barrier
Separator Separator Filter

Tube Array Engine mount Physical barrier

Swirl Vanes Hump & Splitter Filter element
Scavenge Scavenge Flow Pleated surface
Ejector pump Ejector pump Cake buildup
Flow distortion Open passage Maintenance

Factors First−order
Affected Assessment

Mass flow rate Pressure loss

Pressure ratio Separation
Flow stability Efficiency
Weight Weight
Drag Frontal area
Engine power Performance

Figure 1: Flowchart of EAPS.

14
 Table 1: Particle separator list database.

Manufacturer Model VTS IBF

Agusta-Westland AW-119 Koala • •

AW 109 •
AW 139 •
NH-90 •
Bell Bell 205 • •
Bell 206B • •
Bell 206 L3-4 •
Bell 407 •
Bell 429 •
OH-58-D •
Eurocopter EC-130/550 •
AS-350 • •
EC-135 •
EC-120 • •
AS 330/332 •
EC 225/725 •
Oryx •
Gazelle •
Puma/Superpuma •
SA 315B Lama •
SA 316 Alouette •
Tiger •
McDonnell-Douglas MD-500 •
MD-900 •
15
MD-902 •
MD-600N •
MDHI OH-6 •
AH/MH-6J/M •
Sikorsky UH-60 A/L Blackhawk •
UH-60S •
SH-3 •
helicopter configuration greatly complicates the matter.
Once the particles are separated, there are two possibilities: 1.) they
remain attached to the barrier filter (or rebound back into the air stream);
2.) they are scavenged out through a scavenge system (as discussed later). In
the latter case, the design and integration of the engine must avoid particle
re-ingestion.
To justify the use of Stokes’ law in the derivation presented here, the
Reynolds numbers of the particles are relatively small. For inlet velocities
of the order of 10 m/s, a 10 µm particle has a Re ≃ 7; a 100 µ m has a
Reynolds number as low as 70. Although the Reynolds number increases
linearly with the inlet speed, it is still relatively small. However, more
accurate drag models do exist, including those for non-spherical particles
which employ a shape factor, such as the model by Haider & Levenspiel [9]:

24   b + Re
3
Cd = 1 + b1 Reb2 + (2)
Re b4 + Re
where the coefficients bi are a function of a shape parameter. Equation 2
is valid for Reynolds numbers below 105 , which clearly includes the cases
discussed in this paper.

3. Vortex Tube Separators

Vortex tube separators are one of a type of air cleaners known as cen-
trifugal particle separators. A vortex tube contains a set of static helical
vanes at its inlet, and a secondary inner tube of reduced diameter located
at its base. A scavenge passage extracts the dirty air by use of a blower,
as the clean air continues on to the engine. Modifications of this design
can lead to improved particle separation efficiency, better compactness, and

16
 Figure 2: Vortex tube separator according to US Patent 4,985,058 [11].

a lower pressure loss penalty. Furthermore, these tubes are habitually ar-
ranged in arrays – panels of tubes which serve as sides of a box or cylinder
to be placed in front of the engine inlet. There is much scope for how these
are arranged, often dependent on the helicopter to which an array is being
fitted, as discussed further in Ref. [10].
As particulate-laden air enters a vortex tube separator, it is first met by
a set of helical vanes, which impart a radial and tangential component of
velocity to the flow, inducing rotation. Particles in the air are of greater
specific gravity, and so experience a greater centrifugal force in this rota-
tion. Owing to the effects of inertia, this causes the particles to be thrown
outwards towards the periphery of the tube. The vanes bestow a similar
fate to heavier particles too, by virtue of a design which deflects or trains
particles on impact radially outwards. A second, narrower tube in the base
of the device physically separates the flow into two streams; the core air
flow continues to the engine inlet, whilst the particulate-laden “dirty air” is
scavenged to the atmosphere.
The diagram shown in Fig. 2 is one example of such a vortex particle

17
separator, illustrating one embodiment of Ref. [11]. The outer tube, labelled
12 in Fig. 2, has an inner diameter of 18mm, a total length of 60mm, and a
vortex generating region of length 20mm; however it will be seen that these
dimensions differ. The area labelled 20 in Fig. 2 is known as the separation
region, in which the vortex forms a clean air core. Adjacent to this and
common to all VTS is a second tube of smaller diameter but co-axial with
the main tube, through which the clean core air 26 flows. In this example
it is tapered, increasing in diameter downstream, but in other inventions
the diameter may remain constant. The design often depends on the means
for removing the dirty air. The centrifuged particulate matter 22 proceeds
through the annular orifice between the inner and outer tube and arrives
at a scavenge passage, such as that labelled 46. The dirty air is then often
scavenged away through holes, either in the base of the passage, or the tube
walls 48, and proceeds into a chamber common to all dirty air tube outlets
and discharged to the environment. A fan or blower is usually provided
to fluidise the scavenge air flow, which constitutes from 5% to 20% of the
primary airflow. The primary airflow entering the device in this example is
based on a mass flow 4.4 g/sec, but this will clearly vary among devices of
differing dimension, and local pre-inlet flow conditions.
As with all devices of intake protection, there is a constant battle between
achieving good separation efficiency for a minimum pressure loss. Aside from
altering the arrangement of the tubes, patents often pertain to inventions
which improve separation efficiency for no extra loss of pressure, and vice
versa. For example, U.S. Patent 4,985,058 in Fig. 2 above describes a new
approach to fixing the secondary, inner tube, which eliminates the need
for the spokes found in prior art that detrimentally affect the rotational
component of the flow. This improves the separation efficiency of the system.

18
When operating at a pressure drop of 0.96kPa, this device is claimed to be
capable of achieving a total mass efficiency of dust removal of about 98%,
for AC coarse test dust with a scavenge flow of about 10%. Other patents
aim to improve the system efficiency by modifying the helical vanes, or the
arrangement of the tube. The importance of such adjustments is embodied
in a sentence found in Ref [11]: “If the separation efficiency of the inlet
system increases from 94% to 95%, the life expectancy [of the engine] is
doubled, and if the efficiency then increases to 97%, the life expectancy is
doubled again.”

3.1. Scavenge Flow

The design of the scavenge flow means has been found to be very im-
portant in dictating the overall efficiency of the system. In Ref. [12] it is
reported that “The efficiency of a vortex separator is a function of the cen-
trifugal forces developed, the length of the spinning zone, and the proportion
of the scavenge flow to clean air flow. Of course, an optimum clean air flow
is always the primary desideratum. Therefore the scavenge air should be
controlled at the minimum to give effective particle separation.”
The problems connected with controlling the scavenge flow appear to
depend on the design of the vortex tube base. For example, a device may
employ a conduit in the side wall of the scavenge passage for extracting the
dirty air to a desired location. If the orifice becomes clogged, there is risk
that flow will re-circulate and enter the clean air flow tube, exacerbating an
already underperforming device. Furthermore, owing to the high rotational
velocities, any particles that fail to be extracted will continue to circulate
inside the scavenge passage, abrading the walls. This could lead to catas-
trophic failure. Inventions such as adding a split washer-style flange at the

19
scavenge entrance alleviates problems such as this, but inevitably leads to
efficiency losses. Similarly, an increase in complexity not only affects manu-
facturability, but increases the possibility of turbulence within the scavenge
passage, leading to the risk of re-ingestion into the clean air flow.

3.2. Helical Vane Design

Whilst the main task of the helical vane is to turn the air and cause a
vortex, there have been many modifications to the vane leading edge, airfoil
shape, hub length, chord length, vane trailing edge and vane surface, all
driven by the need for greater separation efficiency and a lower pressure loss.
In almost all cases, there are four vanes in a set, each with a chord length
of at least 90 degrees measured along the tube circumference such that each
vane overlaps its neighbour. According to Monson & Rosendahl [13], such
an overlap ensures the application of swirl-force to each dust-air particle,
either aerodynamically or by direct deflection.
The leading edge angle of attack determines the swirl angle of the influent
air; a larger angle of attack produces greater swirl per unit travel at the
expense of greater pressure drop. Monson & Rosendahl [13] continue by
pointing out that the desired angle of attack for a good balance generally
lies in the range of 55 to 70 degrees. If space constraints dictate a shorter
hub, then a high angle of attack might be preferred to achieve the suggested
vane overlap of at least 90 degrees.
A patent study has revealed some of the parameters that can be improved
to achieve a greater efficiency. For example, in U.S. Patent 3,517,821 [14] it
has been shown that employing an ellipsoidal leading edge and modifying
the high pressure surface to gradually alter the flow path of the air into
a spiral lowers pressure losses. A similar modification of the low pressure

20
surface improves the pressure coefficient further, by reducing the intensity
of a vortex which forms adjacent to the leading edge of prior art vanes due
to flow separation. The same inventors also determined experimentally that
feathering the trailing edge narrows the width of the wake trailing off the
vanes, which lowers the pressure losses proportionally.
It is clear that such improvements are in most cases specific to an existing
VTS, and since no mathematical links have so far been found, modifications
are invariably established through experiments. Other inventors have de-
veloped ideas to improve efficiency for no loss of pressure, such as a second
set of vanes located where the clean air tube usually lies, to separate finer
particles [15]; or grooves spanning the vane length which serve to divert wa-
ter droplets and other particles that adhere to the vane surface, to the tube
periphery [16]. While no general mathematical theory has been established
for helical vane design in VTS, no one can doubt the underlying scientific
principles behind these improvements.

3.3. Vortex Tube Arrangement

As previously discussed, VTS are invariably arranged in a vortex cleaner

array, defined by Roach [17] as an assembly composed of a plurality of vortex
cleaners mounted together as a unit with their axes arranged in parallel, or
a group of such assemblies. It continues to state that six rows of vortex air
cleaners is the maximum number of rows generally used in air cleaners for
good cleaning efficiency, but earlier inventions pertain to panels containing
much larger numbers of VTS arranged side by side to comprise a panel,
such as in U.S. Patent 3,449,891 [18] depicted in Fig. 3. This particular
patent refers to several embodiments of an idea in which panels of tubes
are arranged to form a box or cylinder 40. Such an arrangement creates a

21
 Figure 3: Vortex cleaner array with bypass door from U.S. Patent 3,449,891 [18].

chamber which, when a bypass door 60 is opened, can provide axial airflow
to the engine inlets 20. This permits almost full performance of the engine
when operating in clean environments, and an emergency system should the
separator become clogged.
Improvements in the arrangement of VTS have been seen most recently
to be driven by abating noise from the engine, as stricter regulations at air-
ports are introduced. One source of rotorcraft noise is that which emanates
from the compressor, especially at high mass flow rates during take-off. How-
ever, if rotorcraft are expected to land in dusty environments too, a particle
separator is essential in addition to existing noise attenuation devices. One
such device, by Roach [17], in which sound absorption panels inside the clean
air chamber are positioned just aft of the VTS, are arranged in curved, dog-
legged, or in angled formation. This arrangement abates sound waves by

22
deflecting the airflow, but unfortunately leads to additional pressure losses
in this embodiment of around 0.65kPa, which is an increase of almost 50%
on the pressure losses from the particle separating device. Nevertheless, it
is an increasingly essential piece of hardware for rotorcraft, and the pressure
losses are somewhat alleviated by deceleration of the air in the chamber.
Such an invention highlights the increasingly complex and novel approaches
being proposed in the industry to further enhance the capabilities of VTS
devices.

3.4. Vortex Tubes Theory

We now proceed with a basic fluid dynamic model to predict the main
performance parameters of the VTS. Figure 4 shows a sketch of a typical
vortex tube. The inclined surface denotes a deflection vane. The inlet
section is denoted with “1”; section “2” is downstream the vanes; section
“3” is at the base of the collector; section “4” is a conventional outlet surface;
on the scavenge side, it is a port connected with a tube, and thence with
the outside atmosphere. The collector has a conical shape; in this form, it
creates a constriction to the passage of the flow through the scavenge port,
and thus a loss in pressure.
The separation efficiency is defined as the portion of the particulate that
is scavenged out of the inlet. This portion can be determined in terms of
mass, volume and numbers. A number of other definitions are required.
First, the scavenge flow: this is the flow driven off the separator through
the annular passages; this flow contains the large size particulate, as well as
some air. The pressure difference from the inlet to a conventional section
of the scavenge flow is called ∆ps , or scavenge pressure loss. The scavenge
flow is usually referred to the inlet “clean” flow (also called primary flow);

23
 mp

helix

1 2 3 4

Figure 4: Vortex tube separator; Compare with Fig. 2.

for convenience it is given as a fraction of the latter quantity. The pres-

sure difference between the inlet and a conventional section of the primary
flow is called ∆pp . The performance of the vortex tube separator can be
characterised by these two parameters.
The initial mass concentration of the particulate is

  !
ṁp ρp V̇p
c= = (3)
ṁ ρ V̇
and will be an assigned value in our analysis. The initial conditions at
section 1 are all known. However, note that if the effective cross-sectional
area is assigned, then the average inlet velocity is calculated from the mass
flow rate, and vice versa. Since in practice there are several vortex tubes in
the separators, it is convenient to work in terms of mass flow rate per unit
inlet area, m1 .
We will need a further assumption regarding the transport of the particu-
late. In our first-order analysis we assume that the particulate is transported
by the mass flow at the same average speed. Thus, in the case in which the
effective cross-sectional area A at the inlet is assigned, the average inlet

24
velocity for both air and particulate is calculated from

ṁ + ṁp ≃ ρ(A − Ap )U i + ρp Ap V i (4)

or

ṁ(1 + c)
Ui ≃ (5)
ρA + ρp Ap
having assumed that Ap << A. If we assume that the particles are trans-
ported with the same velocity as the air flow, then

    
ṁp ρp Ap Up ρp Ap
= ≃ (6)
ṁ ρ A U ρ A
from which we find that the ratio between cross-sectional areas is

 
Ap ρ
=c (7)
A ρp
The ratio between the density of the particles and the air is of the order
of 2 · 103 . Therefore, Ap ≃ 0.5 · 10−3 cA, or Ap << A. This result justifies
Eq. 5.

3.5. Equilibrium Radius

The first aspect to consider is the particle dynamics in the vortex tube.
The particles enter the tube with an axial velocity; they are rapidly deflected
by helicoidal vanes that cause a sudden increase in angular momentum.
At the exit of the vanes, the particles will have a rotational component
and a residual radial velocity component. Further downstream, the flow
establishes itself to a steady-state condition, and the radial component ceases
to be dominant. The particles’ flight path is helicoidal, around a radius r
that can be determined with a first-order analysis of the body forces.

25
 The velocity of a particle has components Ur , Ut , Ux (radial, tangential
and axial, respectively). Each particle is characterised by a diameter dp
(or radius rp ) and has a specific mass ρp . Such a particle is subject to an
aerodynamic force, a buoyant force and a centrifugal force. In its helicoidal
motion, the drag is

D = 3πdp µUr (8)

which expresses the Stokes’ law for a spherical particle at a small Reynolds
number. The centrifugal force is

Ut2 1 U2
Fc = mp = πρp d3p t (9)
r 6 r
The buoyant force is calculated from the difference in density between the
particle and the air flow:

Ut2 1 U2
B = −V ρ = − πd3p t (10)
r 6 r
Assuming that the process has reached a steady state, the balance of forces
is

D + Fc + B = 0 (11)

which is solved for the unknown radius of curvature of the particle’s flight
path. The solution is

(ρp − ρ)d2p Ut2

r= (12)
18µ Ur
Equation 12 expresses the equilibrium radius of curvature for a particle of
diameter dp having a tangential velocity Ut and a radial velocity Ur . Thus,

26
the equilibrium radius increases with the square of the particle’s diameter,
and with the square of the tangential velocity, all other quantities being
constant. For any given particle’s ensemble, it is convenient to design vanes
that maximise the gain in angular momentum.
The critical diameter dpc of the separated particles depends on the di-
ameter of the collector, do . Particles that have a trajectory with radius of
curvature r > do /2 are likely to be separated. The critical diameter is found
by inverting Eq. 12, with r = do /2:

 1/2
9µdo Ur
dpc = (13)
ρp − ρ Ut2
Other effects intervene, namely the effects of random turbulence and the
effects of collisions between particles and walls; the latter event may cause
large particles to be ingested rather than separated, as a result of them
bouncing back from a collision with a wall. However, these are statistical
cases that are not accounted for in a first-order analysis.
For a given inflow condition, the tangential velocity at the exit of the
vanes (section 2 in Fig. 4) is due to the shape of the vanes. Thus, the ratio
ζ = Ut /Ux should be a design parameter. With this definition, we have a
more suitable expression of the critical radius in terms of the inlet velocity

(1 − ρp )d2p 2 Ux2
r= ζ (14)
18µ Ur
The rotational and radial components of the particle depends on the
construction of the device. The axial velocity component depends on the
turboshaft, and can be assessed from

 π 
ṁ = n d2vts ρUx (15)
4

27
where dvts is the diameter of the vortex tube and n denotes the number of
cylindrical separators.

3.6. Particle Distribution

The type and distribution of the particulate depends on the environment
of operation. There is great variability in dust and sand characteristics. For
this reason, some standards are available to aid the engineering analysis,
for example, US Military Standards MIL 810D (dust and sand), MIL 8593
(sand), British Standard BS-1701 (fine and coarse), ISO 12103 PT 1 (ul-
trafine, fine, medium, coarse), and several others. A convenient way of
characterising the statistical properties of the particulate consists in plot-
ting the percent in weight versus particle size (see for example Duffy [19]).
In fact, such a distribution permits a rapid assessment of the percentage of
particulate below or above a threshold size.
Without lack of generality, we assume that the particulate has a log-
normal distribution

(ln x − µ)2
 
1
PDF(x, µ, σ) = √ exp − (16)
xσ 2π 2π 2
where µ is the mean particle size and σ is the standard deviation of their
own respective natural logarithm. This is a well-known distribution, whose
characteristics are given in several mathematics textbooks. If the mean µ1
and the standard deviation σ1 are known from a particle sample, then the
values of µ and σ in Eq. 16 are

   
1 σ1 2 σ1
µ = ln µ1 − 1+ 2 , σ = ln 1 + 2 (17)
2 µ1 µ1
Assuming a confidence level 1−αc = 0.99 (e.g. αc = 0.01), the corresponding
interval of the log-normal distribution is

28
 x1 = exp(µ − σq ∗ ) x2 = exp(µ + σq ∗ ) (18)

where q ∗ is the (1 − αc /2) quantile of the normal distribution. If pα =

1 − αc /2, then

√
q ∗ = Φ−1 (pα ) = 2 Erf−1 (2pα − 1) (19)

The cumulative density function is

 
1 log x − µ
CDF(x, µ, σ) = Erfc − √ (20)
2 σ 2
The cross-sectional area, the mass and the volume follow the same log-
normal distribution. The cumulative density function, Eq. 20 allows the
calculation of the various characteristics in terms of percent. Hence, if the
independent variable is the particle diameter d, the efficiency of separation
becomes

η = 1 − CDF(dpc ) (21)

Equation 21 expresses the fraction of particle mass that escapes filtration.

This quantity is uniquely determined by the particle size distribution and
by the critical diameter.

The fraction of scavenged mass flow and the fraction of cross-sectional

area are known from Eq. 20. The total particulate mass flow rate is also
known from the initial concentration. From Eq. 7, Eq. 3 and Eq. 20 we find

h i h i ρ 
Aps = 1 − CDF(dp ) Ap ≃ 1 − CDF(dp ) cA1 (22)
ρp

29
3.7. Pressure Effects
Call ∆pr the change in pressure in the constriction, and ∆pc the change
in pressure in the core/collector. The loss in pressure through the tube,
from the front-facing section to the collector, is called ∆pe . We assume
that there is no difference in pressure between the outer flow (with the
largest particulate) and the core/primary flow. Thus, on the scavenge side,
the change in pressure through the separator is ∆pe + ∆pr ; the change in
pressure through the core/primary side is ∆pe + ∆pc . Since the flow is most
likely separated inside the collector, it is reasonable to assume that there is
only a minimal change in pressure; hence we assume ∆pc ≃ 0. The ratio
∆ps /∆pp is a design parameter and can be considered as a known value.
This ratio can be written as:

∆ps ∆pe + ∆pr ∆pe + ∆pr ∆pr

p= = ≃ =1+ (23)
∆pp ∆pe + ∆pc ∆pe ∆pe
From the above equation, we solve in terms of ∆pr

 
∆ps
∆pr = − 1 ∆pe (24)
∆pp
One can assume that the change in pressure inside the collector is not negli-
gible, and set this quantity equal in value but opposite to ∆pr . In the latter
case,

p−1
∆pr = ∆pe (25)
p+1
In practice, the value of ∆pr is included within these two extremes. The
scavenge flow rate is calculated from

s
∆pr
ṁs ≃ κ1 As (26)
ρ

30
with the pressure drop taken from Eq. 24. The factor κ1 is an efficiency
factor for the orifice, and As is the effective cross-sectional area available to
the mass flow. The latter quantity is calculated from the difference between
the geometrical area and the area occupied by the particulate (a blockage
effect), Aps :

As = A4 − Aps (27)

When A4 ≃ Aps the scavenge flow is choked; only the particulate moves
through the scavenge passage. The area occupied by the scavenged par-
ticulate is calculated from Eq. 22. By combining Eq. 22, Eq. 27, Eq. 26
and Eq. 24, it is possible to provide an estimate of the scavenge mass flow
rate as a function of the design parameter ∆ps /∆pp . All parameters being
constant, the change in scavenge flow rate is according to

 1/2
∆ps
ṁs ∝ −1 (28)
∆pp
However, from Eq. 24 it can be inferred that the larger the pressure loss
upstream the collector, the larger the scavenge flow rate. For a given particle
size distribution, the performance of the vortex tube depends on several
other factors, in particular the geometry of the collector itself (front-facing
diameter and exit diameter), and geometry of the inlet vanes (and hence the
rotational velocity component).
The determination of the pressure loss upstream the collector, ∆pe ,
which occurs up to the front of the collector, is a key factor in assessing
the scavenge flow rate, and thus the loss of inlet flow. This pressure is a fac-
tor in Eq. 24. We associate the loss in pressure to a pressure drop coefficient
κ. The pressure drop is written as

31
 1 2
∆p = κρU i (29)
2
The pressure drop coefficient depends on the geometry of the assembly, and
cannot be easily calculated. Essentially, this quantity depends on the friction
coefficient, and possibly also on a shape factor. We define a friction factor
Ro

Lv
Ro = f o (30)
Dh
where fo is a friction coefficient, Lv is the length of the vortex, Dh is the
hydraulic diameter. The friction coefficient is estimated from the assumption
of fully turbulent flow from

fo = 0.314Re−1/4 (31)

The Reynolds number is calculated on the mean air flow with the hydraulic
diameter as a reference length, Re = ρU i Dh /µ. The estimation of the
vortex length is problematic, because it depends on several factors, including
geometry and flow conditions. We assume that this length is equal to the
vortex tube, e.g. Lv ≃ L.
A calculation following the method presented is shown in Fig. 5. This
figure displays the separation efficiency as a function of the tangential ve-
locity component, for a fixed inlet mean velocity, as defined by Eq. 5. The
rotational velocity component depends on the vane design.
Figure 6 shows the scavenge flow rate relative to the inlet mass flow, as a
function of the pressure ratio ∆ps /∆pp . This curve is only weakly dependent
on the separation efficiency, because the latter parameter depends only on
the critical diameter. In fact, this is a drawback of the one-dimensional

32
 1

0.95
Separation efficiency

0.9

0.85

0.8

0.75

0.7
0 0.2 0.4 0.6 0.8 1
Ut / U

Figure 5: Predicted separation efficiency for vortex tube separator, as a function of the
tangential velocity component.

33
 10

Scavenge mass flow rate, %

6

0
1 1.5 2 2.5 3
∆ps / ∆pp

Figure 6: Predicted scavenge flow for vortex tube separator, as a function of the pressure
ratio.

theory. One effect of the pressure drop ∆pr is to scavenge also particles that
would tend to be sucked into the collector.
Some results accompanying US Patent 3,520,114 [20] (a separator some-
what different than the one modelled, yet still based on the vortex tube
concept) imply that there can be a gain of 7 to 8% in efficiency when the
scavenge flow reaches about 8 to 9%. Above this value of the scavenge flow,
there is no gain in efficiency. Results shown by Ballard [21] imply that the
separation efficiency of a vortex tube decreases as the flight speed increases,
which is an apparent contradiction of the result shown in Fig. 5. In fact,
as the forward speed increases, the engine power requirement goes down; so
does the inlet mass flow rate. Thus, for a given inlet area, the average inlet
speed decreases (Eq. 5), and the separation efficiency decreases, a result in

34
line with the present theory.
In spite of the simplified one-dimensional theory, the results obtained are
in line with the claims from the patent owner, e.g. separation efficiencies of
the order of 90%, perhaps up to 93%, with scavenge flow rates out to 10%.

3.8. Summary of VTS Technology

In conclusion, particle separation with a vortex tube is a complicated

process whose efficiency depends on the design of the complete system. The
role of the vanes is important in imparting both tangential and rotational
velocity components. The shape of the collector is essential in determining
the pressure losses both on the scavenge and the core side. The scavenge
pressure is important in the determination of the scavenge flow rate, and
hence the loss of net mass flow into the engine. The loss in pressure between
the front-facing section and the collector has a more limited effect on the
overall performance of the device, since the loss of mass flow ultimately
causes the engine to deliver a reduced shaft power.
A change in particle structure (mean diameter, standard deviation and
shape) causes the system to perform with a different efficiency. However,
this is a common problem in all separator technologies. Advance knowledge
of the theatre of operation would allow better EAPS design. In principle,
there can be several different retrofit VTS systems for the same helicopter.
Other problems have been identified. These include a relatively poor
performance in icing conditions (not included in our analysis), and a re-
quirement that the inlet flow be as normal as possible to the face of the
vortex tubes. Aside from damage caused by larger particles, the vortex
tube separators have a relatively consistent separation efficiency, which is
one clear advantage.

35
4. Inertial Particle Separators

The term Inertial Particle Separator refers to a device which is integral

to a gas turbine engine, fitted at the air inlet for the purpose of cleaning
influent, particulate-laden air. A radial or tangential component of velocity
is imparted to influent air, before ducting causes a change in direction of the
flow stream. The linear momentum of particles entrained in the air hinders
such a rapid change of direction, which allows for their easy separation
from the core air stream. The highly concentrated stream of particles is
encouraged into a scavenge conduit and extracted away, as the clean air
continues to the engine. IPS change in shape, size and type, depending on
the engine mass flow rate or whether they are applied to a radial or front
facing inlet. Devices may contain swirl vanes too, which augment separation
by the deflection or centrifuging of particles. Furthermore, improvements to
existing IPS can be made possible by new technologies such as active flow
control and adaptive surfaces, which can modify the device during operation
to optimise performance to a current flight condition.
The most clear-cut distinction between IPS devices is in their applica-
tion, to either a front facing inlet or a radial inlet. A front facing, axial inlet
IPS device imparts a radial component of velocity to the incoming flow,
whilst accelerating the flow through an annular duct, and uses an annular
splitter to separate the flow into dirty and clean air. A radial inlet scavenges
off particulate from the outer walls of a spiral or U-shaped duct, via “scoops”
located in the periphery of the duct, relying on the principle that particles
will be centrifuged outward as the flow turns through a duct. Swirl vanes
are often employed in addition to front facing inlet to impart a tangential
component of velocity to the flow and further enhance separation. Design of

36
the scavenging system is also something which is heavily deliberated over,
requiring much thought to minimise weight and maximise compactness. For
example, the scavenged particulate may be exhausted through a radial spi-
ral, or an axial duct. Further application of vanes is also sometimes seen, as
an additional separator, for de-swirling purposes or to prevent particulate
being regurgitated back into the engine.
A review of technology up to 1970 was done by McAnally & Schilling [22],
who selected eight EAPS concepts and evaluated them for ten different
rating factors. The best concepts were the “semi-reverse flow” and “powered
mixed-flow”, that were later demonstrated on test rigs [23].

4.1. General Features

Front-facing IPS are comprised of a central body, co-axial with the engine
surrounded by a cowl, and a splitter positioned to divide the airflow into
a dirty flow stream and a clean flow stream. The central body, labelled 16
in Fig. 7, directs influent air through an annular duct of decreasing area
between itself and a cowl 18. The shape of the central body is such that the
radius of the annular duct increases with axial distance from the inlet to a
point, and then decreases to the engine inlet at 14. This shape creates a peak
15 at which the flow makes a rapid turn, remaining attached to the inner
surface of the annular duct by viscous forces. The geometry of the annular
duct causes acceleration of the particulate-laden air up to the peak, which
adds linear momentum to the particles. At the peak, most of the particles
are unable to turn with the flow due to their inertia, which projects the
particles radially outward. An annular splitter 17 bifurcates the airflow,
segregating the particles into a scavenge conduit 19, along with 10% to 30%
of the air flow. The particles are then discharged to the atmosphere, as the

37
 Figure 7: Front-facing IPS according to US Patent 4,389,227 [24].

clean air follows the inner wall of the central body to the engine.
Another objective of the IPS is to deflect heavy particulate or other
foreign objects into the scavenge conduit. This type of pollution may be
too heavy to be entrained in the flow, and so designers must ensure that if
particles are to bounce off surfaces, their rebound will be directed into the
scavenge chamber.
Radial inlet or scroll type IPS devices may be integrated into an en-
gine, such as in U.S. Patent 6,134,874 [25] shown in Fig. 8, or may have
been designed as an addition to a forward facing inlet as in U.S. Patent
3,993,463 [26] illustrated in Fig. 9. In the former case, it can be seen that
the particle separator has been designed integral to the engine, by the en-
gine manufacturer. It is unlikely therefore, that similar embodiments exist,
although the concept of designing an air cleaner into the engine is often per-
formed by the manufacturer. The latter type of radial separator is designed

38
 Figure 8: Front-facing IPS according to US Patent 6,134,874 [25].

more as a retro-fit to an existing engine, and is added to a front facing inlet,

where space constraints or otherwise prevent the use of an axial IPS.
In the embodiment featured in Fig. 8, air is drawn into the axial com-
pressor through one of two axially symmetric radial inlets 11 with a concave
duct 10. The forward surface 12 forces the air into an arcuate path, which
centrifuges heavier particles through the bypass duct 14. A cusp lip 15 acts
as a flow splitter, for diverting bypass airflow into the bypass duct. Dirty air
is collected from both radial inlets by an annular conduit 19 and extracted
to the atmosphere by exhaust driven jet pumps.
The radial separator shown in Fig. 9 is affixed to a forward facing engine
inlet. Air is drawn in radially, through a helicoidal duct 100 which spirals
inwardly to the engine inlet 110. The heaviest particles are initially guided
into a scavenge conduit 115 which is separated from the main flow path

39
Figure 9: Radial IPS according to US Patent 3,993,463 [26].

40
106 by a wall 114. The resulting scavenge channel 104 follows the duct
outer wall and is further supplied with dirty air by “louvers” 116 which
scavenge centrifuged particles from the duct periphery. The dirty air is then
exhausted to the atmosphere, providing approximately 80% to 85% of the
influent air as relatively clean inlet air to the engine.
There are advantages and disadvantages to both axial and radial sepa-
rators. Axial separators are often selected for engines with dynamic or ram
intakes, which are aligned perpendicular to the airflow to augment pressure
recovery. They or their ducted inlets are located ahead of the main rotor
mast, which minimises the risk of Exhaust Gas Re-ingestion (EGR). This
is in contrast to radial separators, which are found on engines with “static”
intakes, and are often located aft of the rotor mass. Static intakes receive
airflow through air inlets aligned parallel to the airflow, which when coupled
with a radial inlet of the type shown in Fig. 8 can lead to increased flow
distortion at the compressor inlet. However, radial inlets are more compact
and if integrated into the engine as in U.S. Patent 6,134,874 [25], carry a
significantly lower weight penalty over front-facing axial IPS devices.
In addition to the three different types of IPS, there are many other
manifestations of the principle, or additions to existing designs. The main
design drivers are increased separation efficiency or a lower pressure loss.
This may include a novel centre body design, the addition of swirl vanes,
an improved scavenging system, but in most cases patented developments
pertain to the actual separation technique.

4.2. Centre Body Design

The main role of the centre body is to impart a radial component of

velocity to the oncoming flow. Another objective may be to provide an

41
accurately designed surface off which heavier particles will rebound into the
scavenge chamber. It is important that any particles that are too heavy
to be carried by the airflow and experience the change in velocity, do not
bounce off the central body, then again off the cowl and into the clean air flow
conduit. The IPS featured in U.S. Patent 3,148,043 [27] demonstrates how a
design can achieve this objective. One could argue that the rebound effect is
not only complicated by the combination of shape, size and material of the
dust, but also by the surface quality of the outer walls. Thus, as these walls
are eroded by the particulate, the rebound process is altered; inevitably, the
IPS will not perform as by design.
Another problem experienced by these devices is the accumulation of
moisture on the central body, which possesses no coefficient of restitution,
but is nonetheless composed of water droplets too large to be entrained in
the airflow. The problem arises when water droplets adhere to the central
body upon impact, coalesce with other droplets, and stream inboard along
the surface. Without a discontinuity, the water would eventually reach the
engine.
In other embodiments, the problem of moisture may be alleviated with
a different technique. For example, U.S. Patent 4,389,227 [24] employs V-
shaped fences at the central body peak, whose arrangement is such that
moisture accumulates and eventually becomes dislodged from the surface
from increasing mass. The device is designed such that the resulting trajec-
tory carries the water into the scavenge conduit. The central body may also
be of a different shape, depending on the inlet size and mass flow require-
ments. It is thought that in order to properly control the directional flow of
the intake air, the maximum diameter of the cone has to be at least 75% of
the opening diameter the device [27].

42
4.3. Scavenge System Design
The scavenge air exhaust system is the part of the separator that pro-
vides the means to extract the scavenged dirty air. Design drivers include
providing sufficient suction, sufficient flow capacity, acceptable operating
life, acceptable weight, good reliability, overboard exhaust and integration
with engine and airframe, but the key requirement is the need to integrate
the scavenge system into the basic separator [19]. The scavenge system is
comprised of ducting from the conduit at the splitter, an impeller to provide
suction, and an exhaust, but the arrangement and sizing of these will vary
depending on the engine. However a patent study has revealed the great
scope of possible scavenging methods.
The separator featured in U.S. Patent 3,148,043 [27] is an example of
one in which the dirty air is scavenged away tangentially. The dirty air does
not continue in an axial direction but instead is exited tangentially with the
aid of an impeller. While this design is compact and does not carry the
same integration demands as axial exhausts, there is a greater requirement
of power from the suction source to duct and to change the direction of the
scavenge flow. Nonetheless, the compactness has great advantages, not only
benefits in space but also weight. The difficulty arises in arrangement of
the ducting to achieve this advantage, which can yield especially complex
solutions if the separation system is complicated, such as in U.S. Patent
6,698,180 [28]. In this case, a splitter first divides the flow into two streams,
which are then imparted with a radial component of velocity, the upper
radially outward, the lower radially inward. This flow splitter is labelled
50 in Fig. 10. It is surrounded by an inner cowl 34 and outer cowl 35,
which thus respectively define an inner and outer IPS device of conventional
arrangement. This “double” separation technique raises the efficiency, but

43
 Figure 10: IPS with tangential scavenging means according to Snyder [28].

requires a complicated scavenging system, as may be invoked from Fig. 10.

The scavenge chamber 38 of the inner separator receives dirty air, which is
routed through one of a plurality of hollow support struts 124 to a scroll duct
122. The scroll duct exhausts the dirty air tangentially to the atmosphere
under the influence of a blower or eductor. The process is the same for the
outer separator.
One risk associated with this type of extraction, is the possibility that
particles in the scavenge chamber may become re-ingested in the clean air

44
 Figure 11: IPS with anti-rebound vanes in scavenge chamber according to Hull [29].

flow. There are inventions that employ a plurality of overlapping vanes in

the scavenge chamber, such as in U.S. Patent 3,832,086 [29] shown in Fig. 11.
This configuration relies upon swirl vanes prior to separation creating cy-
clonic flow in the separator, but stops particles rebounding back into the
airflow and scavenges away the dirty air relatively quickly, in comparison to
scroll type evacuation means.
In other embodiments, the scavenge system discharges the dirty air axi-
ally. However, owing to space and weight limitations, the scavenged air often
first enters an annular chamber, before exiting through a single exhaust duct.
This means the mass flow will vary circumferentially with distance from the
exhaust vent, which can lead to unequal extraction rates and a risk of re-
ingestion from those circumferential locations experiencing the least suction.

45
Furthermore, the varying mass flow distribution affects the circumferential
pressure distribution of the clean air flow, which replicates itself at the com-
pressor inlet. An uneven pressure distribution at the compressor inlet leads
to lower efficiency. One solution to this problem is to create a labyrinth of
ducting, with a plurality of vents spread circumferentially which connect to
the exhaust vent through ducts of the same length [30]. This ensures that
each vent experiences the same pressure differential, which thus extracts the
same mass flow from each portion of the annular scavenge conduit. This
may require a higher rated impeller, but the result may atone for this.
The requirement of an impelling device, such as a blower, eductor or
pump is a disadvantage that most IPS devices carry in comparison with
other air cleaners. The dirty air is usually fluidised with approximately 15%
to 30% of the influent air, depending on the device. This figure is important,
since drawing too much or too little air could affect the separation efficiency
or the engine performance respectively. The use of an impelling device
ensures extraction of a predetermined amount of influent air by setting up a
pressure differential between the device and the scavenge chamber. Power is
drawn off the engine to operate the impeller, which reduces its performance
capabilities. One method of negating this is to use the exhaust to set up
the pressure differential. The particle separator described by McAnally [31]
provides suction by disposing a nozzle in the engine exhaust, parallel to
the flow, connected by conduits to the scavenge chamber at the separator.
A vacuum created in the engine exhaust draws the dirty air through the
conduits, which is mixed and discharged with the exhaust gases and has the
added benefit of a cooling effect.
While such a configuration allows extraction without power, there are
problems. For example, the engine exhaust must protrude further than

46
is necessary, to provide enough length for the two air streams to coalesce
and generate sufficient vacuum. This adds to the engine length, increasing
the weight and cost which offset the saving of not using a powered suction
source. However, if the scavenge air and exhaust gases are mixed better,
the potential of this method could be maximised. The particle separator
in U.S. Patent 4,265,646 [32] connects the annular scavenge chamber at the
separator via two conduits to an annular exhaust which is pleated with the
exhaust nozzle of the engine, creating a series of rivulets. Due to the large
shear forces introduced by the large boundary interface, appropriate mixing
is achieved between adjacent rivulets in a short distance.

4.4. Hump and Splitter Design

The arrangement of the hump and the splitter are delicate and essential
design parameters which have a great influence on overall separator perfor-
mance. The hump, or peak of the inner body, provides the means to turn
the flow, and the splitter divides the air flow into clean and dirty streams.
It is generally agreed that the spacing of these two components and the an-
gle through which the flow is turned dictates the separation efficiency and
pressure drop. For example, in U.S. Patent 6,508,052 [28], it is said that:
“The distance between the splitter and peak of the inner body affects the
size of particulate for which separation occurs. A larger distance generally
correlates to less particle separation of smaller particles... In addition, the
rate at which a particle separator turns the air also affects the size of the
particulate for which separation occurs. Slower turning of the air generally
correlates to a less particle separation of smaller particles”. This is mirrored
in U.S. Patent 6,702,873 [33]: “Particle separation efficiency may be mini-
mally increased by decreasing the size of the clean intake channel opening

47
 Figure 12: IPS design parameters according to McAnally [31].

or by decreasing the angle between a fluid inlet portion of the system and
the hub.” So by varying these parameters, an optimised separating system
can be devised.
A good example of how the splitter/hump spacing affects separation and
system efficiency (pressure loss) is provided in U.S. Patent 3,766,719 [31].
According to research carried out for this invention, for a good balance of
the two efficiencies, the axial distance between the splitter and the hump
should be substantially greater than the radial distance between the two, as
measured from the central axis. These dimensions are illustrated in Fig. 12,
where A represents the axial distance and B, the radial distance. Dimension
C is the channel width.
The slope of the surface 20 is also an important parameter, for deter-
mining the angle through which the flow turns. A steeper slopes achieves
greater separation efficiency, but also increases pressure losses whilst intro-
ducing the risk of flow separation. It can be invoked from Fig. 12 that

48
since the dimension C must be of a pre-determined value for proper engine
bound flow, the ratio A/B indirectly establishes the preferred slope of the
wall 46. So any changes to this ratio will greatly affect the performance of
the separator.
Tests performed on this particle separator reveal the effects of altering
the ratio of A/B, whilst providing an insight into the performance of such
devices. According to the patent, a ratio for A/B of 2.0 achieves a separation
efficiency of 88.0% for a pressure drop of 1.30 kPa; a ratio of 4.0 achieves
83.5% and suffers a pressure drop of 0.62 kPa. An acceptable compromise
was deemed to lie between these values – a separation efficiency of 85%
for a pressure of 0.75kPa is equivalent to a system efficiency drop of just
three-quarters of percent, and is achieved with a ratio for A/B of 2.5. This
trade-off is common to all air cleaning devices.
Another observation in U.S. Patent 3,766,719 [31] is that increasing the
flow velocity up to the annular throat at the hump leads to greater separation
efficiency. This is owed to inertial effects. Faster flowing particulate-laden
air has a higher momentum, which causes a greater portion of the flow to
be scavenged, and centrifuges a higher mass of particles. For the same sep-
aration efficiency at lower speeds, a greater portion of scavenge flow would
be required, which may not be possible due to engine requirements. How-
ever, increasing the flow speed makes turning the air more difficult, thereby
increasing pressure losses. For example, tests show that a throat velocity of
47.8 m/s results in a pressure loss of 0.60 kPa but requires a 40% scavenge
flow to achieve 85% separation efficiency. Meanwhile, a throat velocity of
73.2 m/s leads to a pressure loss of 1.5 kPa but only requires 17% scavenge
flow to achieve the same separation efficiency, of 85%. The velocity at the
throat at zero airspeed is determined by its annular cross-sectional area.

49
Recalling the influence of splitter/hump parameters on flow efficiencies, it
can be appreciated that small alterations to the geometry of the separator
can greatly affect its performance

4.5. Recent Novel Improvements

It is evident that the trade-offs described above are delicate compromises,

based an average inflow conditions of stable, settled air. In reality, the
influent air to a helicopter engine is erratic and non-uniform, thanks to
varying flight conditions and altering downwash flow. It has been shown that
flow conditions inside the particle separator are determined by the geometry
of the device, which dictates the performance of the separator. Therefore the
performance of an IPS is likely to be influenced by in-flight inlet conditions.
It is this problem which is driving the most recent developments in IPS
design.
By incorporating new technologies, inventors have been able to propose
IPS designs which can adapt to optimise performance during operation. In
all cases, there needs to be a detection system employed, to understand
what changes are needed. This may be a side optical device, a velocimetry
device, a thermocouple, a “sand sniffer” along with a particle analyser, or
some other particle sensing device placed aft of the splitter in the clean air
duct. For example in U.S. Patent 6,702,873 [33], a controller is electronically
coupled to such a sensor, and can fine tune the efficiency and power usage by
adjusting a fluid parameter such as channel wall shape, splitter length, clean
fluid intake opening size, or other fluid parameters. The parameter adjusting
system controls the hub and the splitter, both of which are manipulated via
shape memory alloy actuators and worm gears. As the worm gears rotate,
a series of plates realigns to create different rigid shapes. This is known as

50
an adaptive surface, and allows the power consumption by the impeller and
pressure losses to be decreased when higher travel speeds cause a greater
discharging of scavenge air as a result of inertial effects.
In a similar vein, European Patent 1,908,939 [34] employs an inflatable
boot at the hump, to create a temporary “flow path boot” which narrows the
annular throat area. This is depicted in Fig. 13. When the gas turbine engine
is operating in relatively clean environments, the boot remains deflated, flush
with the inner wall 36 of the central body. In incidences of high sand or dust
concentrations, control valve 54 is opened, which feeds bleed air from the
compressor into the flow path boot 50. The new contoured surface of the
separator system creates a sharper turning angle for the particulate-laden
air, and causes a greater acceleration of the influent air. This contributes to
greater separation efficiency, at the expense of an increase in pressure loss.
However, the air that is pumped into the flow path boot to retain inflation, is
discharged through tangentially orientated exit slots, on the leeward surface
50, which returns energy into the boundary layer causing clean fluid flow to
adhere more closely to the top surface 51. Eliminating the risk of separation
also eliminates the risk of flow distortion at the engine inlet.
Another use of flow control is found in U.S. Patent 7,296,395 [35], which
incorporates a zero mass active flow device 100 just aft of the hump, as
depicted in Fig. 14. This keeps the boundary layer adjacent to the inner
wall 226 attached and energised, by pulsing over a range of frequencies, at
a small angle β to the flow direction. In this embodiment there are 5 such
devices, arranged co-radially, which can allow a greater flow turning angle
without the risk of separation at certain inflow conditions

51
 Figure 13: Adaptive IPS according to European Patent 1,908,939 [34].

Figure 14: IPS with active flow control according to U.S. Patent 7,296,395 [35].

52
4.6. IPS Theory
This type of separator is more amenable to a computational fluid dy-
namics modelling; in fact, all the published technical literature on particle
separators focuses on IPS systems. Since ultimately the IPS work on the
principle of separating the particulate and the “clean” air through bifur-
cating tubes, the prediction of phenonema such as collision and rebound
are essential. One such example of theory is available in Hamed et al. [36].
These authors used a combination of deterministic and stochastic particle
bounce models with Lagrangian tracking on a fully turbulent solution of
the Reynolds-averaged Navier-Stokes equations. The particle’s path was in-
tegrated within the RANS solution up to the collision or bouncing point.
The rebound was stochastic and produced new initial conditions for particle
tracking. To calculate the separation efficiency, several thousand particles
have to be tracked from the inlet. The work of Vittal et al. [37] follows a sim-
ilar approach, and focuses on the concept of a “vaneless” separator. Particle
paths were predicted through a Lagrangian tracking, although the method
only accounted for boundary layer corrections. The comparisons with tests
indicate that the separation efficiency was up to 90% with a fine sand and a
scavenge flow rate in excess of 14%. For a coarse sand, the efficiency was 5%
lower. Musgrove et al. [38] addressed the computational design of an IPS
with louver-type channels placed downstream the combustor, and a collec-
tion chamber for the separated particles. However, this technology does not
appear to be useful for turboshaft engines, for which the particle separation
must be upstream the compressor.
Saeed & Al-Garni [39] developed a numerical method based on an in-
verse design approach. The method uses various levels of approximations,
including the reduction of the IPS to a set of airfoil-like bodies. The parti-

53
cles are tracked with a Lagrangian method. Crucially, this theory lacks an
appropriate model for particle rebound from a solid surface, and is based on
an inviscid flow model, an unlikely event in the best of cases.
The most recent theoretical work on IPS is that of Taslim et al. [40].
These authors used CFD methods coupled with particle dynamics to predict
the scavenge efficiency of a conventional inlet design. The main contribution
of this work was the model of the particle impact, in particular the restitu-
tion coefficient and the inelastic effects. They concluded that extremely fine
particles (smaller than 10µm) cannot be practically separated.
It is evident that the field of IPS analysis by computational fluid dynam-
ics modelling is ripe with academic study. This is due to the ease with which
IPS devices are suited to this type of analysis. However, despite the scope for
geometrical changes and the various solvers that be used, the field appears
rather limited in diversity. In all cases reviewed, including those not cited
in this paper, the particle separator is axial type, and the separator means
as per conventional hub and splitter arrangement. Yet, as discussed earlier,
in reality there is a plethora of IPS designs, ranging from radial inlets to
vaned scavenge chambers. Furthermore, it can be invoked from such papers
that the performance of a hub and splitter are very much dependent on the
local flow conditions, which in turn are determined by the engine mass flow
requirements. Therefore, each study case is limited to the engine for which
the IPS system is designed. The sensitivity of IPS design to local conditions
renders universal analysis even more difficult when considering the real life
situation, in which the particulate will undoubtedly differ from those test
sands used to verify CFD data. While this may be a common problem in all
areas of particle separator analysis, it highlights the case-specific techniques
that are needed for IPS theoretical anlysis. Therefore, it is concluded that

54
IPS theory is not conducive to the more holistic approach adopted in the
analysis of VTS and IBF technologies.

4.7. Summary of IPS Technology

IPS are currently the most compact, lowest weight system of foreign ob-
ject damage prevention and engine protection. The distinct advantage over
their counterparts in the field of intake protection is their compactness and
low pressure loss, which they sacrifice for lower separation efficiency. This
can be judged by looking at the volume flow rate per unit area of intake
protection. For example, vortex tube separators can achieve airflows of just
0.5 to 1 kg/s per square metre, as opposed to IPS devices which typically
achieve airflows of over 5 kg/s [41] per square metre. This translates to a
fivefold increase in intake area, which has an obvious drag penalty. This
advantageous feature of IPS devices becomes important when considering
the additional frontal area that is needed to allow for the relatively high
scavenge flows of 15% to 30% intake air. Pressure losses are small in com-
parison with other protection. However, typical separation efficiencies can
range from 50% to just 85%, which is a substantial shortfall in comparison
to IBF and VTS devices.
IPS are mostly designed by the engine manufacturer, as an integrated
device. This is because system geometry is highly influential on local flow
conditions, therefore accurate contouring with the compressor inlet is nec-
essary. IPS are found either as axial separators added to the intakes of
front-facing engines, or are incorporated internally if the engine has a radial
inlet. Both utilise the inertial separation technique, but have advantages
over one another. Most notably, front-facing IPS devices can recover some
lost pressure through ram effects, and have inlets which are located further

55
away from the engine exhaust, reducing the risk of hot gas re-ingestion. The
same cannot be said of radial intake engines, but they are generally more
compact, which reduces weight and axial length. The choice of such a device
depends upon the power and therefore engine requirements of the rotorcraft
in question.
Finally, it has been shown that the IPS design is a balance between
separation efficiency and pressure loss. Parameters that affect these char-
acteristics concern the device geometry, most notably the hump shape and
splitter shape. New technologies, such as active flow control and adaptive
surfaces, are achieving optimised performances whilst in operation, modify-
ing the device structure with feedback analysis and intelligent systems.

5. Inlet Barrier Filters

Inlet barrier filters are employed throughout the rotorcraft industry to

provide protection to engines from potential blade erosion and FOD caused
by particulate ingestion. They were first developed in the 1960s for the U.S.
Army helicopters operating in South East Asia, although this technology
was later abandoned for the IPS design, as described in § 4. The method of
particle extraction differs from other devices such as the IPS, by physically
stopping particulate matter with a porous screen. This technology yields
exceptionally high separation efficiencies of up to 99.3%, for a relatively
low initial loss in pressure. However, this loss increases with time, owing
to the gradual formation of a cake upon the surface of the filter, which
further impedes the flow. To reduce this problem, the filter is inspected
and maintained regularly with wash cycles; eventually, it is replaced. The
filter element is comprised of some layers of woven cotton, sandwiched in a

56
pleated wire mesh. IBF systems are easily adapted, a fact that, along with
their exceptionally high separation efficiency, contributes to their growing
attendence on many modern rotorcraft.
The IBF is installed at the engine intake of a helicopter to filter all
engine-bound air. On older models of helicopter, such as the UH-60 Black
Hawk, they may be added as an appendage; on newer models, such as the
Eurocopter AS 350, they are designed into the airframe as a fully integrated
device. To reach the engine inlet, air must pass through what is known as
a barrier filter. A barrier filter is a panel typically comprised of a blended
fabric, pleated between two layers of epoxy-coated wire cloth. Particles of
size larger than the fabric pores cannot pass through the filter, and remain
on the medium until servicing.
Any restriction to the airflow at the inlet of an engine will result in a
performance loss, regardless of what measures are put in place to reduce
this effect. Engines are designed to receive a consistent and stable flow of
clean laminar air; deviations from this state prevent the engine operating
at full performance. Therefore, a first design consideration is to ensure that
the filter element is properly sized to permit an adequate quantity of air to
the engine without a large loss of pressure. This is a general design driver
for all types of intake protection. Unfortunately, pressure loss is inevitable,
although a worse state for the engine arises from fluctuating pressure losses.
The many modes of helicopter flight cause different levels of ram pressure.
For example, when a helicopter transitions from forward flight to hover, the
loss of ram pressure causes a drop in engine efficiency, which is detrimental
to the stable operation of the engine. Therefore, it is best to design a filter
whose performance is relatively independent of the directional motion of
the helicopter. The type of rotorcraft to which the filter will be applied

57
is also a major consideration for designers. If an existing helicopter is to
be retro-fitted with intake protection, an important factor of the design is
its integration into the airframe. Drastic changes to the airframe contours
could require re-certification of the rotorcraft, which is both expensive and
time consuming. A solution that blends into the existing airframe would
therefore be ideal. One risk that can be reduced by blending the filter with
the airframe is that of icing, a condition in which ice builds up and blocks
pores in the filter, causing a loss in performance that may worsen over time.
Ice will build up more readily on forward-facing surfaces. Therefore, a way
of reducing this effect is to orientate the filter so that the flow is parallel to
the oncoming airflow.
The term “inlet barrier filter” is applicable to a product that consists of
both a filter medium and a means of attachment to the aircraft. In addition
to the filter, an IBF consists of a cowling to replace an existing section of the
airframe (or be incorporated into a new design), a frame with attachment
points, and often a hydraulic-powered bypass door to allow unrestricted air
into the engine in case of filter failure.
The filter itself is a panel of pleated material, sandwiched between two
epoxy-coated wire meshes. The pleat provides a large surface area, whilst
the wire mesh provides reinforcement against foreign object damage and
protects the filter material. Several combs provide structural support to
the filter, to retain its shape. The filter is mechanically bonded to the
frame through adhesion or physical connection, and is sealed with a potting
material, which prevents unfiltered air from seeping through gaps in the
join. Once the air has passed through the filter panels, it reaches a chamber
from where it is further drawn into the engine via aerodynamically-profiled
ducting, as depicted in Fig. 15.

58
Figure 15: Example IBF System for Eurocopter AS 350, with open bypass door and
ducting to engine.

59
 A pressure differential sensor monitors the pressure across the filter and
notifies the pilot when the pressure loss breaches a limit set by the manu-
facturer. In the event of the filter becoming over-clogged, a hydraulically-
powered bypass door can be opened to allow unrestricted air into the engine.
This opening is often rearward-facing, to minimise the risk of foreign objects
reaching the engine in this unprotected state.

5.1. Filter Element Design

The filter is the most important piece of equipment in the system, since
increasing the separation efficiency by just a few tenths of a percent can
drastically improve the overall performance of the engine. In addition to
providing high separation efficiency, the material must also be resistant to
damage by water and other liquids it main encounter. The filter medium is
often manufactured from polyester, felt, or most commonly, woven cotton.
In the case of cotton, the filter is constructed from three to six overlapping
layers, woven into a grid pattern. To further improve efficiency, the filter
may also be impregnated with oil, which not only helps to capture finer
particles by providing a “tack barrier”, but also functions as a good indicator
of usage by changing from red or green to brown or black, with increased
contamination. The use of oil also bestows upon the filter an ability to repel
water, which helps to prevent absorption and prolongs life.
The sizing of the filter is important. It has been found that the pressure
drop across a filter is reduced by increasing the effective surface area of the
filter. Furthermore, it has been shown that in order to achieve optimal life in
erosive environments, the filter should be sized such that the mean velocity
of the air approaching the filter element at the engine’s commercial Take Off
Power (TOP) is less than 30 ft/sec (9.1 m/s), and preferably in the range of

60
about 15 ft/sec (4.6 m/s) to 25 ft/sec (7.6 m/s), as reported by Scimore [42].
The velocity of influent air is determined by calculating the volumetric flow
rate per unit of effective filter surface area. Therefore, if the volumetric flow
rate of the engine is known, the abovementioned mean velocities can be used
as a ratio value to appropriately size the filter element surface area.
The total surface area of a pleated filter is the total area of a single
pleat multiplied by the number of pleats. Sizing may be accomplished by
altering the pleat height, the pitch, or altering the shape of the filter within
the confines of the opening (for example, by curving the surface). A typical
pleat height lies in the range of 1 to 3 inches (2.5 to 7.5 cm), and pleat
spacing may be 3 to 6 pleats per inch (e.g. up to 2.4 pleats/cm). Filter
elements are typically sized with a sixfold total surface area over profile
area. As well increasing the filter surface area for no increase in profile area,
pleating has the added benefit of creating structural rigidity within the filter
element.
The filter stiffness can be further improved by employing a structural
comb, as in U.S. Patent 6,595,742 [42]. The comb has a plurality of teeth
extending across the filter as shown in Fig. 16, which mesh with the pleats.
This helps to maintain filter shape, which allows for uniform flow area and
low, stable pressure drop. There may be a number of combs employed
depending on the filter size, but this may be limited by space required to
facilitate cleaning.
Another contributor to structural improvement is the addition of a wire
mesh to the filter medium, which also helps to reduce abrasion damage. It
is common that the layers of woven cotton are sandwiched together with
an epoxy-coated wire mesh, manufactured from a corrosion- and abrasion-
resistant material such as stainless steel or aluminium. A typical mesh size

61
Figure 16: Pleated filter element with structural comb, from U.S. Patent 6,595,742 [42].

is 60% to 70% open area, as exemplified in U.S. Patent 6,595,742, with a

preferred spacing of 16 wires per inch. The thickness of the wire typically
lies in the range of 0.005 and 0.015 inches (0.02 to 0.06 mm).
Due to there being no self-cleaning or particle removing means installed
in an IBF, the device requires regular maintenance. It is the nature of
IBF systems that with time the filter element will become clogged, and will
continually lose performance due to an increase in pressure loss. To monitor
this, a pressure sensor is placed across the filter element, which measures
the plenum pressure against the ambient pressure.

5.2. Effect on Performance

The overriding benefit of installing an IBF is of course the removal of

particulate from influent air, which would otherwise be particularly trouble-

62
some to an engine. One manufacturer claims to have achieved a laboratory
filtration efficiency of 99.28% with SAE Coarse Test Dust. When replacing
an EAPS system, there are significant improvements to helicopter perfor-
mance. For example IBF systems are both lighter and significantly cheaper
than EAPS systems, and unlike an array of vortex separators, cause little
distortion to the air flow. In fact, it has been said that the arrangement of
pleats actually contributes to stabilising the air flow.
The main pitfall of IBF is a deterioration of performance with time. As
a filter element becomes clogged, it becomes increasingly difficult for the air
to pass through, which leads to an increase in pressure loss. When clean air
is passed through a new filter element, the pressure loss is typically around
0.5 kPa. After 100 hours in a dust-laden environment, this pressure loss
may typically rise to around 3 kPa, which is greater than the pressure loss
suffered by EAPS systems. This pressure loss over the life time of the filter
is discussed further in § 5.4.

5.3. Applications

The application of IBF is evident on helicopters of all type and size.

In the past, rotorcraft such as the UH-60 Black Hawk may have been con-
structed without intake protection in mind. Consequently, the airframe
has had to be modified to incorporate a barrier filter system. U.S Patent
7,192,562 [43] for example pertains to an engine air filter and sealing system
for the UH-60 Black Hawk, which is simply added to the airframe ahead of
the engine inlets, as illustrated in Fig. 17. The addition is a box comprising
three filter panels and a front-facing bypass door.
The blending of intake protection with the airframe is particularly dif-
ficult for larger rotorcraft, as the greater power requirements are typically

63
Figure 17: Engine Air Filter for the Sikorsky UH-60 Black Hawk, from U.S Patent
7,192,562 [43].

64
met by employing two large engines which invariably “jut out”. In contrast,
smaller rotorcraft, in particular those of a single engine can submerge the
engine within the body. Intakes in the sides of the airframe provide the
engine with sufficient air and in many cases can easily be fitted with intake
protection. Engine manufacturers may leave a plenum chamber ahead of the
engine intake such as in the Bell 206B, which allows space for intake protec-
tion to be installed to the customers needs, be it an IBF or EAPS system.
In this helicopter, the intake protection is found submerged within the air-
frame. In other embodiments, such as the MD 500, the filter is blended into
the sides of the airframe. Such a choice allows for a greater filter planform
area, which may be required for larger engines.

5.4. Barrier Filter Theory

There are no technical publications in the open domain on IBF systems.

Thus, it is not possible to extract engineering information on any of the
systems described. By contrast, the subject of filtration itself is vast and
well documented for a wide array of engineering applications. Our knowledge
of IBF systems if thus inevitably derived from such a discipline.
There are two main branches of filtration. Depth filtration is a process
whereby particles are separated throughout the depth of medium as the dirty
air permeates through. They are often composed of a pad of fibres or a bed
of sand, and rely on the adherence of particles to the filter constituents.
As a result of the contaminating nature of this process, depth filters are
not re-usable. This is in contrast to the second type of filtration, known
as surface filtration (or entrapment mechanics), in which the filters behave
more like a sieve. Separation is dependent on the size of the perforations in
the medium. Particles too large to pass through begin to build up on the

65
surface of the filter and form a “cake”, as illustrated in Fig. 18 (this term
will be used throughout). The cake temporally grows, further improving the
separation capabilities of the filter to the detriment of an increasing pres-
sure loss. However, unlike depth filters, the cake can be removed from the
surface of the filter, rendering this type of filter re-usable. For this reason,
and their high separation efficiencies of almost 100%, surface filtration is
the process employed in IBF elements. Unfortunately, the high separating
ability of barrier filters is offset by the consequential loss of pressure across
the medium.
There are several factors used to determine the effectiveness of a surface
filter — the loss of pressure is one of these. The lower the pressure drop,
the better the filter. In IBF theory it is necessary to consider the presence
of cake on the filter, as this augments the filtration process and is an impor-
tant factor in determining overall filter performance. The pressure loss can
be attributed to frictional losses that occur as the air passes through the
interconnected pores within the filter medium or cake. The factors which
affect this process feature in Darcy’s Law for horizontal, laminar flow, which
states that the pressure drop across a homogeneous filter medium is directly
proportional to the filter thickness, the fluid volumetric flow rate and fluid
viscosity; while inversely proportional to the filter area and a factor k known
as the permeability coefficient. Permeability is defined as a measure of the
ease with which fluids pass through a porous medium. It has many applica-
tions, most notably in earth sciences for determining the flow characteristics
of hydrocarbons in oil and gas reservoirs. Whilst the other factors in Darcy’s
law can be attributed to either flow properties or dimensions of the filter,
the permeability is a direct property of the medium matrix. According to
Bear [44] the relevant solid matrix properties are mainly grain (or pore)

66
size distribution, shape of grains (or pores), tortuosity, specific surface and
porosity. These properties are very difficult to measure, especially in the
case of a cake comprised of such a wide variety of particulate of unknown
properties. For this reason, many studies have arisen to formulate empir-
ical, semi-empirical or theoretical methods to determine the permeability
coefficient in the formation of cake.
Nevertheless, it is possible to ascertain the effect of certain cake charac-
teristics on the permeability of a porous medium and thus the corresponding
pressure drop for given flow conditions. For example, a study by Chen &
Hsiau [45] reports that a larger superficial velocity increases the compres-
sion stress on the cake, leading to a greater compaction and thus resistance,
the consequence being a higher pressure loss. Conversely, it was found that
an increase in the average thickness of even just 1mm could result in an
increase of 0.686% in collection efficiency. Despite this experiment being
performed with dust from a coal-fired power station, the principle is evident
for extension to other particulates.
In contrast to cake studies, the permeability of the barrier filter medium
is considered a set value, determined by the material, the weave, and the
yarn properties. The permeability of a fabricated cloth is based on the flow
of air, and usually quoted in cubic feet per minute per square foot, for a
pressure drop specified in inches of water, usually 0.5 inches (0.75 cm) as in
the widely used “Frazier Scale”. Put differently, this value evaluates a filters
ability to transmit air under a prescribed pressure gradient. It disregards
the filter thickness and air viscosity, and is quantified empirically by appli-
cation of Darcy’s law. The permeability of a filter is often considered as
constant, although Rushton (1977) (as cited by Purchas & Sutherland [46])
reports that it may decrease with time as a result of a phenomenon known

67
as blinding, whereby particles become trapped within loosely woven yarns
and evade removal by cleaning thereby increasing the resistance.
The filtration medium used in modern-day IBF is woven cotton. Ad-
ditionally, it is sandwiched between two epoxy-coated wire meshes, which
provide some protection against FOD, but mainly help to support the filter
in its pleated configuration. Another parameter influencing the performance
of filters in the “Dirt Holding Capacity”, which is the ability of a filter to
retain solids while operating below a pre-defined pressure drop. A larger
dirt-holding capacity increases the maintenance interval of the filter, and
therefore service life. Both the pleats and the mesh may influence this as
well as the fabric material. However in first order analysis, we are ignoring
the pleats and mesh, instead modelling the filter as an equivalent flat filter
six times the projected area of the barrier filter element.
IBF filters are often positioned flush with the airframe. Therefore, the
ram effects of forward flight can be assumed as negligible. Filters are some-
times placed immediately below the rotor disk. However, this is often within
a radius where the downwash is low, and therefore the ram effects can be
assumed as negligible here too. Therefore, in our first order analysis, we
can assume that the intake volumetric flow is dictated solely by the engine’s
requirements, and that the static pressure at the filter inlet is that of the
standard atmosphere. Furthermore if the mass flow rate of the engine is
known, the intake velocity can be found for a given filter area, as volume
flow rate per unit area. It is known from an IBF filter manufacturer that
the optimum intake velocity in a compromise for low pressure drop, is 10
m/s.
A practical model for the filter is based on the application of Darcy’s
law. The mass flux through a filter of area A is

68
 kA p1 − p2
Q̇ = (32)
µ ∆x
where k is the permeability of the medium, ∆x is the thickness of the filter
and µ is the dynamic viscosity of the air. This equation, applied to the case
shown in Fig. 18, leads to a velocity U through the filter

   
p1 − p2 k p2 − p3 k
U= = (33)
µ ∆x f ilter µ ∆x cake
By eliminating the intermediate pressure p2 from Eq. 33, we find

p1 − p3
U= (34)
µ [(∆x/k)cake + (∆x/k)f ilter ]
It is generally assumed that the permeability of the filter does not change;
hence, the performance of the filter depends on the characteristics of the
cake, or its resistance

∆x
α=µ (35)
k
As the thickness of the cake increases over time, the pressure drop at the
engine’s inlet increases. The change in inlet velocity U is more modest,
and in principle there should not be any change for an engine working at
constant mass flow rate.
The change of the thickness of the cake with respect to the time of
operation is calculated from a mass balance equation:

Vc
∆xc = (36)
Af
where Vc is the volume of the cake. Note that Eq. 36 shows the effective
filter area Af , rather than the inlet area A of the engine. Data gleaned from
filter manufacturers indicate that Af /A ≃ 6.

69
 The mass can be calculated from the volume of separated dust or par-
ticles. If η is the collection efficiency of the dust or particles, and if the
underscore “ps” denotes the separated particles, then

mps ṁps
η= = (37)
mp ṁp
With this definition of separation efficiency, Eq. 36 becomes

Vps mps
∆xc = = (38)
Af ρc Af
where ρc denotes the average density of the cake. The density of the cake is
deemed to be lower than the compacted particles. Uncompacted sand can
have a density as low as 1,200 kg/m2 . The change in thickness over time is

d∆xc ṁps η ṁp ηc ṁ

= = = (39)
dt ρc Af ρc Af ρc Af
Equation 39 is valid only as long as all the separated particles remain trapped
in the cake; in fact, this is unlikely to happen, since an indeterminate portion
will probably bounce back or fall off when hit by incoming particles, or are
even dusted off by external factors, such as the rotor downwash and the
effects of gravity. These events greatly complicate the prediction of cake
growth, and thus the pressure losses through the IBF. In order to include
this effect, if f denotes the fall-off and rebound rate (0 < f < 1), then the
actual thickness growth is

d∆xc ηc ṁ
= (1 − f ) (40)
dt ρc Af
The transient problem can be solved as a system of ordinary differential
equations. The unknows are: the thickness of the cake, the inlet flow ve-
locity, the pressure drop, the separation efficiency and the permeability of

70
 Figure 18: Flow through a surface filter.

the cake. In first order analysis we assume that the latter parameter is a
constant. Also, it is reasonable to assume that the inflow velocity is not
affected by the thickness of the cake, because it depends on the mass flow
requirement of the engine (assumed to be constant; no ram pressure effects).
These assumptions lead to a considerable simplification, since we only have
two differential equations to solve: Eq. 39 and a differential form of Eq. 34.
Further assumptions are required. First, the manufacturers of IBF report
(although they do not prove it) that the efficiency of separation is constant;
sometimes the efficiency is improved by the presence of the cake. Thus, the
separation efficiency η is a known constant. Second, there is no information
to assess key characteristics of the cake, such as porosity and consistency,
although data are available for sand beds, hydrology and such applications.
Third, there is no information to assess the effects of particle rebound and

71
fall-off.
The solution of Eq. 39 with a constant permeability of the cake leads to
a thickness increasing uniformly with the time, according to

ηc ṁ
xc (t) = (f − 1) t (41)
ρc Af
With the average data known, we estimate

ηcṁ
≃ 2 · 10−7 [m/s] (42)
ρc Af
which would lead to a cake build-up of the order of 0.72 · 10−3 m/hr (0.72
mm) under dust concentrations c ≃ 10−4 , and in absence of rebound and
fall-off events. The actual pressure losses as a function of the operation time
depend on the choice of the parameters in Eq. 39. Now call

"    #
∆x ∆x
R=µ + (43)
k c k f

Differentiation of Eq. 34 yields

dU 1 d∆p ∆p dR
= − 2 =0 (44)
dt R dt R dt
or

d∆p ∆p dR
= (45)
dt R dt
The solution of this equation is nearly straightforward. If the underscore
“o” denotes the clean filter at time t = 0, then:

∆p (∆x/k)c
=1+ (46)
∆po (∆x/k)f

72
Hence, the increase in pressure drop is linear with time, and its gradient
depends on the combination between cake and filter characteristics. The
exact physical behaviour of the permeability with the pressure depends on
the type of medium, as discussed by de Nevers [47]. Even in the event of a
clean filter, we need to find an estimate of the initial pressure drop ∆po .
An estimate of the permeability of the filter can be found from the
solution of Eq. 32 in terms of k, for a known filter area, and initial pressure
drop (taken as a value quoted by manufacturers)

  
∆x ṁ
k= µ (47)
∆p ρAf
Equation 47 contains the inverse of the pressure gradient through the filter,
the mass flow rate per unit area and the dynamic viscosity of the air. The
parameters to be used in Eq. 47 are: ∆p ≃ 0.5 kPa; ∆x ≃ 5 · 10−3 m;
ṁ/A ≃ 20 kg/s/m2 , Af ≃ 6A. With these data, the filter permeability
is k ≃ 6 · 10−10 m2 at standard sea level conditions. However, it must
be pointed out that there can be considerable variations around this value
of the permeability, as a consequence of different filter pleats, wire mesh,
lubrication and atmospheric humidity.
An estimate of the permeability of the cake is given by the equation
(Krumbein & Monk, as cited by Bear [44])

k = 0.617 · 10−15 d2 [m2 ] (48)

where d is the particle’s mean diameter in µm. The permeability of a cake

corresponding to a particle mean diameter of 100 µm is of the order of
10−11 , or 10 Darcy (1 Darcy ≃ 10−12 m2 ). However, there can be variations
of at least two orders of magnitude in permeability due to particle size and

73
porosity, with an uncertainty of at least 10 Darcy. The increase in porosity
alone can increase the permeability by one order of magnitude. In absence
of technical data for this particular application, it is reasonable to assume
that the permeability of the cake is of the same order of magnitude as the
filter itself. This assumption can be verified by reverse engineering. In
fact, data from IBF manufacturers imply that the pressure loss increases
to about 3 kPa after 100 hours of operation. With the further assumption
of cake thickness ∆xc ≃ ∆xf , Eq. 46, solved in terms of the ratio kf /kc
yields kf /kc ≃ 5, or kc ≃ 0.2kf , kc ≃ 10−10 m2 . The result obtained is
about one order of magnitude higher than the permeability calculated with
Eq. 48, which in fact was derived for sedimentology applications. However,
if we also take into account the fall-off effects (as given by the factor f in
Eq. 40) the most reasonable estimate of the cake permeability appears to be
kc ≃ kf . The estimated pressure loss through the IBF is shown in Fig. 19
for three different values of the fall-off coefficient.

5.5. Summary of IBF Technology

IBF are widely used in the rotorcraft market due to their success at
prolonging engine life through successful filtration of contaminated influent
engine-bound air. The IBF System is installed ahead of the engine inlet,
forcing influent air through a porous filter element. The filter element is
constructed from three to six layers of material such as woven cotton, and
impregnated with a specially formulated oil which aids separation by act-
ing as a “tack barrier”, and helps to repel water. The filter medium is
sandwiched between two wire mesh screens which enables the element to be
pleated, a process which adds structural rigidity whilst increasing effective
filtration area sixfold. The performance benefits of intake air filtration are

74
 5
f = 0.9
f = 0.8
f = 0.7
4

3
∆p / ∆ po

0
0 20 40 60 80 100
Hours of filter operation

Figure 19: Estimated pressure loss in IBF as a function of operation time.

offset by a pressure loss of around 0.5kPa on first installation, which may

rise to 3 kPa as the filter becomes clogged with usage. Typical service life
can range from 1,000 to 5,000 hours depending on the environment in which
the IBF system is operating. The barrier filter should be serviced accord-
ingly every 50 to 300 hours depending on the environment of operation, to
clear the filter of any cake build-up.
The pressure loss associated with IBF is dependent on the permeability
of both the cake and the medium, and the dirt holding capacities of the
filter. In first order analysis the latter has largely been neglected, and several
assumptions have been made in order to predict the deterioration of pressure
with time. The initial drop in pressure due to frictional effects is based on
the permeability of the clean barrier filter, a property of the fabric of which

75
the medium is composed. According to the analysis, the pressure drop will
rise linearly with time, at a rate determined by what is defined as a “fall-
off and rebound factor”, utilised to consider the likelihood that particles
will bounce off or be dusted away by external influences. This analysis is
limited by the fact that such a relationship would allow the cake thickness
and pressure drop to grow with time, without severely impeding the flow.
Put simply, the filter would never clog.
In reality, this situation is improbable. Dickenson [48] indicates that
at constant working temperature, the pressure drop will, after an initial
non-linear jump, rise constantly with time eventually rounding into a much
steeper rise, the point at which the element has become too clogged for fur-
ther efficient working. He attributes the rise in pressure drop to a cumulative
build up of contaminants on the filter element which reduces permeability
in the direct proportion to the degree of contamination.
It is important to point out that the definition of permeability is often
ambiguous. In this instance, it includes the effect of cake thickness on flow
restriction which in Darcy’s Law is a separate factor in the calculation of
pressure drop, as shown in Eq. 32. In the previous analysis, the permeability
is regarded as a property of the porous medium matrix, the thickness and
area of which are regarded independently. It has been shown that pressure
drop is directly proportional to cake thickness, which could explain the lin-
ear portion of Dickenson’s prediction. The “knee” which evades prediction
in first order analysis is a consequence of a time-dependent permeability
coefficient.
The assumption of a constant permeability coefficient is actually based
on a number of assumptions that relate to the properties of the cake matrix.
Firstly, it is assumed that the cake is both homogeneous and isotropic,

76
properties which are unlikely to exist in reality. In the case of the former, the
particulate is unlikely to be composed of just one material; in the case of the
latter, as stated by Bear [2], in most stratified materials the resistance to the
flow is smaller along the planes of deposition than across them. However,
these factors are highly difficult to predict, and therefore in the event of
consideration could be rendered constant by statistical models. Instead,
a more likely cause of the temporal deterioration of permeability is the
spatial variation of porosity. With time, particles which may have been small
enough to pass through the barrier filter will find it increasingly difficult to
penetrate the cake, due to the presence of other particles whose interstices
are too small too pass through. A continuation of such a build-up of particles
will eventually lead to a situation where the all the pores are blocked and
air can no longer pass through the cake. Incidentally, this also improves the
overall separation efficiency of the filter.
Another disregarded factor that influences the deterioration of perme-
ability is the variation in the local flow conditions. In reality, owing to
changes in engine requirements and aircraft manoeuvres, the superficial ve-
locity is likely to alter. The study by Chen & Hsiau [45] showed that cake
resistance increases in direct proportion with flow velocity. This is due to
compaction effects, which serve to reduce the cake porosity, or cake density,
thus restricting the volume of free space through which the air can flow. In
a similar vein, temperature may not only influence permeation by altering
the air viscosity, but if low enough can lead to the formation of ice within
the cake. The consequence is a solidification of the cake, causing blockage
or partial blockage of the element and an abnormal pressure rise.
It is clear therefore that the coefficient of permeability k is no simple de-
terminer of filter performance. Owing to a wealth of influences, the cake is in

77
constant flux, leading to several permutations of composition. In the absence
of technical data for this particular application, it is impossible to consider
such factors, and simple assumptions have had to be made. However, it
has been shown that the pressure drop, permeability and cake thickness are
closely related, and are affected by local flow conditions and particulate con-
stituents. A further study may include consideration of the cake porosity
and composition, environmental factors, and if possible properties of the
filter element, thus permitting more accurate analysis to substantiate the
overriding trend that pressure drop increases with operation time.

6. Turboshaft Simulation

The various EAPS technologies discussed in this paper cause a variety

of conditions at the inlet of the turboshaft. VTS cause a moderate (even
negligible) loss in pressure, but a relatively large loss in mass flow rate
(the scavenge flow); IPS cause some loss in inlet pressure, accompanied by
a relatively large scavenge flow. IBF cause a relatively large loss in inlet
pressure, but no loss in inlet mass flow. Additionally, there can be some
power requirement to operate some of the systems. The IBF is notable for
not requiring engine power in the form of air bleed.
In this section we proceed with the simulation of a turboshaft engine
with a one-dimensional flow model. The purpose is to investigate the main
phenomenological effects on the different parts of the engine, and to estimate
which EAPS effects are stronger. The simulation is carried out with the gas
turbine simulation program GSP [49, 50].
The model includes a fuel control option, an inlet pressure loss (or ram
recovery) and an air bleed option at the compressor. The pressure loss is

78
set to a value estimated for the EAPS. The air bleed option is used to set
the fraction of inlet mass flow drawn off the engine core as a result of the
scavenge flow. This flow is assumed to be lost. The power turbine load has
an input value. The compressor has an externally controlled rotor speed
and the power turbine has a power balance at rotor speed option. The GSP
program calculates the fuel flow required to maintain the specified power
turbine load and rpm2 . A diagram of the turboshaft model is shown in
Fig. 20; TT5 denotes the temperature of the combustion gas at the exit of
the combustor (or entrance into the gas generator turbine). The analysis is
carried out in steady-state mode for variable control load. A few key engine
parameters are analysed, in particular the net turboshaft power and the
combustor temperature.
Figure 22 shows the predicted net shaft power for a turboshaft model
similar to the General Electric T700. The atmospheric conditions are stan-
dard day at sea level. The inlet pressure loss is relatively large, correspond-
ing to about 4kPa, as a result of dirty IBF. This pressure loss has been
coupled with a scavenge flow up to 9%. Thus, the inlet parameters on the
engine are extreme for this specific atmospheric condition. Figure 21 shows
the gas generator maximum temperature for the same case. The data clearly
indicate that for a given mass flow rate, there is a loss in net power that is
clearly dependent on the aero-thermodynamic conditions at the inlet. For
the same fuel flow, there is a corresponding increase in gas temperature
across the engine.

2
At the date of writing, the program is available from the Internet address:
www.gspteam.com.

79
 Shaft Control Load
Scavenge Flow
(Air Bleed) 2

Fuel Control Unit

5

1 3 4 TT5 7 8
9
6

EAPS Inlet Compressor Combustor HP Turbine LP Turbine Exhaust

Figure 20: Turboshaft model.

1500
SEA LEVEL, Standard Day

No EAPS
1200
Net Shaft Power, kW

900
ms = 9%
ms = 6%
ms = 3%
600

300

∆p / ∆po = 0.96

0.02 0.04 0.06 0.08 0.1 0.12

Fuel flow, kg/s

Figure 21: Predicted turboshaft power performance with and without EAPS.

80
 1000
SEA LEVEL, Standard Day

ms = 9%

Gas generator temperature, K

ms = 6%

900 ms = 3%

800
No EAPS

∆p / ∆po = 0.96
700
0.02 0.04 0.06 0.08 0.1 0.12
Fuel flow, kg/s

Figure 22: Predicted turboshaft gas generator temperature with and without EAPS.

7. Conclusions

This paper has presented some key multi-disciplinary aspects in particle

separation for helicopter turboshaft engines. Through a review of the current
technical literature and first-order fluid dynamics theories, it was possible to
establish the main characteristics of various types of filters. There remains
considerable lack of technical data on all types of systems; these data could
be used for research, design and operational analysis of rotorcraft in polluted
environments.
Particle separators cause a number of problems on the turboshaft perfor-
mance, in particular relatively high loss of inlet pressure (with the exception
of the VTS), and inlet mass flow (with the exception of IBF). However, with-
out exception, there is a degradation of filter capability over time, some of

81
which is attributable to clogging of separator parts, and the other is at-
tributable to damage and erosion caused by the particulate. Some research
has highlighted side-effects, such as sensitivity to icing.
A turboshaft simulation has indicated that the loss in engine performance
can be considerable, even in absence of impellers and other powered scavenge
means. Therefore, the use of particle separators should be limited to cases
when the risk of dust ingestion is high. For this reason, turboshaft engines
with embedded IPS come with some performance penalty by design. A
better option would be a system that can be temporarily disconnected to
allow for unfiltered air when the risk of engine damage is minimal.
Ultimately, the choice between separators relies on a systems analysis
that must include the effects on the engine, on the vehicle, on the perfor-
mance under critical atmospheric conditions, and on the ground logistics.
In fact, there are aspects as different as overall weight, installation time,
access to the engine and reliability.

References
[1] P. Stallard, Helicopter engine protection, Perfusion 12 (1997) 263–267.

[2] AGARD, Engine-airframe integration for rotorcraft, Tech. Rep. LS-148,

AGARD, Neuilly sur Seine (1986).

[3] J. Seddon, E. Goldsmith, Practical Intake Aerodynamics, 2nd Edition,

Blackwell, 1999.

[4] J. Calcaterra, Experimental and Numerical Simulation of FOD, RTO-

TR-AVT-094, NATO-RTO, 2004, Ch. 3.

[5] J. van der Walt, A. Nurick, Life prediction of helicopter engines fitted
with dust filters, J. Aircraft 32 (1) (1995) 118–123.

82
 [6] J. van der Walt, A. Nurick, Erosion of dust-filtered helicopter tur-
bine engines. Part I: Basic theoretical considerations, J. Aircraft 32 (1)
(1995) 106–111.

[7] J. van der Walt, A. Nurick, Erosion of dust-filtered helicopter turbine

engines. Part II: Erosion reductions, J. Aircraft 32 (1) (1995) 112–117.

[8] J. Warren, C. Gorton, S. Hoff, F. Alby, Effects of Sand and Dust on

Small Gas Turbine Engines, RTO-TR-AVT-094, NATO-RTO, 2004,
Ch. Annex B.

[9] A. Haider, O. Levenspiel, Drag coefficient and terminal velocity of

spherical and non spherical particles, Powder Technology 58 (1) (1989)
63–70.

[10] U. A. M. Department, Engineering Design Handbook. Helicopter En-

gineering. Part One. Preliminary Design, Pamphlet A-700200, Alexan-
dria, VA, 1974.

[11] W. Prinsloo, P. Villiers, M. Dijken, Vortex tube separating device, US

Patent 4,985,058 (January 1991).

[12] J. Campolong, Vortex separator, US Patent 3,895,930 (July 1975).

[13] M. DR, R. TE, Centrifugal separator, US Patent 4,008,069 (February

15, 1977).

[14] D. Monson, D. Keller, J. Rothman, Deflecting element for centrifugal

separators, US Patent 3,517,821 (June 30, 1970).

[15] H. Brockmann, Multi-stage cyclone separator, US Patent 4,289,611

(September 15, 1981).

83
[16] T. Simpson, Separator, US Patent 4,255,174 (March 10, 1981).

[17] C. Roach, Vortex air cleaner assembly with acoustic attenuator, US

Patent 4,050,913 (September 27 1977).

[18] H. Shohet, A. Amelio, J. Kowalonek, Engine inlet air particle separator,

US Patent 3,449,891 (June 17 1969).

[19] R. Duffy, B. Shattuck, Integral engine inlet particle separator. Volume

2. Design guide, Tech. rep., General Electric Co., Cincinnati, OH (Aug
1975).

[20] P. e. a. Pall, Vortex air cleaner assembly having uniform particle re-
moval, US Patent 3,520,114 (July 14, 1970).

[21] J. Ballard, Impact of IPS and IRS configurations on engine installation

design, in: Engine-Airframe Integration for Rotorcraft, Vol. AGARD
LS-148, Essex, UK, 1986.

[22] W. McAnally, M. Schilling, Investigation of feasibility of integral gas

turbine engine solid particle inlet separators. Phase I. Feasibility study
and design, Tech. Rep. USAAVLABS TR-70-44 (Aug 1970).

[23] W. McAnally, M. Schilling, Investigation of feasibility of integral gas

turbine engine solid particle inlet separators. Phase II. Feasibility
demonstration, Tech. Rep. USAAVLABS TR-71-13 (April 1971).

[24] J. Hobbs, Gas turbine engine air intake, US Patent 4,389,227 (June 21,
1983).

[25] M. Stoten, Integral inertial particle separator for radial inlet gas tur-
bine, US Patent 6,134,874 (October 24, 2000).

84
[26] S. Barr, Particle separator for turbine engines of aircraft, US Patent
3,993,463 (November 23, 1976).

[27] D. Richardson, W. Wilkins, Moisture and particle removing means for

engines, US Patent 3,148,043 (September 8, 1964).

[28] P. Snyder, Particle separator for a gas turbine engine, US Patent

6,698,180 (March 2, 2004).

[29] T. Hull, H. Nye, Particle separator with scroll scavenging means, US

Patent 3,832,086 (August 27, 1974).

[30] J. Jenkins, J. Sperinck, M. Buller, Inlet particle separator, US Patent

4,702,071 (October 27, 1987).

[31] W. McAnally, Particle and moisture separator for engine inlet, US

Patent 3,766,719 (October 23, 1973).

[32] B. Weinstein, W. Steyer, Foreign particle separator system, US Patent

4,265,646 (May 5, 1981).

[33] P. Hartman, High particle separation efficiency system, US Patent

6,702,873 B2 (March 9, 2004).

[34] T. Higgins, Adaptive inertial particle separator, EU Patent 1,908,939

(April 9, 2008).

[35] P. Hartman, C. Meadowcroft, Engine inlet air particle separator with

active flow control, U.S. Patent 7,296,395 B1 (November 20, 2007).

[36] A. Hamed, Y. Jun, J. Yeuan, Particle dynamics simulations in inlet

separator with an experimentally based bounce model, J. Propulsion &
Power 11 (2) (1995) 230–235.

85
[37] B. Vittal, D. Tipton, W. Bennett, Development of an advanced vaneless
inlet particle separator for helicopter engines, J. Propulsion 2 (5) (1986)
438–445.

[38] G. Musgrove, M. Barringer, K. Thole, E. Grover, J. Barker, Computa-

tional design of a louver particle separator for gas turbine engines, in:
Proceedings of ASME Turbo Expo 2009: Power for Land, Sea and Air,
Orlando, FL, 2009.

[39] F. Saeed, A. Al-Garni, Analysis method for inertial particle separator,

J. Aircraft 44 (4) (2007) 1150–1158.

[40] M. Taslim, A. Khanicheh, S. Spring, A numerical study of sand separa-

tion applicable to engine inlet particle separator, AHS Journal 54 (4).

[41] S. Suhr, Preliminary turboshaft engine design methodology for rotor-

craft applications, Master’s thesis, Georgia Institute of Technology (De-
cember 2006).

[42] M. Scimone, Aircraft engine air filter and method, US Patent 6,595,742
(July 22, 2003).

[43] J. Stelzer, T. Newman, Engine air filter and sealing system, US Patent
7,192,562 B2 (Mach 20, 2007).

[44] J. Bear, Dynamics of Fluids in Porous Media, Elsevier, 1972.

[45] Y. Chen, S. Hsiau, Cake formation and growth in cake filtration, Powder
Technology 192 (2009) 217–224.

[46] D. Purchas, K. Sutherland (Eds.), Handbook of Filter Media, 2nd Edi-

tion, Elsevier, 2002.

86
 [47] N. de Nevers, Fluid Mechanics for Chemical Engineers, 2nd Edition,
McGraw-Hill, 1991, chapter 12.

[48] T. Dickenson (Ed.), Filters and Filtration Handbook, Elsevier, 1997.

[49] W. Visser, M. Broomhead, GSP: A generic object-oriented gas turbine

simulation environment, in: ASME Gas Turbine Conference, ASME
2000-GT-0002, Munich, Germany, 2000.

[50] W. Visser, O. Kogenhop, M. Oostveen, A generic approach for gas

turbine adaptive modeling, ASME J. Enginering for Gas Turbine &
Power 128 (1) (2006) 13–19.

87

View publication stats