Technical Notes

Design rules for small low speed wind tunnels

R. D. MEHTA and P. BRADSHAW

1. INTRODUCTION lines for the main components of a wind tunnel—the fan,

Even with today's computers, a wind tunnel is an essential wide-angle diffuser, corner vanes, settling chamber, con-
tool in engineering, both for model tests and basic re- traction and exit diffuser (Fig. 1)—based on data from
search. Since the 1930s, when the strong effect of free- successful designs and some original experiments. For
stream turbulence on shear layers became apparent, details of the data correlations see Mehta (1977) and for
emphasis has been laid on wind tunnels with low levels complete details of the experiments and design procedure
of turbulence and unsteadiness. Consequently most high see Mehta (1978).
performance wind tunnels were designed as closed-circuit
types (Fig. 1(a)) to ensure a controlled return flow. How- 2. FANS
ever, as will be seen below, it is possible with care to 2.1. Axial flow fans
achieve high performance from an open-circuit tunnel, The usual arrangement in a closed-circuit tunnel is a
thus saving space and construction cost. 'Blower' tunnels stator ('pre-rotation vanes') upstream of the rotor (the
(with the fan at entry to the tunnel, Fig. 1(b)) facilitate fan proper), designed so that the swirl at exit is zero.
large changes in working section arrangements; to cope In the case of an open-circuit tunnel, swirl present in the
with the resulting large changes in operating conditions, flow out of the fan may be dissipated before the flow
a centrifugal fan is preferable to an axial one. For ease reaches the intake, but a remaining advantage of pre-
of changing working sections the exit diffuser is often
omitted from small blower tunnels, at the cost of a power rotation vanes is that the flow velocity relative to the fan
factor greater than unity. This paper concentrates on the blades is larger than if the stator is absent or located
design of small blower tunnels but most of the infor- downstream of the fan.
mation is applicable to wind tunnels in general.
2.1.1. Fan solidity
A large open-circuit tunnel would be of rather incon- The design procedure outlined by Bradshaw and Pank-
venient dimensions, mainly in length. Also, an open- hurst (1964) is still an adequate guide. The only serious
circuit tunnel requires enough free room around it so problem found in fan design that is not found in the
that the quality of the return flow is not affected signifi- design of wings for low-speed aircraft is the interference
cantly (remember that an open-circuit tunnel in a room is between the flow fields of the blades. This interference
really a closed-circuit tunnel with a poorly-designed re- depends mainly on the 'solidity', the ratio of blade chord
turn leg). The choice may also be restricted by the maxi- to the gap between blades (measured around the circum-
mum available blower size. A working section Re per ference). Providing that the solidity is less than unity
metre of more than about 3 x 106 (a speed of about approximately, interference is small enough to be treated
40 ms_1) is rare in blower tunnels of whatever size, and as a small correction to the performance of an isolated
commercial blowers capable of producing such a speed in aerofoil; for higher solidities the flow cannot be accur-
a section more than about 1 m2 in area are also rare. ately related to that round an isolated aerofoil, and data
The main advantage of open-circuit tunnels is in the for 'cascades' (rows of aerofoils arranged in the same
saving of space and cost. They also suffer less from manner as corner vanes) must be used instead. The
temperature changes (mainly because room volume ^> solidity varies with radius, and in order to use the same
tunnel volume) and the performance of a fan fitted at
the upstream end is not affected by disturbed flow from
the working section. One disadvantage of any open-
circuit tunnel with an exit diffuser is that the pressure is
always less than atmospheric and so spurious jets issue
from holes left unpatched, although this can be remedied
by obstructing the tunnel outlet and creating an over-
pressure in the working section. The main advantage of
a centrifugal blower, as distinct from an axial fan, is that
it performs well over a large range of loads (the whole
blade being at the same incidence and hence operating
at the same lift coefficient). The only advantage of a
suction tunnel, with a centrifugal or axial fan at exit, is
the dubious one that air coming from the tunnel room
may be less disturbed than that coming from a fan.
It is difficult and unwise to lay down firm design rules The Authors: Dr. ft. D. Mehta and Professor P. Bradshaw,
mainly because of the wide variety of requirements and BA, Department of Aeronautics, Imperial College of Science
especially the wide variety of working-section configura- and Technology, London.
tions. An attempt is made here to present design guide- Paper No 7IS.

Aeronautical Journal November 1979 Mehta and Bradshaw

Downloaded from http:/www.cambridge.org/core. Monash University, on 27 Dec 2016 at 11:00:24, subject to the Cambridge Core terms of use, available at http:/www.cambridge.org/core/terms.
http://dx.doi.org/10.1017/S0001924000031985
 - 4 0 b.h p MOTOR WIDE ANGLE DIFFUSES -

U_L

Figure 1(a). T h e m a i n components of a t y p i c a l closed-

circuit wind tunnel.

uniformity of the flow (which is also improved by the

screens in the wide-angle diffuser and the settling
chamber).
Centrifugal • Wide-angle i I L,. c ,

2.2.J. Advantages over other fans

Centrifugal blowers run with reasonable steadiness and
efficiency over a wide range of flow conditions (i.e. vary-
ing tunnel power factor) because the whole blade span
operates at nominally the same lift coefficient. The noise
and pulsations generated by a centrifugal blower are
adequately low, even at off-design conditions, and the
Figure 1(b). The main components of a t y p i c a l blower uniformity of flow varies less with advance ratio, E//cor2
t u n n e l . (Not to scale.) in the notation of Fig. 2. The swirl (exit vortex) produced
by a single-inlet blower is also independent of advance
design procedure for the whole length of the blade it is ratio (dependent on the ratio of rotor to casing width).
desirable to keep the solidity at the root below unity by
mounting the fan on a central nacelle whose maximum
2.2.2. Types of impeller
diameter is roughly half the fan diameter.
The most common type of blading is the backward-facing
2.1.2. Blade design
aerofoil-type (Fig. 2); forward-facing is less efficient. If
the blower efficiency is not too important, these blades
Axial fan efficiencies are of the order of 90% so that
minimisation of losses is not usually important, and the could be designed in the same way as corner vanes or
usual procedure is to choose the blade lift coefficient to cascades by choosing a leading edge angle of 4-5° and a
be as high as is safe, irrespective of lift /drag ratio; values zero trailing edge angle, but a more efficient blade shape
of 0-7 to 0-9 are typical. is that of a cambered aerofoil with finite thickness. In
the present authors' tests on blowers with aerofoil-type
2.1.3. Pre-rotation vanes impellers it was found that the flow uniformity deterior-
Pre-rotation vanes should be run at a lift coefficient not ated with increasing loading. However, with backward-
too far above that for maximum lift/drag ratio because facing 'S' shaped blades (Fig. 2) the flow uniformity was
their wakes pass through the fan; to limit the resulting found to improve with loading, presumably because these
noise, the axial distance between the trailing edge of the blades stall relatively early, leading to increased mixing.
pre-rotation vanes and the leading edge of the fan blades The cost is a higher turbulence level in the outlet flow
should be at least 20% of the vane chord and the number and a reduced blower efficiency.
of fan blades should be different from the number of
vanes. Pre-rotation vane solidities usually fall into the 2.2.3. Splitter plate (tongue)
cascade range. This is an important component which affects the outlet
An alternative to pre-rotation vanes for a lightly flow uniformity and blower noise characteristics. For
loaded fan is a set of straightener vanes downstream of minimum interference with the flow uniformity, the ratio
the fan. of tongue height to casing height needs to be small
For detailed design rules for pre-rotation vanes, fan (<0-3) and the angle and shape carefully designed. The
blades and straighteners see Bradshaw and Pankhurst gap between the rotor and tongue needs to be a minimum
0964). for aerodynamic reasons but optimised for minimum in-
teraction with the outgoing flow and thus minimum
2.2. Centrifugal blowers noise level. The tongue design on most commercially
Centrifugal blowers are normally used to drive open- available blowers is adequate. A badly angled tongue
circuit tunnels from the upstream end: a blower could could be improved upon by adding a cusped fairing
be installed at the exit instead but this has no particular downstream, as shown dotted in Fig. 2.
advantage. Single-inlet blowers could also be used to
drive return-circuit tunnels by installing them in one of 2.2.4. Other features and suggestions
the corners. Single-inlet blowers are found to produce a An inlet bellmouth helps to produce a uniform flow and
vortex-type flow (due to the asymmetric positioning of reduces inlet losses, and an inlet filter (helping to reduce
the impeller) which would aid wall flow attachment in the inlet swirl) is essential to reduce contamination of hot-
wide-angle diffuser. This compensates for the non- wire probes. Large blowers should be mounted on anti-

444 Mehta and Bradshaw Aeronautical Journal November 1979

Downloaded from http:/www.cambridge.org/core. Monash University, on 27 Dec 2016 at 11:00:24, subject to the Cambridge Core terms of use, available at http:/www.cambridge.org/core/terms.
http://dx.doi.org/10.1017/S0001924000031985
 vibration mountings and connected to tne tunnel with a jets and presumably amalgamating to form longitudinal
flexible coupling to reduce vibration, vortices which persist through the contraction. The coali-
Double-inlet blowers (air entering the impeller from tion process is enhanced by variations in B (i.e. non-
both sides) tend to produce a uniformly inclined flow uniform weave) and by irregularities in the screen shape
(without a vortex) which takes a longer distance to re- (i.e. wrinkles). It is therefore essential to inspect and
attach to the bottom wall downstream of the tongue. clean wind tunnel screens regularly.
One should therefore be more conservative in designing
wide-angle diffusers for double-inlet blowers. 3.3. Determination of K
On the whole, commercially available single-inlet (ratio of pressure drop to dynamic pressure)
centrifugal blowers with backward-facing impellers are Although there is no wholly satisfactory method, Wieg-
adequate for driving blower tunnels. hardt's (1953) formula {K = 6-5 [1 -B/B2] [Ud/Bp]-1'3},
Once the maximum required fan static pressure and where d is wire diameter, predicts the right trend; K
volume flow rate have been estimated, the makers' per- decreases with increasing speed up to about Ud/Bv =
formance charts can be consulted. Optimisation between 600, after which it is independent of Re. Collar's (1939)
the efficiency, rpm and required power leads to the formula {K = 0-9 (1 — BlB2)} usually over-estimates K in
blower choice (see section 10). the high Re limit. One needs to be more careful in pre-
dicting X-values for plastic screens since,
3. SCREENS K = f(B, Re, 9 • . • co-planarity . . .).
Wind tunnel screens are normally made of metal wires
interwoven to form square or rectangular meshes. where 8 is angle of screen to incident flow. For 8^0 use
Screens woven from nylon or polyester threads are also
K6 = K cos™ 8, with m = 1 • 0 for screens with B ~ 0 • 6
now being used when the wind loads are not expected to
and m ~ 1-4 for B ~ 0-3.
be very high (UTS of nylon ~ 70, steel ~ 1100, bronze
~700-1100 MNm-a and E of nylon > 3, steel~200,
3.4. Determination of a
bronze ~ 100 GNm - 2 ). The action of the gauze is
(ratio of outlet angle to inlet angle)
described in terms of two parameters: the pressure drop
coefficient, K = f1(B,Rc,8) and the deflection coefficient, For a the f o r m :
« = / 2 (B, K, 6), where B is the screen open-area ratio and B
a=A+
8 is the flow incidence angle, measured from the normal vTiTK)
to the screen.
where A, B are empirical constants, is a better fit than
the generally accepted form:
3.1. Main effects
(for detailed explanations see Mehta 1978) 11
Screens make the flow velocity profiles more uniform by a = V U + -K)
imposing a static pressure drop proportional to (speed)2 Note that the refractive index of a screen (/x) defined as
and thus reduce the boundary layer thickness so that the in optics is equal to 1/a for small 8- For larger 8 use
ability to withstand a given pressure gradient is increased.
A screen with a pressure drop coefficient of about 2
removes nearly all variation in the longitudinal mean D
velocity. A screen also refracts the incident flow towards <x«- i - tan- 1 {tan 8- f-sec'fl C - (E + F0))
0 2
the local normal and reduces the turbulence intensity in
the whole flow-field. For a given open-area ratio, it is C, D, E and F are empirical constants.
better to have a smaller mesh for the reduction of
pre-existing turbulence. Plastic screens tend to yield a Values suggested for the empirical constants by some
more uniform flow beyond the boundary layer edge, limited experiments (Mehta, 1978) are: A =0-66, B =
mainly due to the weaving properties, and produce an 0-31, C = 0-68, £> = 0-62, £ = 1 - 0 , F = l - 5 .
'overshoot' in the velocity profile near the edge, mainly A more complete analysis of the flow through screens
caused by screen deflection angle which is a maximum can be found in Mehta (1978).
at the wall. In terms of tackling a given pressure gradient
or avoiding separation, this overshoot could be beneficial. 4. DIFFUSERS
The flow through a diffuser depends on its geometry
defined by the area ratio (A), diffuser angle (28), wall
3.2. Open-area ratio (/3) contour and diffuser cross-sectional shapes. Other para-
Metal screens with very low B ( ~ 0 • 3) also produce an meters like the initial conditions, boundary layer control
overshoot but this is caused by streamline inclination near method and the presence of separation could also affect
the boundary layer edge. Low B ( < 0 • 57) screens also the flow thus making it very difficult to predict. Almost
produce instabilities resulting from a random coalition of all knowledge acquired about diffusers is empirical. There
are two main types:

4.1. Exit diffusers

"These are fitted downstream of the working sections and
have gentle expansions with a diffuser included angle
usually not exceeding 5° (for best flow steadiness,
although best pressure recovery is achieved at about 10°)
and an area ratio not exceeding about 2-5. It is import-
ant to have a reasonable degree of flow steadiness in the
a.Radial-vane blower e. Backward aerofoil blower c Backward 5-type blower exit diffuser, since otherwise the pressure recovery tends
to fluctuate with time, and, therefore, so does the tunnel
Figure 2. Different impeller types used in centrifugal
blowers. speed if the input power is nearly constant. The design

Aeronautical Journal November 1979 Mehta and Bradshaw 445

Downloaded from http:/www.cambridge.org/core. Monash University, on 27 Dec 2016 at 11:00:24, subject to the Cambridge Core terms of use, available at http:/www.cambridge.org/core/terms.
http://dx.doi.org/10.1017/S0001924000031985
 of these diffusers is well catered for by existing methods
i^see Cockrell and Markland, 1974).
4.2. Wide-angle d iff user
This type is normally installed between the blower and
settling chamber or between the fourth corner and settling
chamber of a closed-circuit tunnel; the cross-sectional
area increases so rapidly that separation can be avoided
only by boundary layer control. A wide-angle diffuser is
a means of reducing the length for a given area ratio
rather than effecting a pressure recovery; generally the
net pressure rise through a screened wide-angle diffuser
is negative but small.
4.2.?. Boundary foyer control methods
The most popular means of boundary layer control is
by installing gauze screens. A screen, besides removing
the direct effects of layer growth and incipient separa-
tion, gives the layer a new lease of life. In a wide-angle
diffuser it is better to use several screens of relatively
small K (less than about 1-5) because increasing K at
one station has little effect on the skin friction at a
station much further downstream. Other types of
boundary layer control methods include splitters, suction
slots, trapped vortices, vortex generators and vanes and
may be preferable in diffusers with very severe geometries
(A > 5, 20 > 50°). For a review see Mehta (1977).
4.2.2. Design charts 11 12 13 14 15
A -
The four most important parameters in a wide-angle
diffuser are A, 20, K and n, where n is the number of Figure 4. Overall pressure d r o p coefficient requirements
for a diffuser w i t h screens.
screens within the diffuser—this includes the screens in-
stalled at the inlet and outlet. Data were collected from
over a hundred wide-angle diffuser designs, mostly must be less than (1-14 Kmm +1-0), giving the minimum
'successful' (no separation, and uniform outlet flow with
an acceptable turbulence level), and charts were plotted required overall pressure drop coefficient. A diffuser con-
for relevant parameters, from which design rules have figuration satisfying both these curves should perform
been derived. In Fig. 3, A is plotted against 20; the successfully provided that certain other design factors are
curves enclosing successful configurations have an ap- kept in mind:
proximately hyperbolic shape. As n increases, the vertex
moves to a higher value of 20 and, to a lesser extent, to (i) Inlet conditions: Thin boundary layers and steady
a higher value of A, thus implying a stronger dependence flow at the inlet are obviously beneficial.
of required n on 20. Figure 4 is a plot of the sum of (if) Screen Positioning: The basic rule is to place screens
pressure drop coefficients of all the screens, JCs„m=2 where the diffuser wall angle changes suddenly, since
(Ap/q), vs A. The straight line EF (A = 1 • 14 K^ + 1 • 0) these are the points where the flow is most likely to
included the maximum number of successful configura- separate. In diffusers where no obvious location is
tions. indicated screens should be equally spaced, remem-
bering that a screen at the diffuser entry (with a
relatively high resistance) is desirable because the
4.2.3. Overall design procedure
angle changes suddenly there.
For a diffuser design to operate successfully it must lie to (Hi) Wall shape: The number of screens required in a
the left of the relevant curve in Fig. 3, giving the mini- diffuser could well be reduced, and the efficiency
mum number of screens required in the diffuser, and A increased, by employing curved walls. Potential flow
methods are sometimes used to determine wall
shapes but it is often easier to design wall shapes by
eye. Straight-walled diffusers (often with curved
screens) are, however, often employed, because they
are easier and cheaper to construct.
(iv) Screen shape: It is an advantage for the screen to
intersect the diverging walls and streamlines at right
angles, so that the refraction of the flow by the
screens does not itself induce separation. Curved
screens can be held in metal frames pressed into
circular arcs and lined with wooden strips so that
the gauze may be firmly embedded between two
frames. It could be more difficult to dish the more
flexible plastic screens (see section 3) which may also
tend to flutter. Another alternative is to use a plane,
10° 20° 30° 40° 50° 60° __ 'variable-iT screen comprising of one screen con-
Figure 3. Design boundaries for diffusers w i t h screens.
centrically superimposed on another.

Mehta and Bradshaw Aeronautical Journal November 1979

446
Downloaded from http:/www.cambridge.org/core. Monash University, on 27 Dec 2016 at 11:00:24, subject to the Cambridge Core terms of use, available at http:/www.cambridge.org/core/terms.
http://dx.doi.org/10.1017/S0001924000031985
 (v) Cross-sectional shape: Most wide-angle diffusers have 1976). It turns out that the shear layer instability in the
either rectangular or square cross-sections for ease near wake has a strength proportional to the shear layer
of construction and since pressure recovery is not too thickness and so for the longest honeycomb, the ratio of
important. It is advisable to fillet the corners in turbulence generated to that suppressed is greatest.
small tunnels, whose designs are likely to be more
adventurous, to reduce the risk of large regions of 6.2. Optimum cell size
flow separation.
For maximum overall benefit the cell length should be
4.2.4. Comparison and verification of design rules
about 6-8 times its diameter. The cell size should be
smaller than the smallest lateral wavelength of the velocity
These design rules compare well with those proposed by variation (roughly 150 cells per settling chamber
Kline et al (1957), Schubauer and Spangenberg (1948) and 'diameter', i.e. 25 000 total, are adequate). The cross-
Squire and Hogg (1944). This is to be expected because sectional shape of the honeycomb cells is usually
many designers have used these rules; evidently the rules hexagonal, but sometimes square or triangular, the shape
are successful, but they may be conservative. The present being chosen mainly for ease in construction. Impreg-
rules also compare well with some experiments and test nated paper honeycombs are adequate for small tunnels.
cases, details of which can be found in Mehta (1977), Aluminium honeycombs made for aircraft sandwich con-
although there is evidence that the rules are not inflexible. struction have more precise dimensions than paper honey-
combs and are to be preferred for high performance
5. CORNER VANES
tunnels and large tunnels where the wind loads may be
Even some open-circuit tunnels have corners, say to expected to be high. The cells of all honeycombs are
deflect the efflux from a horizontal tunnel upwards to often partly obstructed by burrs which can be fatigued
reduce draughts. Rules for the design of vanes for 90° off with an air hose.
corners are uncontroversial and probably rather con-
servative. Thin sheet metal vanes are used on all but the
largest tunnels and, even in the latter, thick aerofoil- 7. SETTLING CHAMBERS
section vanes are used for strength rather than aero- 7.1. General arrangement
dynamic advantage. The ratio of the gap, h, between
vanes (measured from leading edge to leading edge) to The usual arangement consists of a honeycomb (with
the chord, c, should not exceed about 0-25; the vane lift about 25 000 cells) followed by screens, the number and
coefficient is 2h/c. Usual practice is to make the vane as .K-value depending on the turbulence level requirements.
a circular arc, with short straight extensions at leading and If severe yaw or swirl is expected in the flow from the
trailing edges for ease of rolling or pressing. The trailing wide-angle diffuser, it is advisable to install one screen
edge is aligned parallel with the axis of the downstream upstream of the honeycomb, so that the flow angles are
leg and the leading edge is set at a positive 'angle of reduced. A screen with K=l-5 reduces yaw and swirl
incidence' of 4° to the axis of the upstream leg. This angles by a factor of about 0-7 for swirl angles of about
arrangement has superficial logic but differs from estab- 40°. The honeycomb should be installed some way
lished cascade-design practice, and recently Ermshaus and downstream of the wide-angle diffuser exit, so that the
Naudascher (1977) have successfully used a hodograph- flow static pressures and angles have had a chance to
solution design which has a negative angle of incidence become more uniform. Since screens with small /? (less
at the leading edge and over-turns at the trailing edge so than about 0 • 57) tend to produce instabilities, presumably
that the included angle is 105° instead of the conven- in the form of longitudinal vortices, at least one screen
tional 86°. It is not clear whether a significantly higher with a larger ft (>0-57) should be used (in the most
hjc can be used with this design. downstream position) if a truly two-dimensional boundary
layer is required in the working section. Another alterna-
The pressure drop through thin vanes of standard tive is to place the honeycomb downstream of the screens
(86°) design is estimated by Bradshaw and Pankhurst but this at best results in a rise in the turbulence level and
(1964) to be about 1-2 (Uc/v)-1'4, times the dynamic is not recommended in general.
pressure.
6. HONEYCOMBS 7.2. Spacing between screens
Honeycombs are effective for removing swirl and lateral There are two important properties to consider:
mean velocity variations, as long as the flow yaw angles
are not greater than about 10°. Large yaw angles cause (0 For the pressure drops through the screens to be
the honeycomb cells to 'stall' which reduces their effective- completely independent, the spacing should be such
ness besides increasing the pressure loss. that the static pressure has fully recovered from the
perturbation before reaching the next screen
6.1. Reduction of turbulence (i.e. dp/dy = 0).
An incidental effect of honeycombs is to reduce the (ii) For full benefits from the turbulence-reduction point
turbulence level in the flow. Essentially, the lateral of view, the minimum spacing should be of the order
components of turbulence, like those of mean velocity, of the large energy containing eddies.
are inhibited by the honeycomb cells and almost complete
annihilation is achieved in a length equivalent to about It has been found that a screen combination with a
5-10 cell diameters. Honeycombs themselves shed small spacing equivalent to about 0-2 settling chamber dia-
scale turbulence, the level of which is found to be higher meters performs successfully. The optimum distance
when the cell flow is laminar than when it is turbulent: between the last screen and the contraction entry has also
this is attributed to a basic instability of the laminar near been found to be about 0-2 cross-section diameters. If
wakes. Note that the cell flow in most wind tunnel this distance is much shorter significant distortion of
honeycombs is laminar and so Lumley and McMahon's the flow through the last screen may be expected. On the
(1967) analysis, which assumes turbulent flow, will not other hand, if this distance, or for that matter the overall
apply. With a laminar cell flow the net reduction is length of the settling chamber, is too long then unneces-
greatest for the shortest honeycomb (Loehrke and Nagib, sary boundary layer growth occurs.

Aeronautical Journal November 1979 Mehta and Bradshaw 447

Downloaded from http:/www.cambridge.org/core. Monash University, on 27 Dec 2016 at 11:00:24, subject to the Cambridge Core terms of use, available at http:/www.cambridge.org/core/terms.
http://dx.doi.org/10.1017/S0001924000031985
 7.3. Installation of components corner fillets, but rapid termination of these in the
Screens are normally tacked onto wooden frames. More working section must be avoided.
care is necessary when tacking on plastic screens since Two-dimensional contractions are sometimes preferred
these, being more flexible, tend to wrinkle along the lines on tunnels used for boundary layer studies, where the
of tension. The honeycomb is usually just push-fitted into working section is wide but shallow. However, if the
its own frame. A useful arrangement for small tunnels is boundary layers are thick, the plane walls tend to develop
to rest the wide-angle diffuser, screen frames and con- strong secondary flows. Also, 2-D contractions require
traction on a table and clamp them by drawbolts, so that about 25% more length to attain the same uniformity of
frames can be withdrawn easily. On larger tunnels, it is pressure distribution as axisymmetric ones.
advisable to equip the settling chamber (and wide-angle
diffuser) components with castors for ease of removal. 8.4. W a l l shape
Even in tunnels made of metal or concrete, the screens 8.4.1. Theoretical design
are normally installed in separate frames which can be The solution of the Laplace equation or the Stokes-
withdrawn from the tunnel for cleaning or replacement. Beltrami equation is relatively easy for simple geometries
and many analytical solutions have been derived. With
8. CONTRACTIONS the onset of computers many numerical schemes have also
A contraction: been proposed. For a review see Mehta (1978).
There is no wholly satisfactory method of theoretical
(0 Increases the mean velocity which allows the honey- wall shape design, as distinct from analysis. The applica-
comb and screens to be placed in a low speed region, tion of all these methods requires the establishment of
thus reducing pressure losses. some criteria and then the application of trial-and-error
(ii) Reduces both mean and fluctuating velocity variations techniques for which limited guidance is given.
to a smaller fraction of the average velocity.
8.4.2. Design by eye
The most important single parameter in determining Designers have often used the rather unscientific method
these effects is c, the contraction ratio. The factors of of design by eye. Note that the actual form of a con-
reduction, as derived by Batchelor (1970) for c ^> 1, traction contour is not very important except near the
are: ends, and that smoothness in contour shape is much more
important than exact dimensions. In general the wall
[/-component mean velocity: 1 /c radii of curvature should be less at the narrow end and
V or IP-component mean velocity: \/ c each end must join the parallel sections so smoothly
M-component rms intensity: 1 /2c {3 (ln4ci — l)} 1 ' 2 that at least the first and second derivatives of the
V or w component rms intensity: • (3c) 1/2 /2. curve are zero (or very small) at the ends.

(The factor of reduction of percentage velocity variation 9. W O R K I N G SECTIONS

is given by multiplying the above expressions by 100/c). Working section design is totally dependent on the re-
The design of a contraction centres on the production quirements of the individual experimenter. Blower tunnels
of a uniform and steady stream at its outlet, and requires are more flexible in accepting a variety of working
the avoidance of flow separation. Two more desirable sections (with and without exit diffusers).
criteria include minimum exit boundary layer thickness The flow out of a contraction often takes a distance
and minimum contraction length. A design satisfying all equivalent to about 0-5 diameters before the non-
criteria will be such that separation is just avoided and uniformities are reduced below an acceptable level. Also,
the exit non-uniformity is equal to the maximum tolerable if a turbulence grid is installed, it may take up to 10-15
level for a given application (typically ±i% velocity mesh lengths before a homogeneous flow is obtained.
variation outside the boundary layers). These requirements often fix the minimum length of the
working section. The streamwise pressure gradient is best
8.1. Contraction lengths controlled by installing tapered fillets.
It is always possible to avoid separation in the contraction It is advisable to mount removable side panels on
by making it very long, but this results in an increase of pinned hinges on large working sections which makes
tunnel length, cost and exit boundary layer thickness. their 'single-handed' removal easier and safer.
Drag forces, being proportional to (velocity)2, change
8.2. Contraction ratio by twice as large a fraction as the mean velocity; lift
Since the power factor contribution of screens in the forces change because of the change in mean velocity and
settling chamber varies as 1/c2, large contraction ratios because of the influence of tunnel walls on the effective
are advantageous. But large contraction ratios mean angle of incidence. Lift interference on a complete air-
higher construction and running costs besides possible craft model in a rectangular-section tunnel is minimised
problems of noise and separation near the ends. Therefore, if the ratio of working section breadth to height is s/2
contraction ratios between about 6 and 9 are normally (with model span less than three-quarters of the breadth)
used, at least for the smaller tunnels. so most general purpose aerospace tunnels are made with
approximately this aspect ratio. However tunnels for
3.3. Cross-sectional shape measurements in boundary layers, growing on the tunnel
In any contraction with a non-circular cross-section, the floor say, have an optimum breadth-to-height ratio of
flow near the walls tends to migrate laterally, especially about five since all that is necessary is that a reasonable
near corners of a polygonal section. In any case the thickness of irrotational flow shall remain between the
boundary layers near the corners will be more liable to roof and floor boundary layers at the end of the test
separate. However, recent investigations (Mehta, 1978) section (a diffuser with such a non-uniform entry flow
show that this does not cause a problem in a well-designed would not be very efficient). Tunnels for testing building
square contraction; the effect of boundary layer migration complexes or natural terrain at model scale can also have
in a contraction whose cross-sectional aspect ratio changes a large breadth/height ratio; conversely, tunnels for test-
along its length can be reduced by adding small 45° ing isolated tower buildings or smokestacks can have a

448 Mehta and Bradshaw Aeronautical Journal November 1979

Downloaded from http:/www.cambridge.org/core. Monash University, on 27 Dec 2016 at 11:00:24, subject to the Cambridge Core terms of use, available at http:/www.cambridge.org/core/terms.
http://dx.doi.org/10.1017/S0001924000031985
 breadth-to-height ratio less than unity, although the ratio (iv) Loss of total head at the exit of an open-circuit
of model breadth to tunnel breadth must still be kept tunnel.
small to minimise interference. In an open-circuit tunnel, the amount of kinetic
energy lost at the exit and dissipated into heat adds
10. ESTIMATION OF TUNNEL POWER FACTOR to the total losses.
Having decided the size and configuration of a wind
tunnel, the next design step is to estimate the tunnel
power factor, A (equal to HHp0U03A0, where H is the r)Aki =
shaft input power and subscript o refers to working sec-
tion conditions) so that the fan and drive unit can be [ = 1 for blower tunnels with no exit diffuser],
selected. It is difficult, but in fact not essential, to estimate typical value: r)Aki = 0-16 for A =2-5.
the power factor very accurately; adequate extra power Therefore the estimated overall tunnel power factor,
must be installed to cope with a variety of model or
working section configurations, not known in advance. A. = 2 AAn/7}~0-825 for the tunnel considered
The pressure losses in a wind tunnel are due to diffuser (with an exit diffuser), taking 77 = 0 • 8.
losses, resistive components such as screens, and friction
on the tunnel walls. The total pressure loss due to each Once the tunnel power factor has been estimated and
component can be estimated separately, and then summed the required fan static pressure rise determined, one can
and divided by the blower efficiency rj, typically 0-8, to set about the selection of the optimum fan size. The
give the tunnel power factor. Typical values for a tunnel dynamic pressure rise through a blower is usually ignored
similar to that shown in Fig. 1(b) are given below. and can be thought of as a safety factor in the calculations.
The fan outlet flow will be least turbulent when the fan
(/) Losses due to skin friction. is operating near maximum efficiency. Fan efficiency is a
function of the dimensionless flow rate; the pressure rise

"»"&7=(T)7c<:!te coefficient is a (weak) function of the dimensionless flow

rate also, so that requiring maximum efficiency
specifies both dimensionless flow rate and pressure rise
where 5 is the duct local perimeter and remember- coefficient. So for a given required flow rate and pres-
ing that the area ratio is the reciprocal of the velocity sure rise, two equations are obtained which can be solved
ratio. It is normally only necessary to estimate skin to give the fan size and optimum operating rpm. In
friction losses in the working section (A/A0=l). practice the manufacturer's performance charts should
Those in the diffusers are normally accounted for be searched for a fan size (and rpm) giving near maxi-
in the efficiency and the other components do not mum efficiency for the required flow rate and pressure
contribute significantly. rise.
Therefore, TJAXI^: CtSL/A, where L is the work- ACKNOWLEDGMENTS
ing section length. Typical value: T J A X 1 ~ ' 0 - 0 7 , We are grateful to R. W. F. Gould, National Maritime
assuming Cf ~ 0 • 003.
Institute, M. C. Welsh, CSIRO Melbourne, and a referee,
for helpful comments on a draft of this paper.
(//) Losses due to screens, honeycomb and corner-vanes.
REFERENCES
BATCHELOR, G. K. The theory of homogeneous turbulence.
TJAX 2 = X
'A. Cambridge University Press, pp 55-75, 1970.
A BRADSHAW, P. and PANKHURST, R. C. The design of low-speed
wind tunnels. Prog Aerospace Sci, Vol 5, p 1, 1964.
COCKRELL, D. J. and MARKLAND, E. Diffuser behaviour—a
review of past experimental work—relevant today. Aircr
So for a tunnel with four screens (two in the wide- Engng, Vol 46, p 16 (reprint of Vol 35, p 286), 1974.
angle diffuser with A/A0 = 4 and 6 respectively) COLLAR, A. R. The effect of a gauze on the velocity distri-
each with K=l-5 (for £7 = 5-10 m/s), and a honey- bution in a uniform duct. ARC R & M 1867, 1939.
comb with K = 0-5 we have, taking c = 9, typical ERMSHAUS, R. and NAUDASCHER, E. Der Niedergeschwindig-
skeits-windkanal des Instituts fiir Hydromechanik an der
value: T/AX2 = 0 - 1 8 (the screen at A/A0 = 4 contri- Universitat Karlsruhe. Zeitschrift fiir Flugwissenschaften und
butes 0-094)" Weltraumforschung, Vol 1, p 419, 1977.
KLINE, S. J., MOORE, C. A. and COCHRANE, D. L. Wide-angle
(Hi) Loss of total head in the exit diffuser. diffusers of high performance and diffuser flow mech-
anisms. / Aero Sci, Vol 24, p 469, 1957.
LOEHRKE, R. I. and NAGIB, H. M. Control of free-stream tur-
bulence by means of honeycombs: A balance between sup-
7JAX 3 = ( 1 - T 7 D ) pression and generation. ASME J Fluids Engg, Vol 981, p 342,
1976.
LUMLEY, J. L. and MCMAHON, J. F. Reducing water tunnel
turbulence by means of a honeycomb. ASME J Basic Engg,
Vol 89D, p 764, 1967.
where r)D is the diffuser efficiency. This is a loss due MEHTA, R. D. The aerodynamic design of blower tunnels
to the inefficiency of the diffuser in transforming with wide-angle diffusers. Prog Aerospace Sci, Vol 18, p 59,
kinetic energy into 'pressure energy' and is caused 1977.
by boundary layer growth and non-uniformity of MEHTA, R. D. Aspects of the design and performance of
blower tunnel components. PhD Thesis, Imperial College,
the flow. The efficiency of a wide-angle diffuser University of London, 1978.
with screens is generally negative but Ap is small. SCHUBAUER, G. B. and SPANGENBERG, W. G. Effect of screens
For a conical diffuser with A ~ 2 - 5 and 20<~5°, in wide-angle diffusers. NACA TN-1610. 1948.
Cockrell and Markland (1974) suggest TJD = 0 - 8 , but SQUIRE, H. B. and HOGG, H. Diffuser-resistance combinations
in relation to wind tunnel design. RAE Report No Aero 1933,
this may be lower for diffusers with rectangular 1944.
cross-sections, typical value: T/AA3 = 0 - 2 5 for WIEGHARDT, K. E. G. On the resistance of screens. Aero
7jD = 0 - 7 a n d A=2-5. Quarterly, Vol 4, p 186, 1953.

Aeronautical Journal November 1979 Mehta and Bradshaw 449

Downloaded from http:/www.cambridge.org/core. Monash University, on 27 Dec 2016 at 11:00:24, subject to the Cambridge Core terms of use, available at http:/www.cambridge.org/core/terms.
http://dx.doi.org/10.1017/S0001924000031985
 Simple finite elements for dynamic analysis of thick
pre-twisted blades

Dr. B. A. H. ABBAS

1. INTRODUCTION /? rotary inertia parameter =IX.X,/AL2

The study of the effects of shear deformation and rotary T kinetic energy
inertia on the dynamic behaviour of turbine blades has U strain energy
received some attention following the realisation that the z longitudinal axis
vibration induced fatigue failure of rotor blades is a Oa pre-twist per unit length
problem of major concern to the designer of turbo- a pre-twist of blade element
machines. Houbolt and Brooks'1' introduced a rotary oii, ai+i pre-twist at nodes i and i + 1
inertia term into the equation of motion of pre-twisted aL total pre-twist of blade
beams. Carnegie'2' derived the equation of motion, in- /3 non-dimensional total deflection in the x
cluding shear and rotary inertia for a pre-twisted, direction
asymmetrical cross section beam. The application of j8' total slope in the zx plane
finite difference approximation to linearly pre-twisted ij non-dimensional co-ordinate z/l
cantilever beams was developed by Carnegie and Q bending slope in the zx plane
Thomas'31. A. frequency parameter = pALip'i/Elx.x,
Finite element models have been used by Carnegie, v Poisson's ratio
Thomas and Dokumaci'4-5-61 to determine the free vibra- p mass density
tion characteristics of pre-twisted blade excluding the (f) bending slope in the zy plane
shear and rotary inertia effects. ifi non-dimensional total deflection in the y
Abbas and Thomas'7-8) developed a finite element direction
model with nodal degrees of freedom which can satisfy i|»' total slope in the zy plane
all t h e geometric a n d natural boundary conditions of a
thick non-pre-twisted beam. It has been shown in refer- 3. ELEMENT MODEL
ences' 9 - 10-u) that t h e developed element gives excellent A linearly pre-twisted blade element is shown in Fig. 1.
results in the analysis of free vibration a n d dynamic T h e principal co-ordinate system gxy" through centroid g
stability. rotates a t the rate of a 0 radians per unit length as it moves
This analysis presents t w o new finite element models, along t h e blade longitudinal axis z. T h e generalised co-
based o n t h e element developed b y A b b a s a n d ordinates a t each node are ijj, t h e non-dimensional total
Thomas' 7 - 8 ' for the dynamic analysis of thick pre-twisted deflection in the y direction; /3, the non-dimensional total
blades. deflection in the x direction; ij/, t h e total slope in the zy
plane; /3', t h e total slope in the zx plane; <£, the bending
2. NOTATION slope in t h e zy plane; 0, the bending slope in t h e zx
A cross-sectional area plane; i// a n d 8', the first derivatives of the bending
E modulus of elasticity slopes. This results in a n element of 16 degrees of free-
F ratio of flexural rigidities = Ely.y, / BIrx,
d o m thus enabling t h e expression for \fj, /3, <f) a n d 6 to
G modulus of rigidity
*xx> *yy second moments of area of cross section
about x and y axes
*xy product moment of area of cross section
k shear coefficient
L blade length
I element length
P circular frequency

The Author:
Dr. B. A. H. Abbas, Department
of Mechanical Engineering,
College of Engineering, Univer-
sity of Basrah, Iraq.

Paper No 732. Figure 1. A pre-twisted blade element.

450 Abbas Aeronautical Journal November 1979

Downloaded from http:/www.cambridge.org/core. Monash University, on 27 Dec 2016 at 11:00:24, subject to the Cambridge Core terms of use, available at http:/www.cambridge.org/core/terms.
http://dx.doi.org/10.1017/S0001924000031985
 contain four undetermined parameters each which can where F is the ratio of flexural rigidities and equals to
be replaced by the 16 nodal co-ordinates. EIyy,IEIx.x, and a is the angle of pre-twist of the blade
element and is given by
4. STIFFNESS AND INERTIA MATRICES
The strain energy, U, and kinetic energy, T, of an (1) for model I, a = ai + a0,r];
elemental length, /, of a thick pre-twisted blade are (2) for simplified model II, a= ( a i + 1 - a , ) / 2 .
given by
Upon assuming cubic polynomial expansion for \p, /3,
cj) and 6 of the forms
U =
0 0

ilj=^arr]r,l3=^brVr, $= ^crif, 0= ^ <W'

substituting into eqns. (1) and (2) and replacing the coeffi-
cients a„ b„ cT and d r (r = 0, 1, 2, 3) by the nodal co-ordin-
ates, the inertia and stiffness matrices for the two models
+ -^- kAGl
J(£-')V • (1) are obtained.

o 5. BOUNDARY CONDITIONS
i l l The geometric boundary conditions at the root of the
blade are that:

(0the total deflection in the x direction is zero, that

is, /3 = 0;
i- pi j IyyPdV + pl J 4»c&0d7? (2) (ii) the total deflection in the y direction is zero, that
is, i// = 0;
(Hi) the bending slope in the zy plane is zero, that is,
where i// and /} are the non-dimensional total deflections </> = 0;
in the y and x directions respectively; (f> and 0 are the (iv) the bending slope in the zx plane is zero, that is,
bending slopes in the yz and xz planes respectively; /„, 0 = 0;
and /„„ are the second moments of area of cross-section
about the x and y axes respectively and Ixy is the product The natural boundary conditions at the tip of the
second moment of area of cross section, blade are that:
*xx> *yy a n a iT can be written as
(v) the bending moment in the zy plane is zero, that
Ixx = Ix'x/ (F sin 2 « + cos2 a ) is, $ ' = 0;
hv = hv (F cos 2 a + sin2 a ) (v0 the bending moment in the zx plane is zero, that is,
/_, = /,,„ « F - l ) / 2 ) sin 2 a f?' = 0;

TABLE I

Frequency parameter for thick pre-twisted blade

using four element idealisation
a L = 30 F=4 *=5/6 e=0-3

Square root of rotary inertia parameter \/R

Mode Element model
002 004 006 008 0-10

model I 12-32 1204 11-59 11 03 10-39

1 model II 12-33 1205 11-60 11 04 10-40
reference 3 1205 11-57 10-86 10-16

model I 47 04 43-37 38-49 33-36 28-59

2 model II 47-23 43-52 38-60 33-44 28-64
reference3 44-57 39-76 34-40 29-87

model II 472-83 407-49 331-96 263-52 207-65

3 model II 471-31 406-62 331-71 263-66 207-99
reference 3 404-86 315-31 241 01 187-66

model I 1541-18 1033-61 662-61 435-17 297-58

4 model II 1552-48 1037-18 663-14 434-81 29703
reference 3 1121-48 771 01 504-66 343-45

model I 3541-16 2613-10 1856-72 1336-54 985-24

5 model II 3525-68 2611-23 1859-78 1340-68 99007

Aeronautical Journal November 1979 Abbas 451

Downloaded from http:/www.cambridge.org/core. Monash University, on 27 Dec 2016 at 11:00:24, subject to the Cambridge Core terms of use, available at http:/www.cambridge.org/core/terms.
http://dx.doi.org/10.1017/S0001924000031985
 TABLE II

Frequency parameters for thick pre-twisted blade

using four element idealisation
CM. = 6 0 ° F=4 fc=5/6 v=0-3

Square root of rotary inertia parameters \ 7?

Mode Element model
002 004 006 008 0-10

model I 12-49 12-17 11-67 11 05 10-35

1 model II 12-52 12-20 11-71 1109 10-39

model I 44-51 41-32 36-98 32-33 27-91

2 model II 45-20 41-88 37-42 32-65 28-13

model I 505-94 42701 339-74 263-82 203-66

3 model II 50004 423-57 338-64 264-15 204-65

model I 1398-38 979-57 648-65 435-19 302-21

4 model II 143202 992 08 651-23 434-46 300-74

model I 3904-36 2751-14 1895-82 1338-53 967-55

5 model II 3855-32 2744-86 1906-15 1352-42 981 -79

(vii) the shear force in the zy plane is zero, that is, reported by Carnegie and Thomas' 3 ' using the finite differ-
t/>' —</> = 0, and ence method. For thick blades with 30° and 90° angles
(viii) the shear force in the zx plane is zero, that is of pre-twist, the average differences between Ref. 3 results
&-6 = 0. and model I results are 6 • 1 % and 1 • 4 % respectively.
Table I shows that the simplified model gives results
which differ by only an average of 0-3% from model I
6. RESULTS
results for a blade with 30° angle of pre-twist. Tables
Table I presents the first five frequency parameters II and III show that for blades with 60° and 90° angles
(A + pAPp'/EIx'x') of free vibration of a thick blade of pre-twist, the average differences between the two
having a 30° angle of pre-twist {a,) and a ratio of flexural models are not more than 0-8% and 1-5% respectively.
rigidities (F) of 4. The values of the square root of the
rotary inertia parameter (R^Ix'x'jAP) are shown as
*JR = 0-02, 0-04, 0-06, 0 0 8 and 0 1 0 . The value of 8. CONCLUSIONS
shear coefficient (k) is taken as 5/6 and that of Poisson's The finite element model I developed for the dynamic
ratio (v) is taken as 0-3. For model I, the angle of analysis of thick pre-twisted blades gives very accurate
pre-twist is taken as linearly increasing along the length results for the natural frequencies of free vibration. It
of the blade i.e. a = at> + a„r). For model II which is a has the facilities to satisfy all the geometric and natural
simplified model and numerically less involved than model boundary conditions.
I, the angle of pre-twist is taken as a = ( a , + i - a 1 ) / 2 . In order to reduce the analytical work involved in
Table II presents the first five frequency parameters of deriving model I, a simplified approach is developed and
free vibration (A) of a thick blade having 60° angle of used to derive element model II. It is concluded that the
pre-twist and a ratio of flexural rigidities of 4, while simple model gives values, for the frequency parameters,
Table III presents the case of a thick blade having 90° not far away from the values obtained by the first model
angle of pre-twist and flexural rigidities ratio of 4. which is numerically more involved.
Results extracted from Figs. 6 and 8 of Ref. 3 are For thick blades with 30°, 60° and 90° angles of pre-
shown in Tables I and II for comparison. twist, the average differences between model I and model
II results are not more than 0 - 3 % , 0 - 8 % and 1-5%
respectively.
7. DISCUSSION For thick blades with 30° and 90° angles of pre-twist,
This paper presents two new finite element models for the average differences between model I results and the
the dynamic analysis of thick pre-twisted blades. The available approximate results from a finite difference
models are the only models presented so far in which both analysis are 6 - 1 % and 1-4% respectively.
geometric and natural boundary conditions can be cor-
rectly applied. The boundary conditions are \jj = 0, /3 = 0, REFERENCES
$ = 0 and 0 = 0 at the root of the blade and <j>' = 0, 6'= 0, 1. HOUBOLT, J. and BROOKS, G. Differential equations of
\p' = (f) and /3' = 0 at the tip of the blade. motion for combined flapwise bending, chordwise bend-
The convergence of the results obtained from model ing and torsion of twisted non-uniform rotor blades.
I is very rapid and of the same order as the model NACA report 1346, 1958.
developed in Refs. 7 and 8. 2. CARNEGIE, W. Vibration of pre-twisted cantilever blading
allowing for rotary inertia and shear deflection. Journal
It is seen from Tables I, II and III that model II, of Mechanical Engineering Science Vol 6, No 22, pp
which is simpler and numerically less involved than model 105-112, 1954.
I, yields very good results for the natural frequencies of 3. CARNEGIE, W. and THOMAS, J. The effects of shear de-
free vibration. formation and rotary inertia on the lateral frequencies
of cantilever beams in bending. Journal of Engineering
Tables I and III compare the results obtained by the for Industry, Transactions of the ASME, Vol 94, No 1,
present finite element models with the available results pp 267-278, 1972.

452 Abbas Aeronautical Journal November 1979

Downloaded from http:/www.cambridge.org/core. Monash University, on 27 Dec 2016 at 11:00:24, subject to the Cambridge Core terms of use, available at http:/www.cambridge.org/core/terms.
http://dx.doi.org/10.1017/S0001924000031985
 TABLE III

Frequency parameters for thick pre-twisted blade

using four element idealisation
a L =90° F=4 *=5/6 ^=0-3

Square root of rotary inertia parameter VR

Mode Element model
002 004 006 008 0-10

model I 12-76 12-39 11-82 11-12 10-36

1 model II 12-83 12-45 11-89 11-19 10-43
reference 3 12-64 11-84 11-10 10-19

model I 41-10 38-45 34-77 30-73 26-78

2 model II 42-43 39-57 35-67 31-42 27-29
reference 3 4009 35-51 31-33 27-31

model I 55805 457-17 351-92 265-19 199-47

3 model II 545-73 449-96 349-42 265-43 200-91
reference3 456-77 351-71 259-21 196-41

model I 123829 908-37 62605 431-99 305-74

4 model II 1290-69 931-22 632-50 43201 30409
reference 3 92608 630-81 429-49 308-69

model I 4409 03 2950-54 195607 134504 950-11

5 model II 4328-57 2940-68 1973-90 1368-47 971-56

DOKUMACI, E., THOMAS, J. and CARNEGIE, W. Matrix dis- THOMAS, J. and ABBAS, B. A. H . Finite element model
placement analysis of coupled bending—bending vibra- for dynamic analysis of Timoshenko beam. Journal of
tions of pre-twisted beams. Journal of Mechanical Engi- Sound and Vibration, Vol 41, N o 3, p p 291-299, 1975.
neering Science, Vol 9, N o 4, p p 247-254, 1967. ABBAS, B. A. H. and THOMAS, J. The second frequency
THOMAS, J. and DOKUMACI, E. Simple finite elements for spectrum of Timoshenko beams. Journal of Sound and
pre-twisted blading vibration. Aeronautical Quarterly, Vibration, Vol 51, N o 1, p p 123-137, 1977.
Vol 25, pp 109-118, May 1974. 10. THOMAS, J. and ABBAS, B. A. H . Dynamic stability of
DOKUMACI, E. Pre-twisted beam elements based on ap- Timoshenko beams by finite element method. Journal of
proximation of displacements in fixed directions. Journal Engineering for Industry; Transactions of ASME, Vol
of Sound and Vibration, Vol 52, N o 2, p p 277-282, 98, N o 4, p p 1145-1151, 1976.
1977. 11. ABBAS, B. A. H . and THOMAS, J. Dynamic stability of
ABBAS, B. A. H . Mechanical vibration and dynamic fixed—fixed and free—free Timoshenko beams resting on
stability of complex structures by finite element method. an elastic foundation. Paper N o 77—DET—134 an ASME
Ph.D thesis, University of Surrey England, 1977. publication.

Aeronautical Journal November 1979 Abbas 453

Downloaded from http:/www.cambridge.org/core. Monash University, on 27 Dec 2016 at 11:00:24, subject to the Cambridge Core terms of use, available at http:/www.cambridge.org/core/terms.
http://dx.doi.org/10.1017/S0001924000031985