Technical papers

Rocket vehicle loads and airframe design

Introduction
Rocketeers have generally followed a Darwinian evolutionary approach to designing their
vehicles to withstand flight loads: if it breaks, try something else! This approach has worked
remarkably well, but aircraft designers have better methods that allow prediction of the
airframe loads before flight, especially for very high speed/high altitude flights that come to
grief too often.

The major disturbances affecting the trajectory are thrust misallignments and winds. The
wind, especially upper-atmosphere wind (such as the jet-stream) is by far the more significant
of the two, as it affects the flight path, causing vehicle airframe bending and trajectory
deviations.

A lot of the missile design literature from the 1960‟s has now been declassified, so we can
see how they were designed against wind effects.

Recovery system (parachute) loads are not covered here; see our paper „Recovery system
design for large rockets‟ for these loads.

Part 1: Winds, atmospheric turbulence, and gusts

In addition to the compressive loads caused up the fuselage by the rocket motor‟s thrust, the
vehicle is subjected to aerodynamic forces during flight which significantly affect the airframe
design.
As well as the axial loads along the airframe caused by drag forces, surprisingly large forces
normal to the centreline of the airframe (lift forces) can be created by gusts and other
transient wind effects:
U

The sudden gust velocity vector „U‟, when added to V

the vehicle‟s airspeed vector „V‟, causes a small
angle of attack :
It‟s this angle of attack that creates normal forces that
create the side-loads within the fuselage.
Rotational accelerations are also caused as the fins
try to reduce the angle of attack  back to zero.

Ground winds:
Our rocket vehicles are too small to be broken just by the wind as the vehicle leaves the
launcher, because we never launch in overly strong winds. But it‟s important to know how
strong the wind will get as the rocket gains altitude, as the magnitude of any gusts it
encounters depends on the strength of the wind.
The wind gets stronger with altitude as the rocket ascends the wind‟s boundary layer or „wind
gradient‟.
 h2 
P

An empirical engineer‟s formula for predicting the wind at altitude is: V 2  V 1  where
 h1 
V1 is the wind measured at height h1 near the ground, and V2 is the wind speed at altitude
h2. The exponent, P is unfortunately terrain dependant, but for the wide open-spaces around
launch-pads the following rough values will do:
If V1 is measured at h1 = 3 metres above the ground and two-minute averaged readings are
taken, then P = 0.2 if V1 is less than 17 metres per second and P = 0.14 if V1 is greater than
17 metres per second
This formula can be used up to a height of about 150 metres, which is typically the edge of
the boundary-layer: above this height, the wind is constant with further altitude (though see
below).

Rick Newlands & Martin Heywood 1 updated: 12/04/11

 Technical papers

Wind shear and the synthetic wind profile:

Wind-shear is the rate-of-change of average wind speed with altitude: (it‟s a vector quantity,
so includes changes of wind direction with altitude too.) Note that due to fine wind structures
(nowadays suspected to be fractal) the wind-shear values increase the closer you examine
them (vertical scale of distance measured e.g. by weather radar).
A sudden change in wind speed and/or direction causes angles of attack on the airframe
which generate large side loads.
The first large wind shear occurs when the vehicle becomes free of the launcher, and
encounters the wind, but the vehicle‟s airspeed is still very low so this doesn‟t load the
airframe unduly.
nd
Our 2 flight of our FLARE hybrid was hit by a windshear after burnout but well before
apogee; it caused the FLARE to tumble.

In order to judge the effect of windshear, we first have to know the strong winds at high
altitudes, as these worst winds cause the worst windshears.
Various types of models representing the steady wind have been used to design launch
vehicles, such as statistical, stochastic (power spectral), and synthetic models.
By far the easiest to develop and use for essentially vertical trajectories (therefore not for
aircraft) is the synthetic wind profile.
Early American Launch vehicles and Britain‟s Black Arrow were all designed using synthetic
wind profiles.

Construction of a synthetic wind

profile starts with an altitude plot of
worst-case winds. The plot for
Cape Canaveral is shown here,
and was compiled from quasi-
steady (averaged over at least two
minutes) wind values obtained
from weather radar instruments at
the site over several years.

The 95 percentile wind plot shown

here describes wind speeds that
have only a 5% risk of being
exceeded for any launch trajectory
occurring at any time throughout
the year. Note that the plots are for
scalar wind speeds, i.e. the
assumption is that all winds are
blowing in the same arbitrary
direction, which we can specify to
be the worst-case vehicle loading
direction. In reality of course,
winds blow from all compass
directions, though are still
assumed to blow horizontally, as
vertical winds have insignificant
effects on Launch Vehicles.

In tabular form:

Altitude (km) 0 10 14 20 30 60 80
Speed (m/sec) 11 75 75 25.4 25.4 141.2 141.2

Rick Newlands & Martin Heywood 2 updated: 12/04/11

 Technical papers

Note the first wind peak between 10 and 14 Km, this is common all around the world and
describes the jet stream, a high-speed ribbon of air. The worst wind shears tend to occur here
(Ref. 3) so this region tends to be the one that the vehicle structure has to be designed for.

To the wind profile, we then add a severe wind shear, usually at the same altitude as the
strongest winds.
Here are three synthetic
wind profiles (Ref. 3)
devised by the
meterorologist Norman
Sissenwine for windy
regions of the USA. The
latter profiles he released
a few years after the first
when better weather radar
became available.
Note the windshears
(where the curves are
nearly horizontal: shears
are the inverse of the
curve gradient) at around
40,000 feet (12 Km) in all
three profiles.
In the 1959 profile, two
shears are noted: a lower
one (0.0375 fps/per foot)
for a 3000 foot
investigation by radar, and
a higher shear (0.075
fps/per foot) for a 1000
foot investigation. The
lower shear is for the
lower half of the point
(wind increasing to 300),
followed by the sudden
drop in windspeed (the
higher shear of 0.075).
Each profile is for a 99%
probability: actual weather
is expected to exceed
these profiles (wind and windshear) only 1% of the year.

In tabular form, the 1959 Sissenwine profile is (converted to metric):

Altitude (km) 0 5.3 7.5 8.2 9.3 9.9 11.1 23.0 30.5
Speed (m/sec) 10.9 30.6 39.9 46.1 89.7 46.0 39.9 22.4 27.1

You could replace the zero-altitude data in the table above to suit your needs, for example
input your launchsite‟s elevation above sealevel, and use the United Kingdom Rocketry
Association‟s maximum allowed launcher windspeed of 8.94 metres/sec (20 miles per hour).

The Sissenwine profiles are a simple design tool that we can use for high altitude flights, and
are easily fed into a trajectory simulator. Many vehicles including the British Black Arrow,
used a Sissenwine profile for their design even though Black Arrow was flown over Australia
rather than the USA.

Rick Newlands & Martin Heywood 3 updated: 12/04/11

 Technical papers

Here is another, later,

set of wind profiles (1 to
5) for Cape Canaveral,
that were used to design
early versions of the
Saturn 5.
The yellow curve is the
worst case (95%) winds
that we saw earlier.
To this have been added
severe wind shears
(99%) at various
altitudes (curves 1 to 5).
The wind speed
percentile used is lower
than the wind shear
percentile used simply
because the wind speed
can be measured quickly
by radar just before
launch as a safety-
check, and the launch
halted if the wind is too
high, whereas severe
time-consuming number-
crunching is needed to
extract the wind-shear
from radar data.
Since the chances of the
worst-ever wind speed
occurring at the same
time and place as the
worst-ever wind shear
and worst-ever gust are
fairly remote, the above
Synthetic profiles are
conservative.
So NASA reduces the
shear and gust values by
15%, based on
engineering judgment, before applying them to a sim. If curve 1 above was reduced by 15%
then its gradient would be 15% steeper (lower shear) at all points.

Suppose, for example, that burnout (maximum airspeed of your vehicle) occurred at around 6
Km altitude (20,000 feet) so that you are interested in the effect of a windshear at 6 Km. Then
you would use curve 1, which adds a windshear that meets the yellow wind curve at a height
of 6 Km. Note that as you ascend curve 1, the gradient of the curve flattens, showing that the
windshear is getting worse the closer you get to the yellow curve. This describes the
aforementioned fact that the windshear gets more severe the smaller the length scale you
examine it at.
Once curve 1 hits the yellow wind curve, you then just follow the yellow curve upwards from
there.
Construction of the shear curves (1 to 5 above) is described in appendix 1.

Rick Newlands & Martin Heywood 4 updated: 12/04/11

 Technical papers

Gust loads:
The wind shears are at their worst only at high altitudes, so probably can be ignored for low
altitude flights providing your vehicle is sufficiently aerodynamically stable. Gusts, however,
occur at all altitudes so have to be designed for.
The highest aerodynamic loads suffered by the vehicle‟s airframe occur at „max q‟ where q is
dynamic pressure (see glossary), which varies with the square of the vehicle‟s airspeed.
Typically, the highest airspeed, and so the highest max q, occurs just at burnout.
If the vehicle is hit by a strong side-gust at max q, then the aerodynamic loads caused by the
gust can cause structural breakup of the vehicle, if not designed for.
The gust is pictured as a horizontal layer in the sky that the vehicle ascends into.

Gust penetration effects:

The simplest gust representation is the immersion gust, where the vehicle is small enough
compared to the gust wavelength (vertical length) that it can be considered as a point, and the
appropriate aerodynamic loads are obtained instantaneously as the point traverses the gust.
Our vehicles are small enough to use the immersion gust assumption.
However, a more accurate analysis includes penetration effects. This is known as quasi-
steady gust penetration. „Quasi-steady‟ means that aerodynamic inertia effects are ignored,
and steady aerodynamic coefficients are instantaneously reached as each fuselage station
passes the gust front. The vehicle is represented as several lifting sections (nosecone and
fins) separated by essentially non-lifting sections (fuselage) in between.
There is therefore a time lag between the nose entering the gust and the fins entering the
gust, i.e. the nosecone starts lifting before the fins do.

Modelling gusts
There are two approaches taken by
aeronautical engineers for modelling
gusts: a discrete gust, and
continuous turbulence.
In the former, a single strong gust is
added to the synthetic wind profile
just as the vehicle reaches max q.
This was the approach taken with the
design of the UK Black Arrow
launcher for example.

NASA has noticed particularly nasty

sharp-edged gusts of 9 metre/sec
amplitude above the steady wind,
which represent convenient worst-
case gusts.

The Synthetic profile used by the Marshall Space Flight Centre for Saturn 5 design work took
the 95 percentile wind and 99 percentile wind-shear (see previous pages), then slapped a 9
metre/sec gust onto it at the altitude of interest: here a gust (in red) has been added to
windshear curve 3 from the previous diagram.
Note the buildup rate at the leading and trailing edges of the gust of 9 meters per second per
25 meters altitude.
Using engineering judgement, NASA then reduced the gust amplitude by 15% to make it a
7.65 meters per second gust.
Note that your sim must increment in height steps much less than 25 metres or it could miss
the gust buildup curves and experience an „infinite‟ windshear.

For Black Arrow‟s design, a 50 feet per second (15 metre per second) sharp-edged gust was
added to the Sissenwine profile at max q (Ref. 7). This gust profile was rectangular, so was in
effect simulating an infinite windshear (zero graph gradient) though their sim just regarded it
as a perfectly valid step-impulse input. This was known to have been rather conservative, but
on the other hand, the vehicle structure performed flawlessly.

Rick Newlands & Martin Heywood 5 updated: 12/04/11

 Technical papers

I suggest that a less conservative gust might be 15 metres per second, but use the
Sissenwine worst windshears of 0.0375 per sec and 0.075 per second for the leading and
trailing edges of the gust.

The second approach adds a continuous stream of random gusts to the wind profile
throughout the flight, and attention is paid to the flight around max q.
In the aerospace industry, several well-tried formulae have been developed to model gusts. A
continuous series of gusts is known as turbulence.
The effect of turbulence can be modelled by summing the steady flow of air and a random,
zero-mean turbulence velocity. Two central aspects of the turbulence velocity are the
amplitude of the variation (the strength of the turbulence, its wind speed) and the frequencies
at which they occur. Therefore a reasonable turbulence model is achieved by generating a
sequence of random numbers that produces a sequence with a similar distribution and
frequency spectrum as that of real wind.
Several models of the spectrum of wind turbulence at specific altitudes exist.
Two commonly used such spectra are the Kaimal and von Karman wind turbulence spectra:

( ) ( )
( ) ( ( ) )

Here Su(f) is the spectral density function (whatever that is!) of the turbulence velocity and f is
the turbulence frequency,  is the standard deviation of the turbulence velocity (more on this
below), L is a length parameter and U is the average wind speed.
Ref. 1 says that you can ditch the „1‟ on the denominator of either equation with very little loss
of accuracy (4%). Cancelling out the resulting equations gives:

( ) ( )

Or in other words, both equations are proportional to

so both equations can be reasonably modelled by utilizing white noise that has been altered
by a suitable digital filter to give a spectrum of (this is now called pink noise).
Appendix 2 shows how to code this in C++ software.

The use of the standard deviation  tells you something about the nature of turbulence.
Being a natural
phenomenon, it has the
property that if you were
to measure a sequence
of gusts over a few hours
and plot on a graph how
often a gust of a
particular strength
occurs, you‟d find that
they have a Gaussian
distribution as shown
here:

Or in plain English, the majority of gusts are weak, and cluster around the average wind
speed . 68.2% of gusts occur within a width on the x-axis defined as between + and – one
standard deviation , with only an extremely rare occurrence of a gust stronger than three .
In structural engineering, they define the „turbulence intensity‟ as which is the
percentage that one standard deviation of the gusts is of the average wind speed.

Rick Newlands & Martin Heywood 6 updated: 12/04/11

 Technical papers

Rearranging gives the gust strengths:

Buildings will break above an I of 20%, so 15% would seem a suitable design maximum for
our vehicles; this‟ll provide a safety-margin as we wouldn‟t fly in weather that gusty.

The whole point of feeding pink noise turbulence into a simulator is that the noise
encompasses a whole range of frequencies, one of which should excite the natural pitching
frequency of the finned vehicle and produce a large angle of attack response (this pitching
frequency changes with airspeed, and loss of mass out the nozzle).

Which axes plane?

If your sim models a three-degree-of-freedom trajectory (X,Y, and pitch rotation) then the
winds are input into the plane of the trajectory (X, Y).
However, if your sim is a six-degree-of-freedom trajectory then Ref.2 says that you would
apply the varying winds (wind profiles, gusts, shears) at 90 degrees to the plane of the
trajectory (=yaw plane) as this gives the worst loads, and just apply a constant wind in the
plane of the trajectory (pitch plane) to bend the trajectory as the vehicle leaves its
launchtower. However, for our nearly vertical trajectories either plane will do.

The response from the vehicle

Having fed gusts, winds, and windshears into the sim, the sim will produce a vehicle
response. The data that we need output from the sim are: the dynamic pressure q, the normal
and axial accelerations caused on the vehicle, the pitching angular acceleration R around the
vehicle CG, and the maximum angles of attack on the fins and nosecone and the lifts that
these cause. The sim described in our paper „A dynamic rocket simulator‟ will give you this
data.
You then apply these variables to the airframe to see what loads they create as we shall find
out in the next section:

Rick Newlands & Martin Heywood 7 updated: 12/04/11

 Technical papers

Part 2: Airframe loads

Both thrust and aerodynamic loads, which are treated mathematically as forces externally
applied to the vehicle, vary in magnitude and direction, and are resisted only by the vehicle‟s
inertia.
This „inertial resistance‟ causes „inertial loads‟ within the structure as it accelerates in
response to the external forces.
The thrust is taken to act at the nozzle exit.

First approximation
As a first approximation, the French military suggest that the worst aerodynamic loads on the
vehicle in windy conditions are of the same order of magnitude as the launch weight of the
vehicle. This is borne out by sims of level 2 HPR (High Power Rocketry) class vehicles.
However, a gust and windshear occurring at max q will give larger loads so don‟t fly in that
kind of weather!

Aerodynamic loadings:
The distribution of the aerodynamic forces along the fuselage (in directions normal and axial
to the long axis of the vehicle) determines how the vehicle suffers when gusts or windshears
cause sudden angles of attack.

A typical such subsonic aerodynamic Lift force distribution looks something like this:

Note there is negligible „carry-over‟ of pressure from the

nosecone rearwards to the rest of the fuselage. In
contrast, for supersonic airspeeds there is significant
carry-over.

The fin Lift is concentrated over a very small axial distance, so items such as fins, shrouds,
boat-tails, or large protruberances have local normal (lateral) forces that are most easily
modelled as discrete loads acting at their respective centres of pressure.
Strictly, the nosecone is often too long to be dealt with as a concentrated load, but we shall do
so, as nosecones are usually rigid structures: they will deform little.

Mass (inertial) loadings:

D‟Alembert‟s principle (a fore-runner/pre-statement of Newton‟s 3rd law) describes mass as
that which resists acceleration by causing an „inertial load‟ (Newton called this „Inertia‟:
ma = F)
For a rocket vehicle, any unbalanced global external loads (thrust greater than drag, lift forces
caused by a sudden gust-induced angle of attack) produce a global vehicle acceleration:

a global 
F (Newton‟s 3rd law)
m
This acceleration causes local „inertial loadings‟ caused by local masses (such as, say,
equipment or payload masses) acting on their supports within the vehicle structure, which are
straight-forwardly determined using D‟Alembert‟s principle: F = mlocal aglobal

Don‟t forget that the global vehicle mass changes with time due to fuel mass being ejected
out the nozzle during the burn.

To analyse the effect that the inertial loadings have on the vehicle structure, the total (global)
mass of the vehicle is split into a distribution of discrete (local) component masses at various
geometric stations along the length of the vehicle.
Because we are interested in airframe structural performance, the mass distributions we
chose to consider for the analysis of normal components of inertial loadings may or may not
coincide with distributions considered for axial components.

Rick Newlands & Martin Heywood 8 updated: 12/04/11

 Technical papers

For example, the mass of a fuelled propellant tank is supported axially by a bulkhead at the
bottom of the tank, therefore we can represent the axial inertial load of the tank mass
(including propellant) as being concentrated at a single station at the base of the tank,
whereas its normal component is supported by the tank wall, spread over the length of the
tank.

Provided equipment masses are spread evenly around the vehicle central axis, (please make
sure they are!), then structural near-symmetry results in inertial near-symmetry, with
correspondingly negligible inertial coupling between the pitch, yaw, and roll planes, which
means that motion in these individual planes can be modelled separately from one another.

Axial loads:
To evaluate the effects of combined loadings at some airframe cross-section which is
distance „x‟ from the base of the vehicle, all axial loads (thrust T, drag D) are simply added
together:
x
A( x)  T  Dno sec one  D fuselage  Dbase  D fin  S P  a x  m x
X0
The sum is started from x0 (the aft end). Note that it is important to assign the correct signum
to the forces: the thrust is in the opposite direction to the drags so has the opposite sign. Also,
forces causing compression in the airframe are signed negative by convention (tensile forces
are positive).

The first term above, T is the thrust acting at x0.

The Dfuselage term is predominately the skin-friction drag of the fuselage between X2 and X1, its
length. It can be expressed by integrating the drag per metre, or more usefully, drag per
calibre „C‟, where a calibre is x/d, where d is (the widest part of) the fuselage diameter:

The Dbase term is the combined base and boat-tail drag, acting pretty-much at the nozzle
position.

The fins drag term is Dfin, acting at the fins centre of pressure.

S P is the tensile axial force caused by any fuselage compartment internal pressure:
ground-level air trapped within the fuselage as the vehicle rises into outside air that has less
pressure at altitude. (S is fuselage cross-sectional area, and P is the difference in pressure
between inside and outside.)

The last term is the inertial load contribution of all masses tailward of the fuselage section „x‟,
and is multiplied by the vehicle‟s axial acceleration „ax’.
Continuous mass distributions (such as fuselage walls) can be expressed as a mass per
metre or per calibre then integrated over the length (x-x0), as was done for the fuselage drag
(above).

When applying the above equation, it‟s easier to start at the nosetip and work backwards, and
then the drags and masses add up to a maximum at the engine thrust bulkhead.

For example, here‟s a vehicle just before engine burnout. The propellant mass is low, but the
thrust is high. The other main inertial loads are the avionics and their batteries, and the main
„chute mass. Note the change in the graph sign caused by the thrust acting at the engine
bulkhead; the structure tailwards of this bulkhead is in tension while the structure nosewards
of this bulkhead is in compression. The graph slope is caused by the fuselage drag term.

Rick Newlands & Martin Heywood 9 updated: 12/04/11

 Technical papers

You can see that the largest loads suffered by the fuselage occur at the location of the engine
bulkhead.
Note how individual loads are suddenly switched on only when the graph reaches their axial
position; they „don‟t exist‟ before their position is reached. This is the theory of the Method of
Sections which is explained below.

The maximum aerodynamic axial load (max drag) will occur at max q (see glossary) whereas
the maximum thrust will occur at launch. Maximum inertial loads will occur at maximum
longitudinal acceleration: this is typically around engine burnout for a solid motor, but may be
much earlier for the diminishing thrust from a hybrid engine.
Monitor your sim to find the point in the trajectory where the peak load on the engine
bulkhead occurs.

Normal loads:
The loads normal to the long axis of the vehicle, in contrast, cannot simply be added together
like axial load components.

The fuselage structure can be modelled as a cylindrical beam, which is unsupported at each
end (known as a „free-free beam‟).
Using standard engineering beam theory, the normal loads can be reduced to an equivalent
system of Shear Forces and Bending Moments, (see below) and these equivalents can be
simply added.

Rick Newlands & Martin Heywood 10 updated: 12/04/11

 Technical papers

Beam theory:
Visualise a rigid 2-dimensional beam (i.e. of arbitrary cross-section) rigidly supported at the
left-hand end by a wall, free at the other end, and loaded at some cross-section along its
length by a suspended mass „M‟:

If the weight (Mg) of the mass „M‟ is too great, it‟ll shear the beam apart at the point of its
application. The beam is obviously designed to withstand this normal load, and note that the
wall is obviously able to supply an equal load in the opposite direction (action and reaction),
otherwise the beam would fall:

Mg Mg

So the beam is in static equilibrium, i.e. there is no acceleration vertically.

The engineering method known as the Method of Sections can be used to analyse any
longitudinal station along the beam.
The trick is to pretend that you‟ve sliced the beam at the point of interest, and then you apply
appropriate forces in appropriate directions to keep the beam stationary.
From the principle of action and reaction, the „Shear‟ force you pretend to apply at the slicing
point is equivalent to the shearing force actually occurring within the material of the beam at
that station.

For example, if you were to make a pretend vertical cut in the beam at any point between the
wall and the load „M‟, you would have to apply an upward „shear‟ force of magnitude Mg to
support „M‟, so a graph of actual Shear Force (V) within the beam, with beam distance, looks
like this:

Mg

0
wall

Obviously, if you cut it, the cut portion of the beam would rotate downwards about the cut, so
you also have to apply a pretend Moment (B) for rotational equilibrium:

B
where B =-Mg(x2-x1)

Mg
x0 x1 x2

Again, from action and reaction this pretend Bending Moment at the pretend cut is equivalent
to the actual moment occurring within the beam at the imaginary cut station, caused by the
load M.

Rick Newlands & Martin Heywood 11 updated: 12/04/11

 Technical papers

Note the distance dependence in Bending Moment calculations.

In actual fact, for simple beam theory, the Bending Moment graph is simply the integral of the
Shear Force graph with respect to distance along the beam (x) therefore, for the complete
beam shown above (beam itself assumed massless), the graph of Bending Moment with
distance from the wall is:

In this diagram I‟ve used the usual sign conventions for Shear Force and Bending Moment
(opposite to that for concrete structures) which are:

Positive Shear Forces will spin a segment of beam clockwise, and positive Bending Moments
will cause a section of beam to sag downwards. There are lots of examples of Shear Force
and Bending Moment diagrams on the web, usually for static beams.

In the real world, no beam is infinitely rigid, so the bending moment stretches the top surface
of the beam, and compresses the bottom surface of the beam, so it‟s important to know the
Bending Moment maximum values, and where along the beam they occur, because too great
a stretch will tear/fracture the beam, and too great a compression will crumple the beam.
In the above case, the beam would snap where it meets the wall if too great a load was hung
off it, because that‟s where the maximum Bending Moment occurs.

Rocket ‘beam’ theory:

In the above example, the beam was restrained from rigid-body movement and rotation by the
wall, (though a little elastic deflection might occur) so that it was in static equilibrium.
In contrast, for a rocket vehicle fuselage, the only partial restraint against lateral aerodynamic
loadings (and vectored thrust) is the rocket‟s inertia („inertial loads‟, see above).
The Bending Moments created by the Shear Forces (caused by the aerodynamic loadings)
are only partially restrained by the fuselage‟s rotational moment of inertia, and also by the
aerodynamic moment caused by the fins, so the Bending Moments cause the vehicle to rotate
around its centre of mass (C.M. or C of G) with some value of rotational acceleration.
(Remember basic physics: unrestrained „rigid‟ vehicles always rotate about their centre of
mass.)
Also, the lateral loads cause the vehicle to linearly accelerate laterally.
So the vehicle is not in static equilibrium, but we can still use the above beam theory because
it is in dynamic equilibrium: the vehicle‟s lateral acceleration ay is the sum of the lateral loads
on the vehicle divided by its total mass (Newton's second law). This acceleration causes
„inertial loads‟ within the fuselage which load it.

Lateral Shear Force:

This is caused by the externally-applied aerodynamic loads, which themselves are caused by
a gust-induced angle of attack. As the vehicle accelerates laterally and also rotates, further
Shear Forces are caused by „inertial loads‟ from equipment masses.

Rick Newlands & Martin Heywood 12 updated: 12/04/11

 Technical papers

The equation for the Shear Force occurring at (@) some longitudinal station „Xk‟ is

( ) ∑ ∑ ( )

The sum is started at station 0 (the aft end).

Continuing the philosophy of the Method of Sections, the nosecone lift Lnosecone term is only
included if „x‟ is nearer the nosetip than the nosecone‟s centre of pressure, and so on.
Note that I‟ve assumed that the lift is concentrated at the nosecone‟s centre of pressure
because the nosecone is rigid. In reality, it‟s spread over the length of the nosecone, giving:

∫ where C is calibers (x/d) and  is angle of attack.

The boat-tail lift term Lboat-tail is the boat-tail lift force. Note that boat-tail „lift‟ occurs in the
opposite direction to fin lift, so is negative.

The fin lift term is Lfin acting at the fins centre-of-pressure.

The ay summation term is the inertial load contribution of all masses „m‟ tailward of the
fuselage section ‟XK‟, and is multiplied by the vehicle‟s lateral acceleration ay (metres/sec).
Each component mass „m‟ gets „switched on‟ as XK passes it, and acts at its own position
„ ‟.

The last term in the Shear Force equation above represents rotational inertia loads caused by
the rotational acceleration „R‟ (radians/sec) about the vehicle‟s centre of gravity ‘xcg’ . For
example, if the nose rotates such that the nose accelerates in the same direction as the nose
lift, then the inertial Shear Force on the nose‟s mass is increased.
The „ ‟ in this term is the position of each individual mass.
Take care with your defined direction of positive R.

Continuous mass distributions (such as fuselage walls or main parachute) can be expressed
as a mass per metre then integrated over their length, in which case the last two terms are:

∫ ∫ ( )

Bending moment:
The equation for the Bending Moment occurring at (@) some longitudinal station „Xk‟ is
obtained, as usual for a beam-like structure, by simply integrating the Shear Force equation
above with respect to „x‟:

( ) ( ) ( ) ( )

∑ ( ) ∑ ( )( )
Note that the term for the fins starts at the fins position xfin for example.

Rick Newlands & Martin Heywood 13 updated: 12/04/11

 Technical papers

The translational and rotational accelerations are identical to those in the Shear Force
equation above.
The „ ‟ in the last two terms is the position of each individual mass „m‟.

Continuous mass distributions (such as fuselage walls or main parachute) can be expressed
as a mass per metre then integrated over their length, in which case the last two terms are:

∫ ( ) ∫ ( )( )

Example:
As an illustration of how the above equations work, imagine a vehicle where the fins lift and
nose lift are equal, and the CG is halfway between nose and fins centres of pressure. This will
mean that the vehicle doesn‟t rotate as it accelerates laterally. Also assume, for the purposes
of illustration, that the mass of the fuselage tube is zero.
We‟ll examine two loading cases: both are loaded with the same mass „M‟, and in both cases,
the inertial force „F‟ resulting from the vehicle‟s upwards acceleration acting on „M‟ is equal to
the nose plus fins lift.
On the left, the mass is split into two halves which are positioned at the fins and nose centres
of pressure. On the right, the mass is concentrated halfway between the nose and fins
centres of pressure.

As you can see, on the left we have the minimum (zero) bending moment. F1 = the fins lift,
and F2 = the nose lift (so F1 + F2 = F= nose lift + fins lift). The resulting net Shear Force is
zero, so the Bending Moment remains at zero.
On the right, F= nose lift + fins lift again, but this time there are large (maximum) moment
arms between mass „M‟ and the nose and fin centres of pressure, so you get the maximum
possible Bending Moment, occurring halfway along the vehicle.

The Bending Moment diagram shows that it‟s therefore much better to position your
equipment masses at the centres of pressure as on the left diagram: this bends the fuselage
much less. The mass of our engines tends to be rearward anyway, but we can reduce
bending by positioning our recovery systems and their batteries inside the nosecone rather
than some distance behind it.

Rick Newlands & Martin Heywood 14 updated: 12/04/11

 Technical papers

Note from the right-hand Shear Force graph that when analysing the vehicle using the
Method of Sections, each load is only „switched on‟ once the section of interest has been
passed, and the Shear Force remains at that value until the next load is reached.

Similarly, from the Method of Sections, note how the gradient of the Bending Moment graph
only changes once the Shear Force has been passed, i.e. in the above Bending Moment
equation, you don‟t add the nose lift until „Xk‟ is greater than „Xnose‟; if „Xk‟ is less than „Xnose‟
it‟s as if the nose lift doesn‟t yet exist.

Note how the Bending Moment graph starts at zero and ends at zero. That‟s because the
ends of the vehicle are unrestrained.
Also note that a positive Shear Force causes a positive Bending Moment curve gradient, and
a negative Shear Force causes a negative Bending Moment curve gradient.

For the above picture, the Bending Moment is positive, it would cause the vehicle to banana
into a shape that sagged in the middle. (In structural engineering, negative Bending Moment
causes the opposite of sagging which is known as „hogging‟; this may be a nautical term).

A more typical rocket vehicle loading looks like this:

In this diagram, the vehicle is accelerating upwards due to the nose and fin lift, but it would
also be rotating about its CG.

Rick Newlands & Martin Heywood 15 updated: 12/04/11

 Technical papers

The resulting inertial loads from motor mass‟ Mm‟ and avionics mass „Ma‟ are supported
equally by fore and aft bulkheads, and the main chute gives a distributed load along the
fuselage wall.
The point loads cause a sudden jump in Shear Force, whereas the distributed inertial loading
of the main chute is a continuous downwards ramp, which causes the gradient of the Bending
Moment graph to continuously decrease (a parabolic curve).
Again, the maximum Bending Moments occur about halfway along the vehicle here, where
the Shear Force graph re-crosses the X-axis.

I haven‟t added a boat-tail to this vehicle, but it is easily added, acting at its own centre of
pressure. Remember that boat-tail lift is in the opposite direction to nose and fin lift.

Worst loads:
Maximum Shear Forces (and therefore worst Bending Moments) will often occur at the instant
of maximum q times angle of attack. This is where the aerodynamic loads are maximum, and
the lateral and rotational accelerations are also maximum (maximum vehicle response),
giving the highest inertial loads.
Add a gust and/or windshear to your trajectory sim at max q to cause a peak angle of attack
at max q.

Testing
Exactly how much Shear Force or Bending Moment some location along the vehicle can
withstand depends on the material used, and its shape/thickness etc.
Having now gotten the largest Shear Force and Bending Moments from the above graphs, we
need to replicate these forces and moments when the vehicle is stationary on the workbench,
rather than waiting until it breaks-up in the air!
In particular, we need to load all the airframe coupler joints.

In theory, one could apply all of the aerodynamic loads at their correct point of application on
the workbench, perhaps by suspending the vehicle horizontally by its nose and fins.
Then replicate the inertial loads using weights within the airframe.
However, this would be cumbersome; you only need to apply the worst Shear Forces and
Bending Moments at the points that they occur, using weights and levers.

Airframe Structural Design

With most of our HPR vehicles, the stiffened composite skin of the fuselage, or „body tube‟
serves as:
 The main aerodynamic surface,
 The main structural element,
 The walls of the various compartments (which if pressurised improve structural
performance).

Because fuselage radii are usually much larger than skin thicknesses, radial stresses within
the skin are unimportant.
Radial loadings associated with geometric and applied loading discontinuities will be ignored
here, so at zero angle of attack, the only remaining forces acting on the fuselage skin are
tangential, tensile, unit loadings caused by compartment internal pressures : n   Pr ,
(where r is the fuselage radius) and also the axial unit loading resulting from the longitudinal
(drag) loads:
D
nA  where the axial drag load „D‟ is assumed negative in compression.
2 r

Rick Newlands & Martin Heywood 16 updated: 12/04/11

 Technical papers

For thin skins and/or slender fuselages, the skin will buckle and the fuselage collapse under
typical axial unit loadings nA, therefore the skin can be stiffened internally by attaching
longitudinal members (stringers) and discretely spaced transverse circular rings (frames). Or,
one can use a sandwich construction as John Coker (see his websites) does.

The normal Bending Moment (see previously) resulting from aerodynamic/wind disturbances
imposes an axial, varying unit loading:
MB
nb  cos  around the airframe in accordance with elementary beam theory.
 r2
The angular coordinate  is measured around the longitudinal axis of symmetry of the
vehicle, and is assumed zero (therefore giving maximum unit loading) when parallel to the
loading plane (the angle of attack) producing the bending.
Because nb and nA are coincident and parallel, they add algebraically, and the airframe is in
tension when nb + nA >0, but, except for some joints, the controlling mode of structural failure
under combined axial and bending loads is axial compressive unit loading at  = 180
degrees or 0 degrees.

The other force associated with aerodynamic/wind disturbances is Normal Shear Force (see
previously) which introduces a uniformly varying load around the airframe cross section of:
VR
nv  sin 
r
The skin resists this shear unit loading.

Yet another loading is the wind pressure, or aerodynamic pressure acting normal to the skin.
If this external pressure is greater than the compartment internal pressure, then all the
strengthening benefits of a pressurised skin are nullified, and the stringers simply behave as
beam columns, which must be stabilised against collapse by increasing their thickness.

Natural bending modes

As well as the motions of the vehicle as a rigid body, the vehicle also bends out of shape
slightly in response to the various loadings.
What is the effect of this flexibility?
Interestingly, as the fuselage is free at both ends (unrestrained), the forces and masses along
the fuselage cause the elastic bending to acquire a definite, predictable, shape or „ natural
mode‟, but the shape is time-dependant: the fuselage will bend in and out of shape at a set
frequency.
And it‟s not just one mode, it‟s actually a combination of several independent natural modes:
each mode vibrates at a certain (successively higher) frequency which may or may not be set
in motion by gusts and turbulence of a similar frequency.

Basically, the bending modes cause additional Bending Moments; terms which have to be
added to the Bending Moment equation given earlier.
Structural analysis or testing will determine the amount of Bending Moment occurring for each
individual mode deflection i , These moments can simply be summed:
M B ( xk , t )   M 'i  i
The M’ terms simply represent scalar coefficients which, when multiplied by the appropriate
deflection, give the Bending Moment caused by that deflection.

Rick Newlands & Martin Heywood 17 updated: 12/04/11

 Technical papers

These bending modes have particular shapes which roughly look like this, where the solid line
is the central axis of the fuselage, and the graduated x-axis is the unbent shape. Each mode
is a half-wavelength higher than the previous one.

And so on. It‟s worth noting that successively higher bending modes have less and less effect
(smaller amplitude), so can be neglected: if analysing a full-size aircraft or large launch
vehicle, attention would be paid to only the first three or four bending modes.
What the above static pictures can‟t show is that these standing waves oscillate: half a cycle
later, each picture gets turned upside-down.

The bending is important to model, because it will cause the ends of the airframe (nose and
tail) to pull different, possibly larger, angles of attack. And as the fuselage vibrates, this
increases the inertial loads. Also, a bent tail will cause the line of the thrust force to miss the
centre of gravity, so a rotation is caused by the thrust, as well as a sideways thrust
component. Ref. 2 says that the effects of bending on the pitch/yaw moments of inertia are
minor and can be ignored, as can the bending effect on CG position.

Needless to say, the bending modes cause merry hell with any inertial sensors (lateral
accelerometers or rate gyros) within the fuselage. Fortunately, the bending mode frequencies
are fixed and predictable (and measurable with a fuselage-shaker) so these frequencies can
be filtered-out of the sensor data.

Historically, matrix methods and an energy analysis would have been used to calculate the
bending mode shapes (Ref. 2), but nowadays a Finite Element Analysis on computer would
be used, where the mass of each element is included: the distributed mass and inertia along
the fuselage must be lumped into discrete point mass elements. It has been found from trial-
and-error (Ref. 2) that for simple one-dimensional (length along the fuselage only) beam
bending models the required number of fuselage mass stations should be approximately ten
times the number corresponding to the highest elastic bending mode to be calculated.

Rick Newlands & Martin Heywood 18 updated: 12/04/11

 Technical papers

For example, if three elastic bending modes are to be calculated, then approximately thirty
mass stations are required to represent adequately the bending dynamics of the third mode.
More mass stations than this won‟t increase the accuracy noticeably.

For our simple rocket vehicle shapes (no strap-on boosters) this could actually be quite a
simple model which could be run on a home PC, although modelling our composite body-
tubes presents additional difficulties because they‟re anisotropic (stronger in certain directions
and weaker in others).

Applicability
But I‟m not sure how all this would apply to the small, rigid, composite airframes of HPR class
vehicles. They‟re deliberately exceedingly stiff, so may well not bend to any degree worth
bothering about except perhaps in the first bending mode. And the natural frequencies of
solid rocket motors are much higher than the first few bending mode frequencies, so don‟t set
them off.
And HPR airframes have coupler joints at the parachute compartments. These joints have a
degree of rotational free play that is difficult to model mathematically. Then there‟s the mass
of the parachute, which is free to move laterally; again difficult to model.

Frankly, if your HPR-sized vehicle is particularly bendy, you can attach a mechanical shaker
to it to investigate its bending modes: suspend the vehicle in a cradle of elastic bands or
bungees at its C.G. and alter the shaking frequency until it resonates with the first mode (you
can see this, so photograph or video the bent shape it against a graduated background).

Propellant slosh loads

We can ignore the inertial loads from the mass of waves sloshing across the nitrous oxide
tank in our HPR hybrids: HPR vehicles are narrow enough that the slosh mass is a tiny
percentage of the vehicle mass.

Component requirements
The following were taken from our NRC competition build requirements (Ref.6), and serve as
simple guidelines. Safety factors of two are typically applied:

Motor mount strength requirements

The motor mount must be designed to take all of the thrust loads, both axial and lateral via
the motor thrust ring. It must be designed to withstand and transmit into the body tube a force
equivalent to twice the maximum motor thrust without permanent deformation. The mount
must also withstand a lateral force in any direction equal to a thrust misalignment of 5° at the
maximum thrust value cantilevered about the vehicle‟s centre of gravity.

Fin longitudinal loading

Each fin must be able to support a suspended load
from its tip equal to twice the fin mass times the
rocket's maximum axial acceleration occurring during
any flight phase.

Fin stiffness
If the fins are not stiff enough, they can flutter and
tear off, therefore fins should be designed for
stiffness rather than strength. This can be achieved
by giving the fin a little depth so that it is hollow with
composite or sheet metal skins. Or use a sandwich
construction. Ref. 8 says that the chances of the fin
fluttering decreases with the cube of the fin

Rick Newlands & Martin Heywood 19 updated: 12/04/11

 Technical papers

thickness-to-chord ratio, and decreases linearly with the fin shear modulus (effective
stiffness). It also decreases with the square of decreasing fin aspect ratio which is why
rocketry fins have such low aspect ratios. Leading-edge sweepback also helps.
As a stiffness test, each fin must withstand a transverse load equal to the rocket's launch
weight when suspended from the fin tip. When subjected to this load, the maximum lateral
deflection measured at the tip must be less than 10° in either direction.

Fin allignment accuracy

According to Rocket Services UK, accuracy of construction, such as fin allignments, is
absolutely vital for survival of the Transonic zone (around Mach 1). Misalligned fins (more
than 2 degrees off the long axis of the vehicle), as well as creating more drag, can create
vehicle angles of attack (and spinning) that could cause excessive Normal forces, breaking
the vehicle.
It is good practice to investigate the effect of small misallignments in the trajectory sims to see
if these use up any loading safety-margins. (Try misalligned nosecones and fins).

Rick Newlands & Martin Heywood 20 updated: 12/04/11

 Technical papers

References
Ref. 1: Sampo Niskanen, “Development of an Open Source model rocket simulation
software” Helsinki University of Technology Faculty of Information and Natural Sciences
Master's thesis, Espoo, 20.5.2009

Ref. 2: Geissler, E. D, AGARDograph 115 “Wind effects on launch vehicles” advisory group
for aerospace research and development (AGARD) Neuilly-Sur-Seine, France, Feb 1970
Downloadable at:
http://ftp.rta.nato.int/public//PubFullText/AGARD/AG/AGARD-AG-
115///AGARD0G115.pdf

Ref. 3: AD 408259 “Wind shear response for missile systems, comparative study and design
procedure” Lockheed missiles and space company, 1963 (downloadable)

Ref. 4: James R. Scoggins “Wind Design Criteria for the Saturn Vehicle” NASA Marshall
space flight center, Aeroballistics division, 1963 (downloadable)

Ref.5: NASA SP-8035 “Wind loads during ascent” 1970 (downloadable)

Ref.6: Aspirespace NRC technical requirements 2000 (based on Les Jeunes en Espace
strength criteria from their Mourmelon launching campaigns.)

Ref.7: Private communication with our good friend Cliff Maidment, aerodynamicist to the
Black Arrow rocket programme.

Ref.8: Dennis J. Martin, “Summary of flutter experiences as a guide to the preliminary design
of lifting surfaces on missiles”, NACA TN 4197, Feb 1958

Rick Newlands & Martin Heywood 21 updated: 12/04/11

 Technical papers

Glossary:
Geometric definitions:

Angle of attack:  (or Angle of Incidence)

This is usually referred to as „alpha‟, and corresponds to the angle between the airflow
direction (usually the Freestream direction) and some vehicle or fin datum.

Aspect ratio AR:

A wing or fin‟s span divided by its mean chord (see above diagram).

Calibers, Calibres:
In rocketry, vehicle dimensions are usually divided by (compared to) the diameter of the
thickest part of the fuselage so that rockets of different size can be compared: this diameter is
therefore one Caliber.

Centre of Gravity, centre of mass (CG):

The point within the vehicle that is the centroid of mass, the balance point.

Centre of Pressure (CP):

The point on the vehicle‟s surface where the average of all the aerodynamic pressure forces
from some component (e.g. nosecone) act.

Dynamic pressure: (q)

All aerodynamic forces scale directly with the kinetic energy term
 being volume-specific mass = air density, and V = flow velocity (airspeed).
This kinetic energy term is called Dynamic Pressure (q), to distinguish it from its Potential
energy counterpart of static pressure (P).

Reference area: (S)

See Drag (equation)

Shear modulus: (G)

In stress analyses, the shear modulus (or modulus of rigidity), is defined as the ratio of shear
stress (applied load) to shear strain (resulting twist), or in other words the higher the shear
modulus, the less the fin material will twist under an applied torsional load.

Rick Newlands & Martin Heywood 22 updated: 12/04/11

 Technical papers

Appendix 1: constructing a synthetic wind profile

Start with data of windshear versus altitude: here‟s the graph for Cape Canaveral (metres per
second per Km):

Note that there are several curves for different scales of distance (y). For example, the 100m
curve was obtained from an instrument (radar) that could discern wind differences at intervals
of 100m altitude. Note that due to fine wind structures the wind-shear values increase the
closer you examine them (smaller scales of distance) which is why the curves aren‟t equal.
If you were using your own windshear data you might only be able to provide one or two
different scales of distance.

Note that the curves are composed of straight line segments, this is deliberate to allow one to
linearly interpolate between altitudes.

Rick Newlands & Martin Heywood 23 updated: 12/04/11

 Technical papers

Next, convert this windshear data into a wind speed change spectrum, corresponding to a
change in altitude (scale of distance), and plotted as a function of altitude. Wind speed
change is found by multiplying the above windshear data by the above scales of distance:

In tabular form, the above graph is:

Altitude Windspeed (metres/sec) per scale of distance (metres)

Km 100m 200m 400m 600m 800m 1000m 2000m 3000m 4000m 5000m
0.1 10.7
0.2 12.4
0.4 15.5
0.6 15.3
0.8 16.1
1 16.9
2 22.9
3 3.9 6.8 10.3 13.2 15.1 17.2 23.5 28.8
4 3.9 6.8 10.3 13.2 15.1 17.4 24.1 29.7 32.5
5 39.5
10 8.9 13.9 21.7 27.3 31.2 35.2 47.8 54.9 58.7 61.4
14 8.9 13.9 21.7 27.3 31.2 35.2 47.8 54.9 58.7 61.4
20 4 7.0 10.3 13.2 15.2 16.8 20.8 22.9 24.0 24.7
35 4 7.0 10.3 13.2 15.2 16.8 20.8 22.9 24.0 24.7
60 9.5 17.6 31.9 43.4 53.1 60.0 78.3 88.0 95.5 99.6
80 9.5 17.6 31.9 43.4 53.1 60.0 78.3 88.0 95.5 99.6

Rick Newlands & Martin Heywood 24 updated: 12/04/11

 Technical papers

Next, we take a scalar wind speed profile such as the 95 percentile for Cape Canaveral (see
main text).

Then we select a target altitude „hp‟ on this graph upon which to superimpose a windshear.
This might be the altitude that our sim says that „max q‟ occurs, which is the maximum
dynamic pressure (see glossary) suffered by the vehicle and would typically occur just at
engine burnout.

Then starting from altitude hp, we subract the above wind speed change spectrum tabular
data. Start at the smallest scale of distance which is 100 metres, for example if hp is 12Km,
then the above data for 12Km and 100m scale of distance is 8.9 metres per second wind
speed change. So plot a point whose „x‟ is Vp minus 8.9 and whose „y‟ is 100m lower than hp.

Then plot the other scales of distance values in the same way, joining the resulting wind-
buildup curve to the origin of the graph by a straight line that is tangent to the curve:

As the vehicle ascends, let the wind speed follow the red arrows shown here.

To reduce the windshears by 15% simply reduce the wind values in the above table by 15%,
keeping the altitude values the same. This increases the gradient of the curve (reducing the
shear) by 15%

Rick Newlands & Martin Heywood 25 updated: 12/04/11

 Technical papers

Appendix 2: numerically modelling turbulence

As real turbulence almost has a Gaussian distribution, the first step is to generate white noise
with a Gaussian distribution. Some programming languages (such as java) already provide
Gaussian white noise, but C++ doesn‟t, so first we‟ll generate this:

// generate Gaussian white noise (i.e. white noise with a Gaussian amplitude distribution)
// see http://www.dspguru.com/dsp/howtos/how-to-generate-white-gaussian-noise
// The Central Limit Theorem states that the sum of N randoms will approach normal
// distribution as N approaches infinity.
// The drawback to this method is that X will be in the range [-N, N],
// instead of (-Infinity, Infinity) and if the calls to rand() are not truly independent,
// then the noise will no longer be white. Jeruchim, et. al., recommend N >=20 for good
// results.

white_noise = 0.0;

for (N = 0; N < 100; N++) // use 100 randoms (N = 100) for more accuracy
{
// generate random number between 0.0 and 1.0
bob = rand() / 32767.0; // rand() gives random integer between zero and 32767
white_noise += bob;
}

/* for uniform randoms between 0 and 1, mean = 0.5 and variance = 1/12 */
/* so adjust white_noise so that mean = 0 and variance = 1 (this is called standard normal) */
white_noise -= N / 2.0; // set mean to 0
white_noise *= sqrt(12.0 / N); // adjust variance to 1

I checked that this code does indeed produce Guassian white noise: it does.
So we now have the Gaussian white noise (a sequence of Gaussian random numbers). We
now pass it through a suitable digital filter, called the Infinite Impulse Response filter (with two
memory terms) to generate a pink spectrum of turbulence:

// pass white gaussian noise through an infinite impulse response filter (see Ref.1 page 57)
// with two memory terms (called poles) to generate required pink noise

// initialisation, put before main loop

static double memory1 = 0.0, memory2 = 0.0;
static int count = 0;

// main loop
alpha = 5.0 / 3.0;
first_pole = -alpha / 2.0;
second_pole = (1.0 - alpha / 2.0) * first_pole / 2.0;

if (count == 0)
pink_noise = white_noise;
else if (count == 1)
pink_noise = white_noise - first_pole * memory1;
else
pink_noise = white_noise - first_pole * memory1 - second_pole * memory2;

if (count < 5)
count++;

Rick Newlands & Martin Heywood 26 updated: 12/04/11

 Technical papers

// save the last two pink_noise values

memory2 = memory1;
memory1 = pink_noise;

// one standard deviation (SD) of 1000 pink noise values output from this software
// (on my computer, check what values yours gives)
// The most recent versions of Microsoft Excel have an SD function.
const double standard_deviation = 2.179639788;

normalised_pink_noise = pink_noise / standard_deviation;

// input turbulence intensity in percent (see main text)

wind_standard_deviation = (turbulence_intensity / 100.0) * average_wind_speed;

// turbulence speed in metres/second

turbulence = normalised_pink_noise * wind_standard_deviation;

Note that Ref.1 says that this software should be looped continuously at 20 Hertz, which is
slower than most rocket trajectory sims loop. In this case, linearly interpolate the turbulence
values with time before feeding them into the sim.

Rick Newlands & Martin Heywood 27 updated: 12/04/11