Variable Speed Transmission using Planetary Gear System for High

Speed Rotorcraft Application

Sylvester V. Ashok
sylvester_ashok@gatech.edu

Brian Wade
brian.wade@gatech.edu

Daniel P. Schrage
daniel.schrage@aerospace.gatech.edu

Georgia Institute of Technology, Atlanta, GA 30308

ABSTRACT
High-speed helicopter flight poses many challenges to designers. Many of these challenges result from the additive
nature of the helicopter speed and the rotor rotation on the helicopter’s main rotor advancing side. These high speeds result
in increased wave drag, impulse noise and vibrations. These disadvantages can be lessened by slowing the main rotor
during high speed flight. This paper presents a transmission design that slows the main rotor by up to 50% during high-
speed forward flight while maintaining high rotational speed during low-speed flight and hover and maintaining engine
efficiency. The transmission design uses a standard planetary gear system with the sun gear as the input and the planetary
carrier as the output. It accomplishes its speed variation by allowing the ring gear to rotate at variable speeds controlled by
a controller and an electric motor. This design was demonstrated on a theoretical coaxial helicopter. This paper also
covers the sizing and synthesis methodology used to perform preliminary design of this transmission.

forward flight performance choosing a tip speed that is

INTRODUCTION  fast enough to allow the helicopter to maneuver at a hover
The main rotor is the most complex part of a with a safety margin before stall yet slow enough to allow
helicopter. In forward flight it poses further challenges the helicopter to reach its required forward speed.
due to differential tangential velocities observed by the
advancing and the retreating sides. This is further
exacerbated at high velocities where the tip speed at the
advancing side is transonic and retreating side is near stall.
Cyclic feathering helps to alleviate the differential lift, but
the maximum speed of the helicopter begins to be limited
by the speed of the advancing rotor blade tip. As the
advancing blade enters the transonic regime, it develops
wave drag that dramatically increases the power required
from the engine. As this wave drag begins to form,
vibrations also increase due to shock induced and dynamic
stall and the resulting shocks greatly increase the in-plane
impulse noise of the blade. A way to delay these effects is
to slow the main rotor rotational speed. Current rotorcraft
designs slow the main rotor through the engine, forcing
the engine to operate off-design and results in degraded
efficiency. This method, however, is contrary to what is
needed in hover and slow flight. In a hover, the rotor does Figure 1. Aerodynamic, noise, and autorotative
not have the advantage of forward speed. As the rotor constraints imposed on selection of rotor tip speed [2]
speed is slowed, the blades must be pitched to higher and
higher angles of attack until they stall. Typically, Figure 1 (above) shows the optimum tip speed
designers seek to compromise between the hover and consideration for a given forward flight speed limit, in
order to expand this operating limit, one has to maintain
the optimum tip speed while achieving higher advance
Presented at the American Helicopter Society 66th Annual Forum,
May 11-13, 2010, Phoenix, AZ.. Copyright © 2010 by the ratios.
American Helicopter Society International, Inc. All rights reserved.
 Understanding the above limitations, it seems
reasonable that a helicopter would be best served by
allowing the rotor tip speed to be changed in flight from
high speed at a hover to low speed in forward flight. This
can be done to an extent by slowing the engine, as done in
present designs. However, this speed change is restricted
by the operating limits of the engine. All modern turbine
engines are designed around a specific operating speed.
Outside this normal design speed, the efficiency falls
drastically to the point where many engines cannot deliver
enough power to fly the helicopter at the proposed 50%
speed reduction of this paper. Another alternative to
changing the speed of the engine is to slow the output
shaft from the main transmission. The two categories of
transmissions that do this are continuously variable
transmissions (CVT) and multispeed transmissions. CVTs
can operate at any speed between a minimum and a
maximum speed while multispeed transmissions operate
at a finite number of speeds. The conventional two-speed
designs are less complex and do not have as great of a
weight penalty as the CVTs, but they posses inherent
power interruptions during speed change and negative
power flow. Sudden interruptions in power can develop
severe structural loads on the transmission and
interruptions in power flow will cause loss in
performance. Two-speed designs can be adaptable to be
quasi-variable through variable transition assist, which
may be either external powered or internally driven. Speed
range changes for CVTs need to be computer controlled,
sensing both drive speed and engine speed. The design
objectives therefore for the transmission is to achieve the
speed while maintaining continuous positive power flow
without a significant weight penalty. Other design criteria
S: Sun - Input
P: Planet – Intermediate gears
Planet Carrier: Output
Ring: Intermediate gear
Controller: Secondary input
Figure 2. System concept and rotation direction
 Ns   Nr 
 pc   s     r   (1)
 Ns  Nr   Ns  Nr 
to be considered are reliability, system and sub-system
feasibility, life-cycle cost, total volume and part count.
When choosing the speed range needed for the slowed
rotor, the rotorcraft performance and the rotors dynamics
and handling qualities need to be considered. Slowing the
rotor too much will lower the autorotative capabilities and
affect the rotors damping. The slowed speed should be
chosen such that the rotor motion’s natural frequencies are
away from the harmonics of rotation (1P, 2P, etc.). This
puts operating constraints on the motor to control the
speed change and on the optimum speed change rate, for
both slowing and returning to normal operation speed. The
motor also needs to constantly power the controller in
order to maintain a speed if the desired slowed rotor speed
is not slowed by exactly 50%.
CONCEPT OVERVIEW
This transmission design consists of a planetary gear
with sun gear as input and planetary carrier as output as
shown in Figure 2 (below). The planet gears behave as
intermediate gears similar to that in a conventional
planetary system. The ring gear is held stationary under
normal operation, and the planets rotate about it. The
normal operation for this design is categorized as hover,
climb and low speed forward flight (below bucket speed).
The planetary carrier angular velocity is given by the
relationship in equation (1)
Where,
 : Angular velocity
N : Number of teeth
Subscript s: sun
r: ring gear
pc: planetary carrier
 The ring gear speed, ωr is negative, if rotating in the
opposite direction to the planetary carrier as shown in
Figure 2 (above). ωr term in Equation (1) will now be
acting to reduce ωpc. The design concept follows the
principle that the planetary carrier angular velocity, for a
fixed gear ratio has a dependency on the sun gear (primary
input) and the ring gear (intermediate gear for secondary
input). The ring gear, shown in Figure 3 (below), is
therefore designed with internal gearing allowing it to
mesh with the planet gears and outer gearing allowing it to
mesh with the controller gear.

Figure 4. Clutch system

A control mechanism is required for this system to

enable the speed change process. It is the most important
requirement in this transmission design. The functions of
the controller are to:
1. Control the speed of the ring gear after clutch
disengagement. Rotor dynamics and power flow
require that the ring gear start rotating slowly and
increase speed at a rate optimized for rotor
performance.
Figure 3. Planetary gear system – detailed view 2. On conclusion of high-speed (flight) operation,
the rotor is required to return to its normal speed.
A planetary gear offers the advantage of high speed The controller is required to slow down the ring
reductions within a small space. It is adaptable to be a and bring it to a safe clutch engagement speed.
CVT with a motorized controller assist for speed
transition. This is the fundamental part of this type of a The controller can be internally driven or externally
CVT. This mechanism should be adaptable to a drive powered. An internal drive mechanism is one that uses
system of most helicopters. Rotors are designed for mechanical power flow from the drive system to control
operation at particular speeds and this system provides the the ring gear, Figure 5(a) (below). An external drive
capability of having a smooth transition between the mechanism is one in which a motor, powered externally is
required multi speeds. used to perform the speed/resistance control, Figure 5(b)
(below).
This design uses a set of clutch bands meant to hold
the ring gear stationary when required and for this purpose
the ring gear is provided with extended portions on either
side on the outer surface. The clutch works primarily as a
holding mechanism. This mechanism is similar to that
used in traditional automobile automatic transmissions. A
small amount of friction for braking is expected during the
slowing down and engagement sequence (this braking
frictional force is yet to be quantified). These clutches are
designed to be power disengaged, meaning the electric
actuators are spring loaded and power has to be supplied
to release them. For safe helicopter operation, the clutch
system remains engaged without power supply to the
actuators. Actuator pistons are inserted within a spring.
When the actuators receive electric signal, they compress
retracting the bands, this causes the clutch bands to release (a)
the ring gear, which is then free to rotate. System is
designed with multiple clutch bands and actuators for high
reliability and better frictional force distribution as shown
in Figure 4 (below).

(b)
Figure 5. (a) Internally driven control mechanism;
(b) Externally driven control mechanism [1]
Each method of control mechanism has its advantages
and disadvantages. The internal driven mechanism has the
advantage of power regeneration, i.e. the power required
by the controller to let the ring gear start rotating at a slow
rate, when the rotorcraft is entering high speed mode, is
Figure 6. Peregrine – high speed coaxial rotorcraft concept
Transmission Layout
The transmission system for the Peregrine is split into
sections A to E and the gears divided as G1- G12 as
shown in the transmission layout in Figure 7 (below). The
rotor drive system for this entry featured two standard
bled off the drive system directly by the use of the variator
mechanism. Once the operating speed is reached, the ring
gear acts as an output of the planetary system, the variator
can be designed to take advantage of this by regenerating
that power back to the drive train or to an auxiliary thrust
system. The challenges with this system however, is that
variator involves complex shafting and gearing, which
increase the weight and the volume of the system. The
externally driven system is simple, with sufficient gear
ratios for high torque, a sizeable motor can be adopted to
power the system. This will however have a weight and
volume penalty of the motor itself and to provide adequate
gearing for torque generation.
IMPLEMENTATION OF DESIGN
This transmission design was created as part of the
Georgia Institute of Technology’s graduate level entry
into the 2009 AHS Student Design Competition sponsored
by Agusta-Westland to design a unique rotor drive system.
The entry, named the Peregrine, won the first place and is
shown in Figure 6 (below).
turboshaft engines, which feed into a mixing spur
assembly (system A in Figure 7). The power was then
sent forward into the unique dual speed planetary system
(system B in Figure 7) and rearward through bevel gears
systems (systems D and E in Figure 7) to a pusher
propeller. The output from the planetary system was then
fed into a differential system for reduction and speed
matching of the counter rotating rotors (system C in
Figure 7). This is done with the help of a hollow shaft
which powers the lower coaxial rotor from the upper bevel
gear and an inner shaft that powers the upper rotor from
Figure 7. Peregrine transmission layout
Peregrine Performance
Performance analysis was done on the peregrine
helicopter to study the effects of transition velocity on
forward flight performance of the helicopter. Transition
velocity was chosen as 100 knots, with a good cushion on
(a)
the lower bevel gear. This system has no mechanical
dissymmetry in power supply to both rotors i.e. both
rotors operate at the same rpm at all operating conditions
and the same torque when no differential collective is
applied.
performance at various operating conditions. This system
produced a helicopter capable of 222 knot cruise and 250
knot dash speeds as seen in the Figures 8(a) and (b)
(below). Standard sea level conditions did not strike as a
big challenge for hover power requirements, as seen in
Figure 8(a).
Max dash speed 250 knots at IRP
Max cruise speed 222 knots at MCP
 (b)
Figure 8. Required horsepower and available horsepower vs. velocity curve for
(a) sea level and (b) high - hot day – 5000ft, 95°F
The solid red line shows the actual helicopter power
curve including the rotor speed reduction. The blue dash
line shows the power for the helicopter at a low speed
setting only and the yellow dashed line shows the
performance at high speed only. The value of the rotor tip
speed reduction is shown from maximum speed at both
max continuous power (MCP) and at intermediate rated
power (IRP). At sea level standard conditions, the dash
speed moved from 210 knots to 250 knots. This 40 knot
increase was due only to the rotor speed change that took
place in the transmission with the multispeed planetary
system. The operation of the planetary system is
expressed in Table 1 (below).
Table 1. Peregrine transmission settings
Mode Tip Sped Input Output Ring Gear Speed
High Speed 650 ft/s Sun Gear Planet Carrier 0 – Locked
Low Speed 420 ft/s Sun Gear Planet Carrier 391 RPM
Rotor Dynamics
A fan plot for this design was generated using
DYMORE to predict the dynamic characteristics of the
rotor. DYMORE was selected to estimate structural
deformation and dynamic response such as displacements,
strain, forces, and velocity. DYMORE is a finite element
based tool for the analysis of nonlinear flexible multibody
systems. To predict the motion of the rotor, DYMORE
needs the data of aerodynamic forces, structural
properties, rotor motion, and kinematical constraints. This
analysis provides the distance between upper and lower
rotor systems. In addition, the natural frequency of blade
system can be analyzed by the eigenvalues. Since each
rotor system has four blades, 1P, 3P, 4P and 5P should be
avoided and design emphasis was placed on reducing
potential harmonics in this region. In the case of 1P, it
does not cause any problem at all, because the first lead
lag mode is well above 1P (stiff in plane) for removal of
ground and air resonance instabilities. However, Peregrine
is a rotorcraft with variable rotor speed main rotor.
During transition from a rotor speed of 650 ft/s to 420 ft/s,
the initial design had the second flapping mode and the
first lead-lag mode pass through the regime of 4P.
Furthermore, the frequencies of second flapping and first
lead-lag mode were very similar in the low speed mode.
This can cause coupling vibration. Both of these
situations can result in vibrations and instabilities in the
rotor system. Therefore, the blade was stiffened by
adjusting the position and shape of spar. Figure 9 (below)
shows the fan plot for the improved blade. In this design,
the frequencies of second flapping and first lead-lag mode
are always higher than 4P. In addition, the coupling
vibration was also solved by separating the frequencies
during all normal speed regimes.
 Figure 9. Fan plot for Peregrine helicopter

Sizing and Synthesis
A genetic algorithm was used to size and optimize the
gear system to yield the lightest weight system that was
able to withstand all gear stresses of the system. All gears
were initially selected to be helical type gears with 25 0 full
tooth loading due to their gradual loading, quieter running
noise, and capability for high speed, high load operations.
Even with the helical and loading angles selected the
transmission design outlined above still had 25
independent variables. These variables are listed in Table
2 (below). The gear ratio and diametrical pitch (Pd) apply
to the gear set (A-E) and the material selection is done
independently for each of the 12 gears.

Table 2. Transmission design variables

Pd = diametric pitch, number of teeth per inch, for each gear Number of Variables
Pd=[1, 1.25,1.5,1.75,2,2.5,3,4,5,6,8,10,12,14,16,18] 5
Gear Ratios
Engine Gear Box: 1 – 3 2
System A: 1-10 1
System B: 1-10 1
System C: 1-4 1
System D: 0.2 – 5 1
System E: 0.2-5 1
Number of planets
System B: [2,3,4] 1
Material for each gear
Material = [AISI 9310, VASCO X2M, Pyro-wear 53] 12

These parameters were selected based on knowledge
of other gear designs. The diametrical pitch (Pd) set listed
is the most commonly manufactured Pd’s for course gears
that are used in aircraft transmissions. The gear ratio
ranges were selected based on the type of gears used in the
system. Spur gears typically can withstand ratios up to
10:1 and bevel gear systems can withstand ratios up to
1:5. System C was limited to 1:4 due to the high loads it
would see since it was the last stage of reduction to the
rotor. The three material selections are common gear
materials used in the main transmission. All three
materials are steel alloys that are case-hardened and
carburized. Their material properties are shown in Table
3 (below).
For sizing, the following design criteria were selected:
 Engine operated at 3100 HP (3442 Contingency
Power), 23,000RPM (plus a 20% factor of
safety).
 Rotor tip speed: 680 ft/s in High Mode, 420 ft/s
in Low Mode.
  Transmission component life: 3,500 flight hours
Table 3. Gear material properties [5]
Even with these constraints and discritizing the
continuous variables to the hundredths there were over
4.598 (10)34 possible combinations. A computer that
could analyze 1,000 of these combinations a second would
take 1.458 (10)24 years to analyze all the possible
combinations (full factorial analysis). A faster
optimization method was needed. In many typical designs
an initial starting point is determined based on engineering
judgment and the parameters are optimized from that
point. The difficulty with this method is when dealing
with complex and new technologies as with the speed
controller proposed by this paper. Often adequate design
knowledge does not exist for such emergent technologies.
When optimizing from any initial point the optimization
process typically settles into the nearest local optimum
which may, or in many cases, may not be the global
optimum. Therefore, in order to optimize all these
variables to create the lightest weight system that would
meet the stress requirements of the new system, a genetic
algorithm was used. A genetic algorithm (discussed in
detail in appendix section) is a stochastic search method
that is specifically designed for discrete variables and
because it is stochastic it has an increased probability of
finding a global optimum in the presence of many local
optimums. This was very fitting for the transmission
Table 4. Transmission sizing genetic algorithm parameters
General Parameters
Population Convergence Variable Type Probability of Cross Over
Size Criteria Continuous 100
60 15 generations Discrete 100
Constraints
Gear Constraint
5 – Ring Gear Max Diameter = 22 in
7+8 – Rotor Bevel Gears Max Diameter = 30 in
6 – Input to rotor bevel gears Max Diameter = 10 in
sizing since, given the number of variables there was a
very high likelihood of many local optimums and 18 of
the 23 variables were discrete by their very nature (Pd of
each system, material of each gear, and number of planets
in system b.)
Transmission Sizing
A sizing program was written for the new transmission
in Matlab. This sizing program sized each gear mesh
system (labeled A – E in Figure 7). The equations and
gear factors used followed the recommend practices of the
American Gear Manufactures Association (AGMA) as
outlined in [6]. The program’s general design was object
oriented in nature. Each gear system (labeled A – E in
Figure 7) was sized in its own program and the parameters
passed to the main program. In order to ensure the rotor
system and propeller each spun at the required rpm the
gear ratio of systems B and E were calculated given the
input gear ratio of the rest of the gears. This left 23
independent variables. Within each gear mesh sizing
program, the general flow was to step through the number
of teeth of the pinion (up to a max of 200 teeth), calculate
the number of teeth of the gear, and then for each tooth
set, increased the face width until the bending and contact
stress were less than the allowable stress. The program
then estimated the weight for that gear set. This was done
for every tooth combination and the lowest weight gear’s
parameters (teeth of the pinion and gear, face width,
diameters of each gear, and stresses of each gear) were
passed back to the main program which then added up the
weight of the entire system. Thus, the sizing program was
able to optimize the numbered teeth, diameter, and face
width of each gear for the lowest overall weight given the
23 inputs. The genetic algorithm was run as specified in
the above section using Model Center. Model Center is a
graphical user environment to perform integration and
automation of codes, optimization and for experimental
design. Model Center was used to automate the genetic
algorithm and the sizing program. A multiple elitist
strategy with tournament selection helped to ensure the
appropriate amount of exploration while speeding up
convergence to the optimum. A population of 60 was
chosen with 15 consecutive generations with no change
for the top member as the convergence criteria. The other
model parameters for the actual transmission genetic
algorithm are summarized in Table 4 (below).
Probability of Mutation
0.1
0.05
Reason
Engine air intake obstruction
Fuselage width – Drag reduction
Fuselage height – Drag reduction
 Another benefit to using the multiple elitist strategy is also helps to speed up the convergence of the program and
that the cross over probability can be 100%. This means helps the program to be exploitive versus exploratory in
that the bottom 2/3 of the designs is guaranteed to change order to find the best optimum while ignoring other local
but the best designs are preserved in the top 1/3 elite. This optimums.

Figure 10. Genetic algorithm optimization convergence history

The results of the genetic algorithm produced an
optimum gear set for the above parameters. The program
took 113 generations to converge and built approximately
6.24 million transmissions in the process. The
convergence history is shown in Figure 10 (above) for the
final sizing run. The parameters for this gear set are
shown below in tables 5 and 6 (below). As can be seen
from these parameters the constraints in gears 6, 7, and 8
are active (shown in yellow) and gears systems A and C
were sized primarily on contact stress while the rest were
sized based on bending stress (shown in yellow). Also,
note that there is an inherent 20% factor of safety in all
actual stress estimates which was built into the sizing
program.

Table 5. Optimized Gear Parameters

System Parameters Gear Parameters

System Mach Pd Weight No. of Face Weigh
System ID Adv (1/in) (lbs) planets Gear D (in) Teeth Width (in) Material t(lb)
Mixing 1 2.625 21 Pyro-Wear 53 3.04
Spur
Gears A 3.9048 8 32.33 2 10.25 82 6.8 Pyro-Wear 53 29.52
North 3 4.75 38 VASCO X2M 3.33
Planetary 4 5 40 AISI 9310 5.09
System B 4.1053 8 37.29 4 5 14.75 118 2.8 VASCO X2M 13.77
Bevel 6 10 30 VASCO X2M 16.81
Gear 7 30 90 VASCO X2M 60.65
System C 3 3 154.93 8 30 90 3.9 VASCO X2M 60.65
South 9 9.3 93 VASCO X2M 11.45
Bevel 1 D 1.0968 10 25.15 10 10.2 102 2.7 VASCO X2M 13.7
South 11 9 90 Pyro-Wear 53 12.34
Bevel 2 E 1.3222 10 33.536 12 11.9 119 3.1 VASCO X2M 21.2
Eng Acc Eng
Gear Box Eng 1.628 Acc VASCO X2M
 Table 6. Gear Bending and Contact Stress

System Parameters Gear Parameters Contact Stress (psi) Bending Stress (psi)
System System ID Gear RPM Actual Allowable Actual Allowable
1 14128 198,940 200,230 42,592 43,621
Mixing Spur Gears A 2 3634.6 100,190 206,580 38,183 44,688
3 3634.6 197,490 200,100 43,418 43,599
4 3452.9 192,490 203,550 42,964 44,181
North Planetary System B 5 0 63,599 206,700 40,077 44,709
6 885.36 211,110 211,180 39,083 41,344
7 295.57 211,110 213,150 39,083 41,639
Bevel Gear System C 8 295.57 211,110 213,150 39,083 41,639
9 3634.6 198,240 207,720 40,572 41,815
South Bevel 1 D 10 3304.6 198,240 208,170 40,572 40,885
11 3304.6 207,700 209,510 40,489 40,885
South Bevel 2 E 12 2500 207,700 209,510 40,489 41,088

Computer Aided Design model resizing of the transmission to evaluate constraints on

individual gear systems dimensions, based on rotorcraft’s
A parametric computer aided design (CAD) model for structure for each iteration. The final design is shown in
this design was created in CATIA which enabled rapid Figure 11 (below).

Figure 11. Final transmission design (CAD model)

The CAD model enabled some initial finite element calculations. Figure 12 (below) shows an image of stress
analysis (using SIMULIA) to verify theoretical stress analysis performed on the planetary gear system.

Figure 12. Finite Element Analysis on planetary gear system (planet and sun gears shown)
 CONTROLLER ARCHITECTURE
A Power Electronics Module (PEM) was designed to
control the variable speed planetary transmission. This
system is shown in Figure 13 (below). The Pilot has
control over the selection between high and low speed
transitions in this initial concept. At hover and low speed
operation, the PEM is set for high rotor rpm, meaning the
ring gear clutch bands are engaged. For high-speed flight,
the PEM is set for low rotor rpm. The PEM is made up of
a Supervisory Controller (SC), Motor Controller (MC)
and Clutch Controller (CC). SC monitors controller gear
speed and rotor rpm (from the speed governor). The
governor works as a primary input to both the engine
control system (FADEC – Full Authority Digital Engine
Control) and the PEM. The speed switch should have an
input to the FADEC system as well to ensure that the
engine does not try to reset the speed of the rotor. SC
monitors motor torque, to enable accurate inputs to the
MC. The MC powers the motor and supplies precise
voltages to ultimately keep the reference rpm stable. The
CC controls the engagement and disengagement of the
clutches. This system should have a provision for manual
override from the pilot to be able to engage the clutches
for flight safety purposes. This is indicated by the dashed
line in Figure 13. The PEM should also receive an input
from the vehicle electrical system.
Figure 13. Power Electronics Module architecture
High rpm to low rpm transition PEM sequence:
1. Disengage clutch. Power disengagement involves
retracting actuators simultaneously.
2. Power motor to maintain controller and ring gear
somewhere between 0 rpm and an upper limit. The
upper limit depends on the dimensions of the
system and the sensitivity that this upper limit
might have on the rotors performance.
3. Increase controller rpm to slow planetary carrier
according to a predetermined optimum speed
change rate (approximately 5-10 rpm per second
main rotor equivalent)
4. The PEM will now seek to keep the rotor rpm
stable by controlling the motor torque and speed.
Low rpm to high rpm transition PEM sequence:
1. Slow the ring gear at a predetermined optimum
slowing rate (approximately 5-10 rpm per second
main rotor equivalent) by powering the motor.
2. Once ring gear has been brought to a slow speed of
somewhere between 0 rpm and an upper limit, the
clutches can be engaged. This upper limit depends
on the amount of frictional force that can be
developed by the clutches without considerable
wear.
3. Power to the clutch actuators is discontinued and
the clutches engage. The ring gear is now held
stationary and the drive system returns to normal
operating mode
High RPM: 650 fps main rotor tip speed
Low RPM: 420 fps main rotor tip speed
 CONCLUSION
High speed flight has always been one of the
greatest challenges in the Rotorcraft industry and
there is an increased interest in maximizing forward
speed without much compromise on weight and hover
performance. This paper offered a promising design
solution to slow the main rotor using the
transmission. The new design incorporates the well
understood planetary reduction system but achieves
its speed reduction by allowing the ring gear to rotate
from stationary to a predetermined speed. Through a
controller motor and gear, the ring gear’s speed can
be maintained, resulting in up to 50% reduction in
rotor tip speed while allowing the engines to operate
at their design range. This transmission design was
adopted and tested on a theoretical helicopter for the
AHS student design competition. The design
transmission was optimized with a genetic algorithm
to ensure the lightest weight system. The resulting
rotorcraft achieved a top speed of 250 knots. This
was 40 knots improvement over the single speed
maximum. This speed increase was mostly
attributable to the delay in drag divergence of the
advancing blade through the speed reduction. This
helicopter retained the low speed maneuverability and
hover capability of a conventional rotorcraft and was
free of instabilities through its flight regime.
Additional work is needed:
 To accurately quantify the torque
required for the controller mechanism to
produce the desired control.
 To explore other control mechanisms
such as traction based internally driven
system.
 Perform controller simulations to test and
optimize the Power Electronics Module
control system.
 Test the entire design in a virtual
environment before it could move on to
actual physical tests.
APPENDIX
Genetic Algorithm Overview
A genetic algorithm is based off the theory of
evolution. Due to its random nature, it cannot
guarantee an optimum but has an increased chance of
finding a global optimum in the presence of many
local optimums as would be expected in this
transmission optimization. The program builds a
population of transmissions then selects the ―most fit‖
transmissions based on a predetermined criteria or
overall evaluation criteria function. In this case the
overall criteria were weight with all stress
requirements being meet. The fittest parents are then
―mated‖ so that their combination produces a new
population of transmissions that are better. This
continues until there is no improvement in several
successive generations. To increase the search
ability, several offspring are also allowed to mutate or
differ from either parent. Each design is represented
by a binary string. At the beginning of a genetic
algorithm, the design space is discritized into discrete
steps. Due to the nature of the binary string, the
genetic algorithm works only on discrete values.
Continuous variables must be discritized into small
steps. All variables are then converted into binary
strings and those strings are then assembled end to
end to form one long binary sting that represents that
design. Initially, a population of points is randomly
generated. The overall string length is determined by
converting all variables and there ranges into binary
strings and piecing these end to end as specified
above. A population of points, typically 50-100 or so,
are then randomly generated by randomly selecting 0
or 1 to fill each location in the sting. The stings are
then converted back into decimal numbers and these
stings and decimal equivalents are then passed onto
the selection stage. There are many ways to determine
which designs are promoted into the parent
population to mate and create the next child
population. The method used in this paper was a
multiple elitist tournament selection method. This
type of method helps the program converge faster
which was useful in this paper since the sizing code,
written in Matlab, took about 30-60 seconds a run to
size each transmission for a given set of parameters.
For this method, two strings were randomly selected
and their translated decimal equivalents were run
through the sizing program. Their overall function
values (transmission weight in this case) were then
compared. If the overall function values were among
the top 1/3 of all compared thus far they were
automatically promoted to the parent population.
This is the multiple elitist part of the selection
method. If they were not part of the top 1/3 then their
two function values were compared and the one with
the best function value (lowest weight) was then
promoted to the parent population. This was
continued until the parent population was the same
size as the pervious population. The multiple elitist
part of the strategy automatically promoted the top
third of the designs into the parent population. This
ensured the best designs were used in the parent
population and helped to make the optimization more
exploitive and converge faster. Other methods, such
as the roulette method, do not guarantee that all of the
best designs are promoted. The main disadvantage of
this type of method is that it tends to be exploitive
rather than explorative. When there is a greater
chance for lower ranking designs to be used in the
parent population the algorithm naturally bounces
around the design space more but converges slower.
This disadvantage was offset by the tournament
selection part of the method. Two designs were
randomly selected, their parameters run, and there
function values compared. In other methods, such as
again the roulette method, the chance of a point being
selected is directly proportional to its objective
function value. This type of algorithm will quickly
zero in on an optimum which, if done too quickly
without adequate exploration, can many times be a
local rather than global optimum. This method also
has the added advantage of speed in that is a point is
randomly chosen more than once, the algorithm can
recognize this and not run the sizing program for it
again. If this occurs, then all points in the population
will not be fed into the sizing program increasing
speed but sacrificing exploration. Once the parent
population is chosen they are mated to create an
offspring child generation. The parent population is
grouped into sets of two randomly. Crossover takes
place with a certain specified probability. If chosen,
that set is mated and undergoes a two-point cross
over. In this cross over method, the size of the bit to
be crossed over is randomly selected. Thus, there is a
chance that if the two numbers were sequential then
there was a uniform cross over (all bits swapped).
This is illustrated below for two points A and B. In
this example Strsize is the total number of bits in the
string.
Two Point Cross Over Example
Once this new generation is created from the
parents the individual members are tested to see if
they will undergo a random mutation. This mutation
would swap a bit in a single position. For each
member of this new population a random number was
chosen between 0 and 1. If this number was less than
the probability of mutation another random number
was chosen, scaled, and rounded as above to the
length of the string. The bit corresponding to this
position was then changed from a zero to a one or
vice-versa. Once this mutation is completed for all
members, that population is then re-designated as the
new parent population. The process of crossover and
mutation is now performed on the new population.
This is repeated until there is no change in the
objective function of the best design in a child
population for a predetermined number of
generations. Once this occurs the optimization is
considered to have converged.
ACKNOWLEDGEMENTS
The authors of this paper would like to thank all
members of the Georgia Tech graduate design team
for the 2009 American Helicopter Society Annual
design completion. The design was truly a team
effort.
REFERENCES
[1] Stevens, Mark A., Handschuh, Robert F.,
Lewicki, David G., ―Concepts for
Variable/Multi-Speed Rotorcraft Drive
System‖, September 2008, NASA/TM—2008-
215276, ARL–TR–4564, AHS 080273.
[2] Leishman, Gordon J., Principles of Helicopter
Aerodynamics, 2nd Edition, Cambridge
University Press, New York, 2006.
[3] Stevens, Mark A., Handschuh, Robert F.,
Lewicki, David G., ―Variable/Multispeed
Rotorcraft
Drive System Concepts‖, March 2009,
NASA/TM—2009-215456, ARL–TR–4758.
[4] Litt, J.S., Edwards, J.M., DeCastro, J.A., "A
Sequential Shifting Algorithm for Variable
Rotor Speed Control," AHS 63rd Forum,
Virginia Beach, VA, May 1-3, 2007,
NASA/TM-2007-214842.
[5] Bellocchio, Andrew T. ―Drive Systems Design
Methodology for a Single Main Rotor Heavy
Lift Helicopter‖, Georgia Institute of
Technology, Dec 2005.
[6] Norton, Robert L. Machine Design: An
Integrated Approach. Prentice-Hall Inc, 1998.
[7] Handschuh, Robert F., Zakrajsek , James J.,
―Current Research Activities in Drive System
Technology in Support of the NASA
Rotorcraft Program‖, January 2006,
NASA/TM—2006-214052, ARL–TR–3707.
[8] Vanderplaats, Garret N., Numerical
Optimization Techniques for Engineering
Design, 4th Edition.