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

net/publication/298792453

Design and Test of a Small Reaction Wheel for a Remote Sensing

Microsatellite

Conference Paper · October 1997

CITATIONS READS

4 2,124

2 authors:

Massimiliano Pastena Michele Grassi

European Space Agency University of Naples Federico II
88 PUBLICATIONS 292 CITATIONS 150 PUBLICATIONS 1,370 CITATIONS

SEE PROFILE SEE PROFILE

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

ESA EO Future Mission Technologies View project

vision‐aided applications for Autonomous uav flight in Complex Environments View project

All content following this page was uploaded by Massimiliano Pastena on 23 March 2016.

The user has requested enhancement of the downloaded file.

 DESIGN AND TEST OF A SMALL REACTION WHEEL FOR A REMOTE SENSING
MICROSATELLITE
Massimiliano Pastena* and Michele Grassi**
*Dipartimento di Ingegneria Aerospaziale Seconda Università di Napoli
**Dipartimento di Scienza e Ingegneria dello Spazio Università degli Studi di Napoli "Federico II"
Abstract
This paper deals with the design and test of a reaction wheel for a remote sensing microsatellite.
The wheel control torque and angular momentum are numerically evaluated using a code that
simulates the microsatellite orbital and attitude dynamics and control. The wheel design is
performed using a mathematical model based on the minimisation of the system total mass and
power consumption. The design tolerances are also determined studying the effect of the wheel
dynamic loss of balance on the microsatellite attitude dynamics. A mathematical model of the DC
motor behaviour is adopted to derive a control law of the supply voltage as function of the required
control torque. Laboratory tests are finally performed to evaluate experimentally the wheel control
electronics transconductance and the rotor inertia moment.
Nomenclature
b motor mass figure of merit R motor winding resistance
c power supply figure of merit tc wheel cylindrical tolerance
h wheel angular momentum tp wheel perpendicular tolerance
I motor current
I2 microsatellite 2-axis inertia moment U motor supply voltage
x system state vector
kh housing mass figure of merit
SBD dynamic loss of balance
k2 microsatellite 2-axis inertia ratio β microsatellite pitch angle
K m motor torque constant η motor efficiency
J wheel inertia moment θ rotor polar coordinate
L motor inductance φ phase shift
mh housing mass ξr pitch damping factor
mm motor mass τm open loop system time constant
m ps power supply mass τ cl closed loop system time constant
mt total mass ωr pitch eigenfrequency
mw wheel mass ωo orbital angular velocity
m A linear mass 123 inertia principal axes
superscripts:
M c control torque
. derivative
M d dynamic loss of balance torque subscripts:
M m motor torque i input value
M H motor stop torque o output value
n motor angular velocity p project
P motor power consumption max maximum value
Ps motor shaft power 0 no-load value
r wheel radius
*Real Casa dell'Annunziata, via Roma 29, 81031 Aversa (CE), Italy
**p.le Tecchio 80, 80125 Napoli, Italy
Introduction
Microsatellites are a flexible tool to perform scientific and technological research in space.
Nevertheless, obvious limitations in size, weight, on board available power and costs impose
several constraints on the design of the microsatellite sub-systems. The attitude control sub-system,
as one of the more complex sub-systems of a satellite, is particularly affected by these constraints,
especially when the considered applications require a stringent attitude control. Therefore, new
solutions in terms of components and operating logic need to be investigated to reduce costs,
volumes and power requirements.
In this context, at the University of Naples, a low-weight, low-power reaction wheel has been
developed. The wheel, which will be used to control the attitude dynamics of a microsatellite for
remote sensing applications [1], has a total mass less than 1 kg (housing, DC motor and control
electronics included) and an angular momentum capacity of 0.1 Nms.
This paper deals with the design of the reaction wheel. To this end, the wheel on-orbit requirements,
in terms of maximum reaction torque and angular momentum capacity, are first determined using a
numerical code that simulates the microsatellite orbital and attitude dynamics and control. A three-
axis attitude control is considered based on the use of three reaction wheels operated by brushless
DC motors. The wheel design is then performed using a mathematical model in which the system
total mass is minimised. The design tolerances are also determined using a simplified analytical
model that relates the rotor dynamic loss of balance to the attitude dynamics. A control law of the
motor supply voltage as a function of the required torque is finally derived. To this end laboratory
tests are performed to evaluate experimentally the wheel control electronics transconductance and
the wheel inertia moment.
1.Wheel design: mathematical model
The reaction wheel design is performed by a minimisation of the system total mass (power supply
and wheel assembly) in order to realise the best compromise between the wheel mass and power
consumption [2]. The system total mass is expressed as follows:
mt = m ps + mm + mw + mh (1)

Neglecting the contribution of the motor winding inductance, the power supply mass can be
expressed as a function of the power consumption as follows:
2
P 1 ⎛⎜ M m ⎞
m ps = = ⎜ 2 R + M m n ⎟⎟ (2)
c c⎝ K m ⎠
where the first term between brackets is the power dissipated by the motor windings, while the
second one is the motor shaft mechanical power. Considering that the motor parameter R K 2 m is
proportional through a figure of merit b to the motor mass[2] and that the motor angular velocity
can be expressed in terms of the wheel angular momentum, radius and mass, equation (2) becomes:
M m ⎛ b2 h ⎞
m ps = ⎜⎜ 2 ⋅ M m + 2 ⎟ (3)
c ⎝m m r ⋅ mw ⎟⎠
Equation (3) shows that, for given Mm and h , the motor power consumption is reduced as the
wheel radius increases. On the contrary, the wheel assembly dimensions, mass and working stress
increase with the wheel radius. Since for a given working stress the housing mass is proportional to
the third power of the wheel radius[2], the system total mass can be expressed as follows:
Mm ⎛ b2 h ⎞
mt = ⎜ M m + ⎟ + mm + m w + k h r 3 (4)
c ⎜ m2m r mw ⎟⎠
2
⎝
The wheel radius and mass that minimise the system total mass are given by the following
expressions:
 1 1
4
⎛ h⋅ Mm ⎞ 8⎛ 2 ⎞
r =⎜ ⎟ ⎜⎜ ⎟⎟ (5.a)
⎝ c ⎠ ⎝ 3k h ⎠
1 3
⎛ 3⎞ 4 1 ⎛ hM ⎞ 8
mw = ⎜ ⎟ kh 4 ⎜ m
⎟ (5.b)
⎝ 2⎠ ⎝ c ⎠

Figures 1a and 1b show the wheel radius and mass as a function of the required control torque and
⎛W⎞ ⎛ kg ⎞ ⎛ kg W ⎞
angular momentum (the values 5.2 ⎜⎜ ⎟⎟ , 481 ⎜ 3 ⎟ and 5.23⎜⎜ ⎟ have bee taken for the
⎟
⎝ kg ⎠ ⎝m ⎠ ⎝ Nm ⎠
parameters c , k h and b , respectively [2]). In particular, the curves in the figures correspond to
constant values of wheel radius (range from 0.01m to 0.2m) and mass (range from 0.25kg to 5 kg).
Wheel radius (m) Wheel mass (kg)
3 3
10 10

0.2

0.5
Angular Momentum (Nms)

Angular Momentum (Nms)

2 2
10 10

1 1
10 10

0.01 0.25
0 0
10 10
-3 -2 -1 0 -3 -2 -1 0
10 10 10 10 10 10 10 10
Control Torque (Nm) Control Torque (Nm)
Figure 1a Figure 1b

2. Wheel design: numerical model

The wheel required control torque and angular momentum have been evaluated by a numerical
simulation of the microsatellite attitude dynamics and control. The numerical code also simulates
the microsatellite orbital dynamics considering the main environmental perturbations [3]. For the
microsatellite attitude simulation the disturbing torques due to the aerodynamic drag and the gravity
gradient have been considered.
The microsatellite three-axis attitude control has been simulated. It is performed using three
reaction wheels aligned with their axes along the microsatellite geometrical axes ( xyz ). The i-th
wheel control torque has been modelled using a Proportional Derivative (PD) control law as
follows:
M c = −Gi x (6)
where Gi is the i-th row of the gain matrix, evaluated with the optimum control theory[4,5]
Figure 2 shows the wheels angular momenta at the beginning of the attitude control and during ten
orbits. Since the microsatellite inertia principal axes do not coincide with the geometrical axes [1]
the reaction wheel angular momenta slowly increase along the orbit. A maximum angular
momentum capacity of 0.1 Nms (which is reached in about 50 orbits) has been assigned to the
reaction wheels. The numerical analysis also shows that a maximum reaction torque of 8⋅ 10-4 Nm is
required for the attitude control. Table 1 shows the characteristics of the reaction wheel that satisfies
the control torque and angular momentum requirements.
 Figure 2 Wheel angular momenta

wheel mass (gr) 80

wheel radius (cm) 5
maximum shaft power (W) 0.5
maximum angular velocity (rpm) 5300
Table 1 Wheel main characteristics
3. DC motor selection
Power Efficiency Current 4
The DC motor dynamic and 4
x 10
1
electrical behaviour can be 30
Power

described with the following 70

10 3.5
0.9
equations [4]: 0.8
dI (t ) 25 60 eff.
3 Current
U (t ) = RI (t ) + L + K m n(t )
dt (7)
8 0.7
50
dn(t ) 20 2.5
J = K m [I (t ) − I 0 ] = M c 0.6
dt
n (rpm)

40 6
Watt

In order to select a suitable DC 2 0.5

%

15
motor its steady state behaviour 30 0.4
1.5
has been first analysed. By some 4 n
10
mathematics, the motor main 20
0.3
1
parameters can be related to the 0.2
2
motor torque. In particular, 5
10 0.5
equations (8) state the motor 0.1

supply current, angular velocity, 0 0 0 0 0

shaft mechanical power and 0 20 40 60 80 100
Torque (mNm)
efficiency for a given supply
voltage.
Figura 3 DC motor steady-state behaviour
Figure 3 shows the motor main
parameters versus the shaft torque
for a brushless DC motor with the characteristics listed in table 2 [6].
 Mm
I=
Km
U − RI U R
n= = − 2 Mm
Km Km Km
(8)
⎛ R ⎞ 1
Ps = M c n = (M m − I 0 K m )⎜⎜U − M m ⎟⎟ ⋅
⎝ Km ⎠ Km

η=
Ps
=
(M m − I 0 K m ) ⎛⎜U − R M ⎞⎟
⎜ m⎟
U ⋅I U ⋅Mm ⎝ Km ⎠
Since the selected DC motor satisfies the following conditions:
( )
Ps max U p > Ps p
n0 (U p )
np > (9)
2
M H (U p )
M cp max < .
2
it is able to give the required performances with a high efficiency.
nominal tension (V) 48
winding resistance (Ohm) 8.4
maximum shaft power (W) 33
maximum efficiency (%) 77
no load angular velocity (rpm) 22500
no load current (A) 0.088
stop torque (mNm) 115
torque constant (mNm/A) 20.2
winding inductance (µH) 7.6
Table 2 DC motor main parameters
4. Wheel control logic
With reference to equations (7), the transfer function of the open loop system U(s) ↔ M c (s) is
given by the following expression:
Km
s
H (s ) = L (10)
2 R K m2
s + s+
L JL
Figure 4 shows the system response to a step supply voltage. It can be observed that the control
torque rapidly approaches its maximum value ( τ m = 0.1ms ); then it exponentially decreases as the
angular velocity tends to its no load value. This behaviour is confirmed from the analysis of Fig. 3:
for a constant supply voltage the working point on the diagram rapidly moves from the origin to the
point ( M m ≅ M H , n ≅ 0 ). As a consequence, the angular velocity increases and the working point
moves to the point ( n = n 0 , M m = 0 ) along the line n(M m ) . Therefore, in order to maintain the
required control torque between two successive commands, the wheel control electronics must
apply a supply voltage increasing with the angular velocity.
 -6 -6
x 10 x 10
2.5 2.5

2 2

1.5 1.5
Mc (Nm)

Mc (Nm)
1 1

0.5 0.5

0 0
0 1000 2000 3000 4000 0 0.5 1
Time (s) Time (s) -3
x 10

Figure 4 Open loop system step response

A control law for the supply voltage as a function of the control torque is derived from the first of
equations (7) neglecting the contribution of the winding inductance and substituting for the angular
velocity the expression obtained by integrating the second of equations (7):
R
[M ci (t ) + M R ] + K m n(t ) = R [M ci (t ) + M R ] + K m 0M co (τ )dτ
t
U (t ) = ∫ (11)
Km Km J

With reference to the block diagram in Fig. 5, the closed loop system transfer function and time
constant can be derived from equations (7) and (11) as follows:
R
s
M co (s ) L
=
M ci (s ) R
s2 + s (12)
L
L
τ cl = = 0.9µs
R
The wheel control electronics consists of a Figure 5 Closed loop block diagram
servo-amplifier, which drives the appropriate current in the motor windings, and of Hall sensors for
the angular velocity measurement.

5. Wheel design tolerances

Due to the design tolerances and/or a mounting inaccuracy the wheel central axis does not coincide
with the DC motor rotation axis. The wheel static and dynamic loss of balance can be modelled
considering a balanced rotor with an overbalancing mass, m, placed at a distance r from the wheel
rotation axis and at a distance d from the plane perpendicular to the rotation axis and containing the
wheel centre of mass. Due to the dynamic loss of balance, the reaction wheel will produce on the
microsatellite a disturbing torque, rotating with the wheel, whose modulus is given by the following
expression [7]:
Md = (m r d ) n 2 = (SBD)n 2 (13)

In order to evaluate the effect of the dynamic loss of balance a simplified analytical model relating
SBD to the microsatellite attitude dynamics has been developed. Assuming small angles and
circular keplerian orbit, the pitch attitude dynamics is given by the following equation:
 h! SBD 2
( )
β!! + 3ω 02 k 2 β +
I2
=
I2
[ ]
n1 cos (n1t ) + n32 cos (n3t ) (14)

The system response is given by:

⎡⎛ 2 ⎞ ⎛ 2 ⎞⎤
⎜ ⎛ ω2 ⎞ 2 ⎟ ⎜ ⎛ 2 ⎞ 2 ⎟⎥
SBD ⎢⎜ ⎜ r ⎟ 2 ωr ⎟ + ⎜ cos (n t + φ ) ⎜ ωr ⎟ 2 ωr ⎟⎥ (15)
β(t ) = ⎢ cos (n1t + φ1 ) ⎜⎜ 2 − 1⎟⎟ + 4ξ r 3 3 ⎜⎜ 2 − 1⎟⎟ + 4ξ r
I 2 ⎢⎜ n n 2 ⎟ ⎜ n n 32 ⎟⎥
⎢⎣⎜⎝ ⎝ 1 ⎠ 1 ⎟ ⎜ ⎝ 3 ⎠ ⎟⎥
⎠ ⎝ ⎠⎦
where the first term has a frequency equal to the roll wheel angular velocity, while the second one
has a frequency equal to the yaw wheel angular velocity. Figure 6 shows the results of the analysis.
In particular, it is worth noting that the pitch angle amplitude is maximum along the line
⎛ SBD ⎞
n 1 = n 3 and equal to 2⎜⎜ ⎟⎟ . On the contrary, when n1 n3 = K , with K integer, sudden reductions
⎝ I2 ⎠
of the amplitude indicate that the two signals in equation (15) are not in phase. For our application,
a dynamic loss of balance of 40 gr cm 2 has been chosen which produces a maximum pitch error of
3 ⋅10 −4 deg .

Figure 6 Pitch error versus n1 and n3

In order to realise the proposed dynamic loss of balance, it must be correlated to the wheel design

Figure 7a Rotor tolerances definition Figure 7b SBD evaluation

tolerances. In particular, SBD can be expressed as a function of the wheel cylindrical and
perpendicular tolerances. Figure 7a shows the rotor configuration that gives the maximum SBD for
given t c and t p . With reference to the scheme in Fig.7b, SBD is evaluated considering only the areas
that produce loss of balance (areas 1, 2, 3 and 4). The dynamic loss of balance is therefore given by
the following expression:
π π
2 2
[( ) ( )] [( ) (
SBD = 2 ∫ m A r 2 d 2 − m A r 2 d 1 sinθdθ + 2 ∫ m A r 2 d 3 + m A r 2 d ) ]sinθdθ
4
(16)
0 0
Figure 8 shows a photograph of the wheel assembly. Finally, Figure 9 shows the dynamic loss of
balance versus t c and t p . As expected, the cylindrical tolerance gives the most significant
contribution to SBD . The value of the design tolerances corresponding to the selected dynamic loss
of balance is equal to 0.01mm

Figure 8 Wheel assembly Figure 9 SBD versus design tolerances

6. Wheel test
The laboratory test main goal is to determine the wheel control electronics transconductance, whose
knowledge is necessary to generate appropriate voltage commands. The test facility consists of a PC
equipped with a data acquisition card. A software has been developed to control the wheel tests.
1 The voltage commands consist of a
rectangular pulse whose pick value ranges
Command Voltage (V)

0.8

0.6
from 120 mV to 2780 mV. For each value
of the voltage command ( Vc ) the wheel
0.4
angular velocity is determined performing
0.2
the FFT (Fast Fourier Transform) of the
0
0 20 40 60 80 100 120 Hall sensor signals. In addition, the DC
Time(s)
motor supply current is measured by a
100 tester. Figure 10 shows the wheel angular
Revolution per second

80 velocity behaviour for a 1V voltage

60 command. Figures 11 and 12 show the
40 results of the experimental determination
20 of the motor torque and control
0
electronics transconductance. The
0 20 40 60 80 100 120 knowledge of this curves allows the
Time(s)
appropriate voltage command to be
determined for a given control torque.
Figure 10 Voltage command and wheel angular velocity
In order to generate the appropriate
command voltage on the microsatellite on-
board computer a least square fit of the experimental data in Fig. 12 has been performed using a
third order polynomial for M m < 5.8mNm and a first order polynomial for M m > 5.8mNm . Figure
13 shows the per cent error of the least square fit.
2 0.02

0.018
1.8
0.016

1.6 0.014
Transconductance (Ohm)

Motor Torque (Nm)

0.012
1.4
0.01
1.2
0.008

1 0.006

0.004
0.8
0.002

0.6 0
0 1 2 3 0 1 2 3
Command Voltage (V) Command Voltage (V)

Figure 11 Figure 12
An experimental determination of the wheel inertia moment has been finally performed. Since the
tests have been conducted in a laboratory environment, the equation of motion has the following
form:
dn 1
(
= M m − I 0 K m − Cn n − Cn2 n 2
dt J
) for Vc ≠ 0
(16)
dn 1
dt
= − I 0 K m + Cn n + Cn2 n 2
J
( ) for Vc = 0

where the two coefficients Cn and C n 2 also take into account the effects of the presence of the

atmosphere. Estimates of the terms

(M m − I 0 K m ) and −
I0Km
are then obtained as the constant
J J
dn
(n) for each command voltage.
coefficients of the second order polynomials fitting the functions
dt
The distribution of the J measurements versus the command voltage is shown in Fig.14. The mean
value and the standard deviation of the distribution are 2.32658 *10 -4 kgm 2 and
1.41692 *10 -5 kgm2 , respectively.

Conclusions
In this paper the design and test of a reaction wheel for a remote sensing microsatellite has been
presented. The wheel design has been performed by a minimisation of the system total mass (power
supply and wheel assembly) and taking for the required control torque and angular momentum the
values obtained with a numerical simulation of the microsatellite attitude dynamics and control.
A mathematical model has been developed to relate the wheel design tolerances to the rotor
dynamic loss of balance and, therefore, to the attitude dynamics. The model shows that a value of
0.01 mm must be taken for the cylindrical and perpendicular tolerances to keep the attitude error
within 3 ⋅10−4 deg.
Finally, in order to determine the voltage command corresponding to the required control torque the
wheel control electronics transconductance and inertia moment have been measured by laboratory
tests.
 -4
x 10
4.5 2.9

4 2.8

3.5 2.7

Rotor Inertia Moment (kg*m2)

3 2.6

2.5 2.5
Error (%)

2 2.4

1.5 2.3

1 2.2

0.5 2.1

0 2
0 0.005 0.01 0.015 0.02 0 1 2 3
Control Torque (Nm) Command Voltage (V)

Figure 13 Figure 14

Acknowledgements
This work was supported by the Agenzia Spaziale Italiana and the Ministero dell'Università e della
Ricerca Scientifica e Tecnologica
References
1. Grassi, M.,and Pastena, M., Design of a Multi-mission Micro-satellite, Space Technology,Vol.
16, No. 4, 1996, pp. 197-213
2. Michaelis T. "Small satellite reaction wheel optimisation", Proceedings of the 2nd Annual
AIAA/USU Conf. on Small Satellites, Logan (Utah), Sept. 18-21,1988
3. A. Moccia, S. Vetrella, and M. Grassi, "Attitude Dynamics and Control of a Vertical
Interferometric Radar Tethered Altimeter", Journal of Guidance, Control and Dynamics,
Vol.16, No.2, March-April 1993, pp.264-
4. B. Friedland, Control System Design: An Introduction to State-Space Methods, McGraw-Hill
Publishing Company, New York, 1986, pp. 337-347.
5. J.R. Wertz (Ed.), Spacecraft Attitude Determination and Control, D.Reidel Publishing
Company, Boston, 1978, pp. 204-205
6. Minimotor SA catalogue (1996)
7. Guido A. R., Della Pietra L., Lezioni di Meccanica delle Macchine. CUEN, Napoli, pp.431-442
(1991)

View publication stats