Professional Documents
Culture Documents
Proposed for the vibration testing of free flying systems is the Marionette paradigm, a passive gravity com-
pensation scheme of unique mechanical simplicity that does not corrupt performance. Tribological (damping)
and stiffness interference with specimen dynamics are virtually eliminated. Inertial pollution, well controlled
in all cases, is minimal for most practical problems. Moderate kinematics are easily accommodated and few
limitations are imposed on suspension system design. Tolerance against specimen and support imperfections is
naturally high. The concept can be generalized to also accommodate certain adaptive model geometries; scope
may thus be extended to the vibration testing of evolving configurations and to some deployment problems.
The simulation of certain inertial loading conditions in weightlessness — for example, steady state acceleration
— is also possible.
An introduction to the Marionette paradigm — theory, features, and practicalities — is offered. The need
for and the potential of system optimization are discussed and illustrated. The arguments are supported with
demonstration hardware and numerical results.
Nomenclature
I. Introduction
AUGHT between budgetary constraints and a demand for ever higher performance, space structures engineers
C increasingly often turn to solutions that would have been considered too risky in the past. Low weight, flexibility,
and size — for antennas, reflectors, solar sails, etc. — are often cited as characteristic to this trend.
∗ Research Associate, Center for Aerospace Structures, Dep. of Aerospace Engineering Sciences, USB 429, Member AIAA.
† Chief Engineer for Space, Structures and Materials Competency, 4 West Taylor Str./MS 230, Associate Fellow AIAA.
1 of 19
American
Copyright © 2005 by Greschik. Published by the American Institute Institute
of Aeronautics andofAstronautics,
Aeronautics Inc.,
and Astronautics
with permission.
The creative engineering so fostered makes the hardware verification of both concepts and performance a very high
priority — a priority that directly translates into a critical need for ground testing with a fidelity and quality to match
the sensitivity of the subject system. This sensitivity is, indeed, a critical issue as it implies vulnerability to gravity
effects, as well as to the imperfections both of its own construction and of the testing apparatus.
In fact, it was the challenge posed by the sensitivity of certain solar sail booms due to their architecture, size,
and compliance that motivated the writers of this paper to seek innovative means to simulate weightlessness. The
elimination of nondeterministic effects (random slip, friction, damping) was considered imperative. Deterministic
effects such as added inertia that can be precisely quantified were of secondary concern as long as imperfection
sensitivity was little affected, because one primary goal of testing was to benchmark numerical models (rather than to
actually mimic behavior in true weightlessness).
Additionally, a focus on vibration testing was also defined. This focus, however, does not imply a lack of interest
in the complementary support problem — gravity offloading during large kinematics (e.g., specimen deployment).
Rather, modal testing was prioritized simply as a natural consequence of the intended use of test results to benchmark
numerical models. Gravity compensation throughout the deployment sequence, still a desirable capability, was thus
forced to take a back seat.
This research led to the development of the Marionette paradigm wherein the weight of each part of the specimen
is balanced by the other parts, with only optimized fly beams and cords contributing to mass overhead. Damping,
friction, and slip are eliminated, and specimen response is not affected by deleterious stiffness either. The support
architecture is simple, with high tolerance against both specimen and support system imperfections. Kinematics can
naturally involve up to moderate displacements and deformations in both the vertical and horizontal directions, and
certain types of expansion (e.g., deployment) in the horizontal direction can also be accommodated with minor system
modifications. This unique combination of flawless performance, exceptional fault tolerance, and utmost mechanical
simplicity are an attractive alternative to “classic” gravity compensation schemes for many problems.
A. Gravity Compensation
Gravity is ubiquitous in everyday life on the global mechanical scale, rendering the human experience of weight-
lessness and the direct study of this condition impossible on Earth. Accordingly, ground tests to explore the zero-g
performance of mechanical systems inevitably employ various compromise solutions. In fact, the variety of how such
compromises can be negotiated is rich; it can be discussed from several different perspectives.
One way to classify schemes to eliminate gravity effects is in terms of the means employed. For example, one can
distinguish1, 2 physical, buoyancy, pneumatic (air cushion and bearing), and mechanical methods. (The first category,
“physical approach,” includes parabolic flight and drop tests — not really gravity compensation but, in fact, orbital
tests with Earth-colliding orbits.)
A broader categorization can be based on the dimensionality of the specimen motion accommodated. 3 Accordingly,
one can distinguish support systems to permit linear, planar, or spatial kinematics.
One can also rank gravity offloading options based on support structural complexity. The simplest route is to avoid
offloading at all and downscale the hardware4 until, according to constant thickness5 or classic proportional6 scaling
principles, gravity effects reduce to tolerable levels. This approach is practical if sensitivity and quantitative fidelity
are not critical. The next level of complexity, applicable to some sufficiently stiff systems, is specimen alignment
with the gravity field to eliminate direct gravitational bending at the cost of geometric nonlinear effects. The latter
must be accounted for during test evaluation. Further raising complexity, a horizontal specimen orientation free from
nonlinear effects may be used with simplistic distributed lateral support, realized with cord-and-pulley mechanisms,
counter-weights, soft (constant force) springs or “bungee” cords, pneumatic or buoyant bearing, etc. Such simple
support, however, generally affects and sometimes severely restricts specimen response. To overcome this final hurdle,
autonomous force control can be employed at each supported location. Depending on the sensitivity and kinematic
scope of control, the simulation of weightlessness can asymptotically approach perfection and/or can permit large
kinematics in many cases. Hybrid active–passive control systems that combine robustness with high fidelity may offer
the best performance.7, 8
One may observe that the cost of quality is the sophistication of support. It turns out, however, that nearly flawless
performance can also be achieved for many vibration problems with the simple scheme herein proposed: a single fly
beam hierarchy.
2 of 19
general hierarchy
of fly beams: S
rank 3
rank 2
rank 1
forces S1 S2 S3 ...
exerted:
symmetric
full hierarchy: S
rank 3
rank 2
rank 1
forces Sc Sc Sc Sc ...
exerted:
Figure 2: Load tree; whiffle tree. Figure 3: Solar array deployment test setup, Ref.1
Fly beams are typically used in specimen support to simplify the apparatus by reducing the number of external
support locations, as in case of the solar panel deployment test by Fischer, 1 Fig. 3, An example with multiple fly beam
hierarchies is the PowerSail deployment experiment,10 Figs. 4 (a) and (b).
Fly beams have also been used to directly aid gravity balancing with extra weights, cf. the recent testing at NASA
Langley R.C. of the L’Garde Encounter/ISP solar sail boom,11 Fig. 5. In this role, the beams serve as “frictionless
pulleys” to reduce nondeterministic effects and increase the test apparatus sensitivity.
3 of 19
4 of 19
so the model
"feels" no
weight.
Figure 7: Each force exerted at the base of the load tree Figure 8: A. Calder: Mobile, 1963 (Healey Library, Univ.
suspends one part of the whole. of Massachusetts, Boston).
The specimen “feels” no gravity to the extent of how fine the Si force distribution is. This scenario is similar
to a kinetic mobile in the artistic sense of Gabo, Moholy-Nagy, and Calder, Figs. 7 and 8 — except the suspended
elements are here connected and thus don’t float apart, unlike the components in kinetic art. The floating motion in
kinetic displays, however, nicely highlights the similarity between the kinematic constraint enforced by a load tree and
weightlessness.
These relations are independent of the beam rotation α. Figure 9: Fly beam geometry and rotation α.
Next observe that, if the beam tip vertical displacements h1 , h2
(Fig. 9) directly reflect those of the masses (their centers of gravity) hung therefrom, then the dz cg displacement in the
z direction of the combined (resultant) c.g. of both suspended subsystems is
h2 m 2 − h 1 m 1
dzcg = dzF + (6)
m1 + m 2
5 of 19
h2 m 2 = h 1 m1 (7)
dzcg = dzF (8)
This result reveals that fly beam pivoting does not affect the [z height of the combined center of gravity of the suspended
masses. This implies that fly beam support at the fulcrum, dzF = 0, constrains the height of the suspended subsystems’
resultant center of gravity. Via the repetitive application of this principle through the hierarchy, the constraint exercised
by the load tree is seen to keep the specimen center of gravity stationary.
Thus the constraint exercised by the load tree simply maintains (zero) momentum in the z direction. This condition
is equivalent to weightlessness, since it is the principle of the presentation of momentum that inevitably constrains a
closed mechanical system in free flight.
Note also that the above analysis relies only on general fly beam properties. The constraint exercised by the
suspension, therefore, does not depend on the load tree geometry and topology. Specimen response, therefore, should
be little affected by the choice of any particular load tree design, as long as the latter involves a single hierarchy with
external support only at the top fulcrum of the top beam.
6 of 19
7 of 19
(a) Mode 7: f=1.655 Hz. (b) Mode 8: f=2.412 Hz. (c) Mode 9: f=2.983 Hz.
Figure 13: Plate suspended with load tree in gravity field, dominant free vibration modes. (Modes 8 and 9 are identical
to modes 2 and 3 of the free flying model, Fig. 12 (b) and (c), here shown with the displacement fields negated.)
Each fly beam arm has been meshed with three BAR (i.e., uniform beam) elements (resulting in a total of 6 elements
per beam), as has been each cord section between subsequent hierarchy ranks. (Accordingly, cords spanning several
ranks had three time as many elements as the number of ranks involved.) The full mesh thus consisted of 400 plate
and 327 bar elements.
As seen from the vibration frequencies in Figs. 12 and 13, the suspension, indeed, well simulates weightlessness:
the suspended system frequencies are 1–2 % off the free modes. Surprisingly, this error has been calculated positive:
the suspended frequencies are somewhat greater than the true free ones. One would expect the opposite, since the sup-
port system increases system mass — this should lower, not raise, frequencies. It is possible that numerical problems
play a role in this discrepancy; the modal analyses of the suspended system have been plagued by excessive iteration
errors.
In addition to the faithful simulation of weightlessness, load tree suspension also adds pendulum and internal
compliance modes to the system that should not be allowed to interfere with the tested modes of interest. For the plate
model, mode separation has been sufficient for the five pendulum modes, three of which are shown in Fig. 14. (Modes
3 and 5, not shown, are swing and tilt modes like 2 and 4 — Fig. 14 (b) and (c) — in the complementary direction.)
However, the dominant support compliance mode — vertical system vibration, Fig. 15 (a), with f=1.441 Hz — is close
(a) Mode 1: yaw, f=0.0078 Hz. (b) Mode 2: tilt, f=0.184 Hz. (c) Mode 4: swing, f=0.394 Hz.
Figure 14: Plate suspended with load tree in gravity field, pendulum modes of vibration.
8 of 19
ez
ex
G. Alternative Interpretations
The mechanical essence of the Marionette paradigm, formally investigated in the previous section, can be intuitively
grasped in several ways. For example, one can see the concept as:
• A load tree as a support mechanism.
• A mechanical calculator, geometrically programmed to produce the resultant of the combined specimen-support
weight for output at the external support.
• A means to balance specimen weight with itself; defeating gravity effects with the help of gravity.
• Kinetic art with the pieces interconnected.
9 of 19
III. Practicalities
HE systematic introduction of the Marionette concept in Section II has been
T structured with a taste for didactic rigor. In contrast, the illuminating details
and practical concerns presented next are hoped to deepen the reader’s under-
standing of the paradigm with a panoramic view of critical details. Much of the
material discussed is based on observations of demonstration hardware models.
A. Horizontal Dynamics
A fly beam can rotate virtually unconstrained both vertically about its fulcrum
and horizontally, twisting its support cord. Consequently, a load tree can ac- Figure 18: Space experience.
commodate lateral specimen kinematics as well, as long as the latter is perpen-
dicular, rather than parallel, to the current directions of the involved fly beams.
An ideal example is the straight beam shown deformed in Fig. 10; where all fly beams are aligned with the specimen
and, accordingly, all (small to moderate) lateral specimen motion is perpendicular to the fly beams. Global pendulum
effects do not arise because the specimen center of gravity is remains exactly under the top support no matter how the
specimen may deform; otherwise. horizontal momentum would not be preserved. Local pendulum effects within the
hierarchy do not arise either because the fly beams can rotate and move horizontally to maintain equilibrium with all
cords vertically aligned (again, for small to moderate lateral motion).
However, specimen and suspension dynamics may interact laterally in unfavorable ways. Further, minor second
order stiffness and geometric nonlinear effects (cord rotational stiffness and vertical slope effects) that would otherwise
be negligible may still interact in a manner to effect some pendulum stiffness. Therefore, attention is required if
horizontal specimen motion is (also) anticipated: the compatibility of lateral specimen kinematics with fly beam
alignment, suspension system dynamics in the horizontal direction, and the interaction of cord torsional and pendulum
effects should be carefully evaluated.
10 of 19
Figure 19: Frictionless flexure in thin elastic tongue. Figure 20: Deleterious constraint by multiple load trees.
For friction, however, one may argue that the motion between cords and fly beams at their joints this still involves
friction, albeit of the benign rolling nature. This rolling friction can be eliminated (reduced with orders of magnitude)
if flexure joints are used by which, in essence, inelastic (i.e., viscous) flexure elements, the cord ends, are replaced
with elastics ones, thin flexible tongues, Fig. 19.
M = Kα α (10)
Kα = eS (11)
with S denoting the fulcrum load (the combined weight of fly beam and suspended subsystems, Eq. 1).
For a bean in the suspending hierarchy, however, the effect of a fulcrum offset is slightly different, because the
beam orientations are locked by (coupled with) the suspended subsystems (ultimately, the specimen). If the latter is in
a preferred orientation (e.g., it is horizontal) than the fly beams are horizontal, too. In the context of this constraint, the
effects of the offset components ex and ez (cf. Fig. 16) separate: the horizontal offset ex alters the force distribution by
changing the beam arms (r1 , r2 ), while the vertical one ez introduces the geometric nonlinear stiffness Kα via Eq. 11.
The perturbation of the force distribution does not affect system stiffness; it only alters the specimen stress-strain
response to gravity (in the context of the support) and, as seen in Fig. 15 (b), it may cause the entire system to tilt. This
tilt can be easily corrected by the fine horizontal adjustment of the top beam fulcrum alone, with no similar adjustment
required in any other beam.
On the other hand, the Kα stiffnesses through the hierarchy emerge from the vertical ez offsets only. This stiffness
for a particular beam is positive (of restoring, stabilizing nature) if the fulcrum is above the ideal location, e z > 0 (cf.
Fig. 16). However, Kα is negative (of destabilizing nature) otherwise, making the beam “want” to flip over. If the
offset error distribution is not systematic, these stiffness effects tend to cancel out across the system. However, should
the reduction of this error be deemed necessary, it can be achieved by the vertical adjustment alone of the fulcrum
locations. In many cases, this adjustment can be guided simply visually and be performed in the full test setup.
The vertical fulcrum offset in the top beam, on the other hand, is of critical importance. This offset must not
become negative, lest the entire suspension system destabilizes (tilts until further motion is arrested externally or by
the collision of system components). Robustness, therefore, calls for a positive e z > 0 for this beam. This finite ez
11 of 19
(a) Support with load tree. (b) Load tree as force distribution device.
Figure 21: Alternative test setups for a cantilever system.
also generates a soft restoring stiffness against global specimen rigid body motion, which guarantees global system
stability but little affects specimen response.
12 of 19
use of the pantograph mechanism, Fig. 22, because its internal kinematics maintains fly beam arm proportions and,
therefore, does not interfere with the overall load tree design.
To eliminate scissor friction, control to drive each pantograph could be introduced. While horizontal transient
effects may still be beyond the scope of the system, the vibration testing at several phases of quasi-static deployment,
as well as the quasi-static deployment process itself, would be possible.
V. Seeking Perfection
UE to its mechanical simplicity and robust fault tolerance, apparatus design and fabrication for Marionette sus-
D pension can be carried out with simple and approximate means. However, the optimization of these steps for best
performance (for minimal inertial interference with specimen response) is very difficult.
A. System Optimization
In order to develop a practicable method to minimize via the judicious engineering of the suspension system the inertial
pollution of specimen response, a generic measure of inertial interference — a merit function — would be required.
Next, this measure would need to be embedded in a preferably rapid and transparent yet sufficiently rigorous procedure
to assess the relative performance of preliminary design alternatives without much effort. This objective function and
procedure should of course recognize that inertial interference is influenced collectively by (a) the distributions of
specimen mass and of the anchor points, (b) the topology and geometry of the beam hierarchy, (c) the fly beam design,
13 of 19
1. General Considerations
For the testing of stiff heavy specimens such as a small aircraft, probably not much effort needs to be invested in fly
beam design. It may be sufficient to simply cut pieces of lumber into appropriate lengths, and to set arm ratios λ (Eg. 2)
to reflect with “construction precision” the desired equilibrium conditions — provided that the beams do not fail or
deform excessively. Most space structures and lighter aircraft, however, will require more sophisticated engineering.
The task is twofold: minimize inertial effects and eliminate second order stiffness associated with fulcrum offsets.
I NERTIAL E FFECTS . Mass moment of inertia and mass must be distinguished. Minimum mass design will not
achieve minimal moment of inertia because it does not penalize mass with the distance from the axis of rotation; the
beam of least mass moment of inertia will be slenderer than that of lowest mass. The relative importance of mass and
rotational inertias should be specified for beam design, likely in light of the beam location in the hierarchy and the
specimen motion of interest.
G EOMETRIC N ONLINEAR S TIFFNESS . As discussed above, secondary stiffness against fly beam tilt is effected
by nonzero fulcrum offset from the ideal location, and it can be eliminated with vertical fulcrum adjustment after
fabrication. The associated challenge is that of engineering precision: the precision of design and fabrication, and that
of calibration. When responding to these challenges, one should be aware of the following issues.
Fly Beam Mass. Fly beam mass and center of gravity depends on the beam design itself. Except for the simplest
situations, therefore, the design is iterative.
The Line Between the Beam Tips. The ideal fulcrum location on a uniform beam will be on the line between the
beam tips (which happens also to be beam centerline). For a more complex design of variable cross section or one
assembled of various members, this may no longer be true. In the general context, the fulcrum should be at the c.g. of
the system defined by the beam and the masses suspended from its tips, placed at the tips.
Beam Deformations. Mass optimization eliminates unnecessary material and, as a result, reduces stiffness. A
mass optimized fly beam may deform considerably under the load it supports. The fulcrum location should be appro-
priately placed in this, deformed, state.
Tip and Fulcrum Definition. Careful analysis is required to determine the equivalent geometric locations of beam
tips and fulcrum if non-traditional joints (cf. Fig. 19) are used.
14 of 19
3. Design Specifications
The specifications for material and technological details have been defined somewhat optimistically, reflecting, first,
the writers’ intent to explore the limits of optimization and, second, the fact that the beams are likely to be fabricated
in well controlled conditions with precision tooling. Further, environmental and imperfection effects (which influence
the conservativeness of specifications in many indirect ways) will also be limited since the beams will be used in
laboratory conditions.
Cords and Tendons. Unidirectional fibers with little resin are assumed, E=15 Msi, ρ=1500 kg/m 3 . Tensile
strength is Ru =350 ksi. The minimum cord and tendon diameters are dmin,cord =0.5 mil and dmin,tendon =2 mil,
respectively — the actual values can continuously variable from these limits up, as needed.
Joints. Joints integrate flexural stiffness continuity as well as cord grip as needed, with technological details
unspecified. The mid-keel joint mass is expressed as a function of the fly beam span m keel =L×0.5 g/m. The joint at
the keel end, which grips the heavier tendons and the subsystem suspension cords, is 3×, while the joints at the cross
bar ends are 3× heavier than the mid-keel joint. Other details are ignored.
15 of 19
Keel Segments and Cross Bar. The standard strut failure modes of local (wall buckling) and global (Euler) stabil-
ity have been considered with the equations presented in Ref. 12 Cross bar length has not been allowed to drop below
5% of
The Entire Keel. Lateral loss of global stability considered, approximately assessed with an pequivalent simply
supported beam on elastic foundation. The assumption of end pin support reflects the high tendon stiffness. The
equivalent foundation stiffness has been calculated as the average n lateral stiffness provided by the cords. Beam
compression is the average of segment compressive forces.
Cords and Tendons. Checked against tensile failure and tendons required to have diameters at least 8× those of
the cords, in order to guarantee the applicability of keel global stability analysis. As tensile failure turned out not to
be an issue, tendons generally ended up with diameters of d tendon =4 mil, in order to achieve 8× the minimum cord
diameter dcord =dmin,cord =0.5 mil.
Factors of safety. All failure criteria were required to be satisfied with factors of safety of at least two, FS≥2.
L span = 46 cm
S total load = 0.205 N
φ force ratio = 1.0 φ force ratio = 4.0
m mass = 1.080 g m mass = 1.078 g
keel segments:
t = 2 mils
d = 12.2-13 mils
16 of 19
m+ =
1 1
= mf b /msusp
10
1
.1
.01 0.1 0.1
.001 0.01 0.1 1 10 100 1000 0.01 0.1 1 10 100 1000
.0001 total load, S [N] total load, S [N]
.00001
force ratio φ = 4.0 span, L [m] force ratio φ = 8.0 span, L [m]
10 10
1 1
0.1 0.1
0.01 0.1 1 10 100 1000 0.01 0.1 1 10 100 1000
total load, S [N] total load, S [N]
Figure 26: Direct mass overhead m+ = mf b /msusp of optimized fly beams, single keel bipod construction.
17 of 19
VII. Acknowledgments
HIS work has been supported by NASA Langley Research Center, award number NAG-1-03003. The modal
T analysis of the plate model for numerical concept verification has been performed with NE/NASTRAN, with a
license owned and generously made available for this analysis by L’Garde, Inc. The optimization engine used in the
design optimization study is the freely available COBYLA software package. All other tools and software, including
the FE package integrated with COBYLA and the software package for the nonlinear design of compressed struts
involved in the parametric design study, have been developed by the first author of this paper.
References
1 Fischer, A. and Pellegrino, S., “Interaction Between Gravity Compensation Suspension System and Deployable Structure,” Journal of Space-
craft and Rockets, Vol. 37, No. 1, January–February 2000, pp. 93–99.
2 Fischer, A., Gravity Compensation of Deployable Space Structures, Ph.D. thesis, Pembroke College, the University of Cambridge, Cam-
Flying Robots,” 55th International Astronautical Congress of the International Astronautical Federation, the International Academy of Astronautics,
and the International Institute of Space Law, International Astronautical Federation, 3-5, Rue Mario-Nikis, 75015 Paris, France, IAF, Vancouver,
Canada, October 4–8 2004, IAC-04-J.5.08.
4 Salama, M., White, C., and Leland, R., “Ground Demonstration Of A Spinning Solar Sail Deployment Concept,” The 42nd AIAA/ASME/-
ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference and AIAA/ASME Adaptive Structures Forum, Seattle, WA, April 16–19
2001, AIAA 01-1266.
5 Greschik, G., Mikulas, M. M., and Freeland, R. E., “The nodal concept of deployment and the scale model testing of its application to a
University Press, 40 West 20th Street, New York, NY 10011-4221, December 1996.
7 Straube, T. and Peterson, L. D., “A large motion suspension system for simulation of orbital deployment,” The 35th AIAA/ASME/ASCE/-
AHS/ASC Structures, Structural Dynamics, and Materials Conference and AIAA/ASME Adaptive Structures Forum, Vol. 4, Hilton Head, SC, April
18–20 1994, pp. 1959–1969, AIAA Paper 94-1568-CP.
8 Kienholz, D. A., “Simulation Of The Zero-Gravity Environment For Dynamic Testing Of Structures,” The 19th Space Simulation Conference,
nar, Surface Structures For Space Applications,” The 44th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference
and AIAA/ASME Adaptive Structures Forum, Norfolk, VA, April 7–10 2003, AIAA 2003-1828.
11 Lichodziejewski, D., Derbes, W., Reinert, R., Belvin, K., Slade, K., and Mann, T., “Development and Ground Testing of a Compactly Stowed
Scalable Inflatably Deployed Solar Sail,” The 45th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference and
AIAA/ASME Adaptive Structures Forum, Technical Papers, Palm Springs, CA, April 19–22 2004, AIAA-2004-1507.
12 Anon., “Buckling of Thin-Walled Circular Cylinders,” NASA Space Vehicle Design Criteria (Structures) NASA SP-8007, NASA Langley
18 of 19
19 of 19