You are on page 1of 9

Smart Mater. Swct. 2 (1993)240-248.

Printed in me UK

I Shape adjustment of precision truss


1 structures: analytical and experimental
1 validation*
M Salama, J Umland, R Bruno and J Garba
Jet Propulsion Laboratocy, California institute of Technology, Pasadena, CA 91109,
USA
Received 3 July 1993,in final form 30 September 1993

Abstract. The use of a limited number of actuatorisensor pairs to alter the shape
or behavior of a truss structure to other desired states is explored analytically
and experimentally. Feasibility of the concept is established by numerical
simulation. As a demonstration, a sequence of validation tests is performed and
the correlation with analysis is discussed. An existing full scale, space-erectabie
high precision truss structure is used for this purpose. For the most part, the
test results agreed well with the analysis. However, micron-level nonlineanties
were discovered in the truss behavior. The significanceof these nonlinearities
in precision structures and their impact on the basic premise of adaptivity is
discussed.

1. Introduction of an adaptive structure is that, by strategically placing


actuators in the support strucme, the surface shape can
Adaptive structure concepts offer unique design altema- be adjusted to minimize errors produced by uncertainties
tives for filled-aperture space-based astronomy missions in finite fabrication precision and/or on-orbit thermal
such as the proposed Sub-Millimeter Intermediate Mis- deformations. The same actuators may also be used
sion and the Large Deployable Reflector. These missions to prestress the joints of the truss structure in order to
are intended to observe the infrared and sub-millimeter remove nonlinearities that may exist due to small free-
wavelengths. They employ a large, segmented, filled- play in the joints.
aperture reflector supported by a truss structure. A re- The shape control of mss structures has been
quirement of a diffraction limited telescope is that the studied by several investigators. For example, in
telescope wavefront error should be of the order of A/lO, reference [I] optimization of the support truss was
where A is the minimum wavelength of interest. The cor- formulated so as to minimize the maximum actuator
responding reflective surface error is then determined to input. This had the effect of equalizing the corrective
be half the wavefront error, i.e. h/20. Typically, 90% force among all actuators. The actuators were not
of the surface error budget is allocated to the primary optimally placed-but many fixed actuators were used.
reflector. Therefore, for such reflectors, where the wave An approach for the optimal actuator placement was
length of interest is in the 100-1000 micron range, the considered in [Z] by first converting the discrete truss
surface accuracy of the primary reflector must be main- structure model into a continuum model, for which the
tained to the order of a few microns. optimum placement was sought. Using an equivalent
To a large extent, the surface accuracy of a continuum model has the advantage of reducing the
segmented reflector is determined by the dimensional computational cost of optimization in comparison with
accuracy of the structure supporting it. This is because discrete programming methods. However, applicability
the reflector panels, which are made to within one of the resulting placement for the continuum model to
micron RMS of theii correct figure, are positioned above the actual truss is not straightforward. As an altemative
the support structure by a panel actuation mechanism. to rigorous discrete optimization, heuristic techniques
The minimum achievable reflector surface error is, have been used at reasonable computational costs for
therefore, potentially limited by the range of motion the optimal placement [3-5]. Specifically, reference [51
of the panel actuators coupled with the accuracy of employed the simulated annealing heuristic to minimize
the support structure. In this context, the advantage shape distortions as well as the effect of nonlinear joint
* The research described herein was performed at the Jet behavior. Prestressing or preloadmg the truss joints was
Propulsion Laboratory. California Institute of Technology, under also studied in 161 with the objective of removing joint
contract with the National Aeronautics and Space Administration. nonlinearities.

0 1993 US Qovemment
Precision truss structures

Figure 1. Precision truss structure.

The analytical results referenced above investigated and data acquisition system. Discussion of the test
the feasibility of using actuators to adaptively alter sequence, results, and their correlation with the analysis
the shape or behavior of the structure to other states. predictions, and the major findings will conclude the
Experimental studies, however, have concentrated paper.
mostly on the field of adaptive optics, where the figure
of flexible mirrors is controlled by a number of discrete
actuators. For example, a test was performed 171
to verify the ability of piezoelectric film to create a 2. Analysis
predetermined deformed shape of a composite panel. In
PI, nine force actuators were used to adjust the figure Considering the back-up truss shucture in figure I , on-
of a 62 cm diameter spherical thin mirror. orbit static distortions can result from various sources
Using an. existing full scale space-erectable "ss such as: (i) thermal variations and/or gradients due
structure, figure I , the objective of this paper is to show to exposure to sun and shade, (ii) the presence of
analytical predictions and corresponding experimental gravity as the structure is assembled on-ground, and
validations with regard to the ability of a limited number (iii) finite manufacturing imprecision. But apart from
of actuators to significantly reduce initial smctural shape the source of aberrations, the aim here is to establish
aberrations. The structure is a four meter diameter, the degree to which a limited number of strategically
doubly curved tetrahederal erectable truss designed to placed actuators can restore the structural shape to a
serve as a backup structure for a precision segmented desired configuration. Assuming a linear response, the
reflector. It has been the subject of extensive previous objective of the analysis is to find the best locations
testing and characterization [9, IO]. In the sections that and gains for a given number of actuators so that a
follow, the analysis methodology is first described along prescribed distortion can be corrected. The actuator gain
with the simulation results, which the experiments will is defined here as the amount of travel or displacement
aim to verify. We then summarize features of the in an extensible axial member, i.e. active member. To
test structure, its active components, instrumentation, generate a realistic distorted shape of the structure in

241
M Salama et a/

figure I , a finite element analysis is first performed From equilibrium


to calculate the deformations at all degrees-of-freedom
(DOF) of the truss due to an initial 5 K m-' thermal ,!ITP= F (5)
gradient in the horizontal plane of the truss. This thermal
profile is selected from a thermal analysis performed in which F is a vector of extemal nodal forces. Since the
in reference [ 111 for the Sub-Millimeter Intermediate disturbance to the structure is due to the induced actuator
Mission. Subsequent analyses will then address the length changes Ao, and not due to extemal forces, then
problem of optimally placing a limited number of setting F = 0, combining equations (2) through (3, and
actuators and selecting their gains to minimize this initial defining K , = @'KB gives
distortion at selected degrees-of-freedom (or control
DOF) of the truss. This is equivalent to athermalizing U = SA0 (6)
the structure. But rather than requiring the actuators
to restore the distorted structure to its original shape, s =B~K;'B~KB.. (7)
equivalently one may view the negative of the initial
The matrix S is generally rectangular nd x n,. It
distortion as the deformed state which we desire the
represents the actuators influence coefficients at the
actuators to generate. In this paper, both the desired
control DOF, or the sensitivity of the control DOF to
and current configurations of the structure are assumed
a unit displacement gain in each of the actuators. In
to be known deterministically. If this is not the case, the
addition to equation (6). its inverse form may be also
approach described herein can be modified in terms of
required in the minimization of equation (1). This can
expected values and other statistical descriptors. In the
be obtained from
next section, we describe the analytical procedure used
to find the optimal actuator locations and gains required
to correct specific shape aberrations.
Let U * represent the vector of the desired deformation
Since the actuator locations are discrete variables,
state at a specified set of nd control DOF, and let U
a discrete optimization technique such as the simulated
represent the vector of deformation state induced at the
annealing algorithm [4] may be used to provide a
same control DOF by n, actuators which have been
computational framework for selecting the optimal
arbitrarily placed and their gains arbitrarily set to a
actuator locations and gains. This is summarized in
displacement vector, Ao. For any such a set of actuators,
figure 2. Again, description of the actuator locations
we adopt the error function, e, as a measure of the
is indicated by ones or zeros in the columns of the E ,
actuator's ability to achieve the desired shape correction:
matrix. For the initial simulated annealing iteration and
e = [(U' - u ) ~ ( u '- - U ) ] ' / * . all succeeding ones, the error function e, for a typical
(1) iteration r is evaluated in the bold box of figure 2 as
The optimal actuator locations and gains are those that follows:
minimize e with respect to all possible locations and (a) calculate S, from (7).
gains. (b) assuming that U , = U * , compute an approximate set
If n is the total number of DOF and m is the of gains Aor from (81,
total number of members in a linear truss structure, (c) given (a,b) above, find U , from the forward solution
the relationship between the complete set of nodal in (6).
displacements U and the complete set of m member (d) evaluate e, for the current locations from (1).
length changes A is obtained by the displacement
method [I21 from Depending upon the relative magnitude of e, for
A=PU (2) the current iteration and e,-, for the previous iteration,
and depending upon the current value of the annealing
where ,!I is m x n geometric compatibility matrix. parameter 0, the current candidate set of locations in
Additionally, the deformations, U , at the control DOF E. are either (i) accepted unconditionally if Ae, <
are 0, (ii) accepted only if it met the probability test
U = BdU (3) P = exp(-Ae,/O) 2 random, or (iii) rejected
where Ed is f l d x n matrix of zeros and ones that selects if neither of the other two conditions is satisfied.
the rows associated with the control DOF in question. It is the probabilistic acceptance of non-improving
The relationship between the forces P in all members solutions in (ii) that allows the minimization process to
and the length changes A. induced in a set of actuators climb out of local minima, and therefore distinguishes
(i.e. actuator gains) is: the simulated annealing algorithm from the traditional
iterative improving techniques. Acceptance of the
P = K(A - EaAo) (4) current actuator locations implies that these locations are
potentially optimal, and are therefore used as a basis
where K is a diagonal matrix containing the axial for comparison with future candidate locations, as well
stiffness E A / L , and E, is m x n, matrix of zeros and as for generating a new set of candidate locations for
ones which selects the members associated with the the next (r + I ) iteration. A new set of locations
actuator locations. is constructed by perturbing the lastly accepted set of

242
Precision truss structures

. -
... .. .,,I-

.. . .
dl) -11,- a,
I I

20 x , . ..
"
0 , .. ~
x
0 5 10 75 20 25
N. U 01 Aclive Members

Figure 3. RMS error reduction with increasing number of


actuators. (27control WF).

-.'I
wlor,
In view of these results, the following experiments were
designed to verify the general trend exhibited by figure 3.

3. Test set-up
Figure 2. Simulated annealing optimization algorithm.

locations at only one site (selected randomly). That 3.1. Testbed description
site is replaced by one of the unoccupied sites. The The test structure in figure 1 was designed and
frequency of accepting non-improving locations, as well fabricated as a support truss for a 2.4 m focal length
as the iterative process itself are all govemed to a large segmented reflector. The upper surface comer-to-comer
extent by the annealing parameter 0. Guidelines for the is nominally 3.9 m. The truss thickness is approximately
initial selection of 0, and a strategy for its subsequent 0.64 m. There are 150 graphiteepoxy struts, 300
reduction are discussed in [4]. aluminum erectable joint assemblies, and 45 aluminum
The analysis procedure described above was applied nodes forming the doubly curved tetrahedral truss. Of
to the structure of figure 1. In figure 3, the result of the 45 nodes, 27 fall on the upper surface and 18 fall
a parametric study is displayed in terms of the error on the lower surface. The structure is supported by
correction, e, which can be generated using various clamping the three central nodes of the lower surface
numbers of n, optimally placed actuators. In this to a 1130 kg steel block, which represents ground. The
example, the DOF to be controlled were chosen to consist erectable joint and node assembly design [9] is a half-
of the vertical deformation DOF at all 27 top nodes. The scale version of the joint originally developed for the
magnitude of the initial error (corresponding to U* at the Space Station Freedom. The use of this type of joint
control DOF) is that shown when no actuators are used, allows for inspace erection of a precise, highly redundant
i.e. when na = 0. A number of important observations structure. The joint halves at opposite ends of a shut
can be made in connection with figure 3. have opposite thread directions, allowing the strut to
First, for a statically indeterminate structure with n be used as a turnbuckle. Prior to installation, the strut
deformable DOF and m members, where n < m , one length is precisely adjusted and locked in place by jam
needs at least n actuators to produce exactly any desired nuts. The nominal node center-to-node center lengths are
shape at all n DOF Locations of the n actuators must, 0.813 50 meter for the core struts, range from 0.771 47 to
however, be optimally selected from among all possible 0.814 15 meter for the upper surface members, and range
m member locations. Fewer DOF in the controlled set from 0.877 16 to 0.91661 meter for the lower surface
allow one to use correspondingly fewer actuators to members. All lengths are set to within a tolerance of
provide the same degree of accuracy. From a practical f 5 microns.
point, however, one does not need to achieve the desired Extensive testing was performed on this structure
shape exactly, and therefore, one does not need to use during earlier studies [9, lo]. One significant result of
a large number of actuators, even if the number of DOF these tests, is the observed small nonlinearites, most
in the controlled set is large. A remarkable degree of likely due to finite precision in the joints. The static
closeness to the error correction, e, can be achieved by deflection test result in figure 4 is an example, where
strategically replacing a small percentage of the total the RMS of the distorted shape at the top surface nodes
members by active members with optimized gains. For is plotted versus the load applied at one of the comer
example, figure 3 shows that 47%, 70% and 82% of nodes. Two complete load/unload cycles are shown.
the initial error can be corrected respectively, with only For the first cycle, the RMS of the distortion does not
two, four, and six of the total 150 members being active. retum to zero when the load is removed. For the second

243
M Salama et al

Figure 6. LVDTS at 15 control DOF.

Figure 5. Lead screw active member.


relatively rigid support structure (known as a ganhy-
not shown). Their nominal accuracy is f0.25 microns.
cycle, the data points are nearly coincident-and when They are used with a IOW pass filter tO minimize noise
unloaded, the RMS distortion retums to the new starting problems, and a signal conditioner whose bandwidth is
point. m i s suggests that there is angu]ar reorienbtion &250 Hz. In addition to the displacement transducers,
of the joints, and possible joint nonlinearities, which the structure is instrumented with 14 thermocouples to
are being closed/opened as the load/unload progresses. monitor any thermal changes a m s s the upper surface
The impact of these nonlinearities is discussed furrher in and through the thickness of the mSS.
subsequent sections.
33. Data aquisition system
3.2. Actuators and instrumentation To ensure fast and accurate data recording, an automated
The actuators used throughout the experiments are data acquisition system (DAS) was used. It is capable
referred to as lead screw active members and are shown Of accepting Outputs from all LVDTs, load cells, and
In figure 5 as part of a complete assembly with adapters, thermocouples at once. It is also capable of initializing
a load cell and erectable joints. The actuator consists of the test as required. This includes zeroing the LVDT
a motorized micrometer as the primary component of an signals, zeroing the load cell signals to give the
embeddable active member. The motorized micrometer member force due to actuation alone, and recording
contains an M10 x 0.5 lead screw driven by a DC the thermocouple readings as the test progresses. As
gearmotor. A DC load cell is placed mechanically in the actuator forces are incrementally changed, the
series with the motorized micrometer to measure the force, displacement, and temperature data are acquired,
net load in the member. Each load cell signal is averaged over multiple samples, and stored. The DAS
conditioned by an amplifier. A magnetic encoder is is interfaced with a motor controller which operates
incorporated into the gearmotor assembly, but due to either in a manual mode to incrementally change the
rotational backlash in the gearhead, the encoder is a poor force in the actuators, or in an automoted mode in
displacement transducer. Instead, the DC load cell is which the maximum force in each actuator is given
used as the feedback control sensor. This lead screw along with the number of increments. In the latter
active member is more suitable for the present static mode, the DAS adjusts each actuator until the load cell
application over piezoelectrically driven active members, associated with each actuator reads that the desired level
because actuator gains of the order of 1 mm are required has been achieved. An automatic ovemde. is built into
here. This is much larger than can be provided by the DAS, which will cause the system to shut down if
piezoelectric active members. the maximum force level is exceeded in any single load
As in reference [IO], the structure is instrumented cell.
with fifteen linear variable differential transformers
(LVDTS) to measure the vertical deformations in the
Z-direction at the 15-nodes indicated in figure 6. These 4. Results: analysis and experiments
nodes fall on the three major diameters of the upper
surface and serve as the control DOF. The 15 control DOF To demonstrate the shape control procedure described
are a subset of the 27 control DOF used in the calculations in previous sections, a series of validation tests were
of figure 3. Their locations are chosen here only for performed. They involved successive use of optimally
easy access. The LVDTS are positioned in place from a placed two, four, and six active members to create a
244
Precision truss structures
Table I. Optimal 2, 4, 6 actuators and their gains.

2 actuators 4 actuators 6 actuators


Actuator no 38 103 34 97 98 99 34 36 64 97 98 99
Force (Ib) -50.0 22.7 -38.5 -46.0 -47.0 -50.0 -33.9 -18.0 3.0 -46.7 -50.0 -46.9

Table 2. Analysis detormations at the control DOF


(microns).
~

Coni01 DoF Oesired Number of actuators


2 4 6
1 -30.7 -26.0 -28.0 -31.0
2 -30.7 -32.0 -24.0 -29.0
3 -25.5 -27.0 -23.0 -26.0
4 -25.5 -24.0 -25.0 -27.0
5 -14.4 -20.0 -16.0 -15.0
6 -19.3 -20.0 -17.0 -19.0
7 -29.9 - 5.6 -30.0 -31.0
8 -29.9 - 4.3 -28.0 -27.0
9 -19.3
~~ - 9.0 -19.0 -19.0
10 -14.4 - 7.9 -16.0 -16.0
11 -23.5 + 1.0 -21.0 -24.0
12 - 7.2 + 0.7 - 2.8 - 5.0
13 - 7.2 -+ 0.8 - 4.4 - 7.9
14 + 4.7 + 4.7 + 4.0 + 4.0
15 + 4.7 - 0.7 + 5.0 + 3.5
RMS 82.68 63.4 76.98 78.9

incremental complete reversal of the loading, and


then reversion to zero loads-also incrementally. In
each loadinglunloading direction, ten increments were
Figure 7. Optimal locations for 2,4. and 6 actuators typically used and the loads in the individual actuators
were kept proportional. For each increment, the
deformation pattem, U , at the 15 control DOF that closely actuator forces and the corresponding displacements at
approximates the deformation pattem, U*, at the same the control DOF were recorded. The cyclic load/unload
DOF. if the structure had been subjected to the 5 K m-’ procedure allowed the observation of nonlinearities
thermal gradient in the Y direction. During testing, and/or residual deformations remaining when the
no thermal changes of any magnitude were intended structure was completely unloaded.
to influence the structural deformation. The purpose of The single-actuator tests consisted of placing one
the thermocouples was to monitor thermal changes-if actuator at a time at one of the eight locations of
any-from the beginning to the end of each test. table I and figure 7, then loadinglunloading the actuator
As discussed in section 2, U and U * can be made incrementally to f 5 0 Ib. These tests essentially
closest in an RMS sense if the actuators were optimally measured the actuator sensitivity matrix, S , of equation
located and their gains also optimized. These optimal (7). experimentally. Figure 8 is an example of the
locations are depicted in figure 7 and table 1 for the test result for the actuator at location ‘97’. Residual
two, four and six groups of actuators. Notice that while deformations of the order of 4 microns can be observed
the optimal four-actuator configuration is a subset of the when the load in the member is completely removed.
six-actuator configuration, the two-actuator configuration Other than this, the load deformation curve is typically
is not. The corresponding optimal gains for each of the linear. With this preliminary characterization, the
three configurations are also listed in table 1. Consistent following three tests aimed to emulate the analysis
with these, the resulting Z deformations at the 15 control conditions leading to the results of tables 1 and 2 for
W F were calculated. These are listed in table 2. the two-, four- and six-actuator configurations.
A series of four test configurations was conducted In the two-actuator test, one actuator was placed at
which included (1) eight different single actuator location ‘38’ and another was placed at location ‘103’ in
tests, (2) two actuators, (3) four actuators, and (4) figure 7. These were then incrementally loadeflunloaded
six actuators. Prior to each test all load cells simultaneously between (-50.0, +22.7) and (+50.0,
and displacement transducers were initialized to zero. -22.7) pounds. The deformations at the 15 control
Each test then involved commanding the actuators DOF at the maximum load are displayed in figure 9.
in question to adjust their lengths incrementally until A negative sign indicates deformation in the positive Z
their load (measured by the load cells) reached the direction. The two values shown in each box correspond
maximum desired levels. This was followed by to results of the test and analysis (table 2) respectively.

245
M Salama et a/
-19.8 -28.3
p-mq p3xq
-
-

.40-

.60-

.40 .zo 0 20 40
AclYalOl laas Ibl,

Figure 0. Sensitivity measurement for actuator at location


‘97’.
The third value (above each box) is obtained by
numerically combining the results of the appropriate Two Actuators @ 38, 103
singleactuator tests using equations (2).(4) and (6). In Figure 9. Tesffanalysiscomparison for two actuators.
the present case, these are the single-actuator tests for
location ‘38’, and for location ‘103’. The two missing
values are due to failure of the LVDT’s signal conditioner
at these locations during testing with actuator ‘103’. The
numerical combining of the two single-actuator tests
provides another method of correlating the analytical
and experimentally measured sensitivity matrix, S. The
two sets of experimentally derived values in figure
9 are quite consistent among themselves, and when
compared with the corresponding analytical prediction,
one finds generally good agreement. Deviations of
the order of 20% represent the norm, which is to be
expected. Another measure of the quality of agreement
is the RMS of the deformation at the control DOF for
the test, analysis, and desired state. These are 47.8,
63.4 and 82.7 microns respectively. To verify the
optimality of the analytically derived optimal actuator
pair (38,103), the experimental values of the single-
actuator tests for an arbitrarily chosen pair (38,98) were Fow Aclualos @ 34,97.98,99
combined numerically. Notice that actuator ‘38’ is Figure 10. Tesffanalysis comparison for four actuators.
common between the optimal and non-optimal pairs,
The resulting deformations at the 15 control DOF are: same figure by the second set of values inside each
(-7.9, 1.4, -2.4, 1.0, 0.36, 2.7, -0.75, 0.37, 1.8, 2.9, box. Clearly these two sets of results are in poor
-1.3, 2.8, 3.7, 3.8, 9.5). with RMS= 14.8. Compared correlation. However, it was observed during the test
with the desired values in table 2 with RMS= 82.7, the that at points of the loading path when the actuators
optimal actuator pair (38,103) with RMs= 47.8 is clearly were completely unloaded, residual deformations as
better than (38.98). large as 30 microns remained at some of the control
Except for the number of actuators used and their DOF. The possibility that thermal changes that may have
locations and gains (see tables 1 and 2 and figure 7), accrued over the test duration may have caused these
the four- and six-actuator test configurations proceeded deformations was discounted on the basis of analysis
in the same manner as in the two-actuator case. The of the thermal environment. Data recorded by the 14
results of the four-actuator test, figure 10, compare more thermocouples mounted throughout the structure showed
favorably with the analysis results than did the two- temperature changes of the order of 0.5 “C, which can
actuator test. Comparing the individual response values produce no more than a 5-10 micron deformation at
at the control DOF, deviations of the order of 7% are any DOF. Inherent joint nonlinearities are believed to be
typically observed. The RMS of these values for the test the only possible source for these residual deformations,
and analysis are, respectively, 81.5,77.0, compared with and since the analysis assumed a linear elastic behavior,
the desired value of 82.7 microns. subtracting the residual deformations from the total
The test results for the six-actuator configuration observed values (above each box) should provide a more
are shown in figure 11 at the control DOF above the rational basis for comparing the test and analysis results.
box, while results of the analysis are indicated in the The so-modified test results are represented by the first
246
Precision truss structures

41.2 68.8 5. Discussion and conclusions


)%XXiP 139.0:29.1(
Using a realistic space erectable truss smcture, the
analysis and tests described herein demonstrated the
ability of a limited numbers of active members to
produce the desired shape corrections with reasonable
accuracy. In a series of tests, successively increasing
number of actuators were used to produce the desired
shape correction with successively increasing degree of
accuracy. Correlations with the analytical predictions
were good under all test conditions where nonlinearities
were negligible.
The most severe nonlinearities were observed in
the six-actuator test, where residual deformations of up
to 30 microns were observed when all actuators were
completely unloaded. In spite of the fact that joints of
this truss were specially designed and built to a high
degree of precision, they appeared to be the cause of
Six Acmators @ 34,3664.9'7.98.99 the observed micron level nonlinearities. Coupled with
random variations in the geometric dimensions of the
Figure 11. TesUanalysis comparison for six actuators. truss members, and angular tolerances in the orientation
of the various members meeting at a joint, it is plausible
that small nonlinearities in some of the joints would
react preferentially depending upon the load distribution
throughout the truss. This may explain the small residual
deformations (of the order of 5 microns) in cases of
single, two, and four actuators, where the relatively
localized loading could exercise only a few nonlinear
joints, as compared with the residual deformations of up
to 30 microns in the six-actuator case, where the loading
is more distributed and could have dominance over a
larger number of nonlinear joints.
I
I Although these micron level nonlinearities obscured
0 2 4
Nu"
6 8
0, Act"ill0rs
10 12 ?d an otherwise good correlation with the linear analysis,
their presence adds special significance to the use of
Figure 12. RMS error reduction with increasing number of adaptive concepts in precision structures. Adaptive
actuators comparison of analysis and test results for 15 concepts are especially aimed at structures with such
control OOF. sources of uncertainties in behavior. In a closed-loop
setting, once these residual deformations are sensed, they
value inside each box in figure 11. The blank spaces can be corrected adaptively by the active members to any
indicate no data due to failure of the LVDT'S signal desired degree of accuracy. The use of active members
conditioners at these locations. The modified test values in precision structures can also ease, and may even
show greatly improved correlation with the analysis provide an attractive altemative to, requirements that
values. The RMS of the modified test deformations are certain structural or optical components be manufactured
now 88.4 microns compared with 78.9 for the analysis from 100% dimensionally stable materials.
and 82.7 for the desired deformations.
To summarize the comparison between analysis
and test results for the two-, four-, and six-actuator Acknowledgments
cases in a form similar to figure 2, we take the RMS
of the difference between the desired and achieved The assistance of several colleagues, R Helms, R Losey,
deformations at the 15 control DOF to represent the C Miller and D Moore, is gratefully acknowledged.
remaining RMS error. The results are displayed in figure
12 as percentages of the RMS shape e m r still remaining References
when one uses zero, two, four or six actuators to correct
the truss shape. It is encouraging that apart from [I] Padula S L. Adelman H M and Bailey M C 1987
the nonlinearities described earlier, the present analysis Inkgrated structure electromagnetic optimization of
and test results correlated well; both confirming that a large space antenna reflector NASA TM-89110.February
I987
limited number of strategically placed active members [Z] Burdisso R A and Haflka R T 1989 Optimal location
could indeed be used to correct shape aberrations with of actuators for correcting distortions in large buss
reasonable accuracy. structures AIAA J. 27 1406-11

247
M Salama et a/

I31 Haftka R T and Adelman H M 1985 Damping and control control experiments with a functional model for
of spacecraft structures: selection of actuator locations large optical reflectors Proc. First USlJapan Conf.
for static shape control of large space structures by on Adaptive Structures (Maui, HI,November 1990)
heuristic integer programming Comput. Struct. 20 (Lancaster, PA: Technomic) pp 615-30
575-82 [91 Bush H.Herstrom C, Heard W, Collins T, Fichter W,
[41 Chen G S , Bruno R and Salama M 1991 Optimal Wallsom R and Phelps J 1990 Design and fabrication
placement of activelpassive members in mss structures of an erectable truss for precision segmented reflector
using simulated annealing AIAA J . 29 1327-34 application 31st AIAA Structures Structural Dynamics
[SI Bruno R, Salama M and Carba J 1992 Actuator and Materials Conference (Long Beach, CA, April
placement for static shape control of nonlinear truss 19901 pp 45442
structures Pmc. Third Int. CO$. on Adaptive Structures [lo] Umland J W and Lou M C 1992 Precision segmented
(San Diego, CA, November 1992) (Lancaster, PA reflector primary support structural testing 33rd AIAA
Technomic) pp 433-45 Svuctures Structural Dynamics and Materials Cor&
161 Das S K, Utku S,Chen G-S and Wada B K 1990 (Dallas. TX,April 1992) paper 92-2535
A mathematical basis for the design and design [Ill Tsuyaki G and Mahoney M J 1991 The precision
optimization of adaptive trusses in precision control segmented reflector program: on-orbit behavior of the
Pmc. First USlJapan Conf. on Adaptive Structures sub-millimeter imager and line survey telescope 26th
(Maui. HI, November 1990) (Lancaster, P A Technomic) AIAA Thennophysics CO$. (Honolulu, HI,June 1991)
pp 660-88 I121 Martin H C Introduction to Matrix Methoh of Structurol
[7l Kuo C P and Bruno R 1990 Optimal actuator placement in Anulysis (New York McGraw-Hill) pp 298-9
an active reflector using a modified simulated annealing
technique Proc. First USlJapan Conf. on Adaptive
Structures (Maui. HI.Navember 19901 . (Lancaster.
. PA
Technomicj pp 105G57
[81 Tabata M. Itoh N. Mivawaki K. Satori A. Ive M.
..
Yamashita Y,Noglchi T and Tanaka W i99O'Shape

248

You might also like