Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

D O U B L E - L A Y E R TENSEGRITY G R I D S : STATIC L O A D

R E S P O N S E . I I : EXPERIMENTAL STUDY
By Ariel Hanaor, 1 Member, ASCE

ABSTRACT: Tensegrity structures are freestanding prestressed cable networks in


which the cables are prestressed against a discontinuous system of bars. In double-
Downloaded from ascelibrary.org by Universidad De Sevilla on 06/15/15. Copyright ASCE. For personal use only; all rights reserved.

layer tensegrity grids (DLTGs), the bars are confined between two parallel layers
of cables. This is the second paper in a two-part analytical and experimental study
of a type of DLTG. The first part, presenting results of a first-order linear ana-
lytical model, indicates that these structures possess low stiffness and low bar force
efficiency. The experimental investigation of a small-scale model indicates that
actual response is significantly nonlinear and that both stiffness and bar force ef-
ficiency are higher than indicated by the linear model. Member forces due to the
applied load are generally higher than the linear model indicates. A nonlinear an-
alytical model is generally in good agreement with the results. The concept, con-
sisting of independent prismatic units, possesses a high degree of structural re-
dundancy. Load-bearing capacity is practically unaffected by the loss of a member.

INTRODUCTION

The concept of double-layer tensegrity grids (DLTGs) has been introduced


in the first of this two-part study (Hanaor and Liao 1991). The type of grid
studied consists of independent tensegrity prisms (Hanaor and Liao 1991)
joined together to form two parallel layers of cables prestressed against a
discontinuous system of bars confined between these layers (Fig. 1).
The complex appearance of these structures notwithstanding, their con-
struction is simple. The individual prismatic units are constructed and pre-
stressed independently, and then joined in their prestressed condition. Each
unit has one state of prestress (and one internal mechanism) and can be
prestressed by elongation of one bar, although elongation of all three bars
in each unit is usually required to maintain the geometric regularity. The
simplicity of the connection of a rigid bar to cables [Fig. 2(b)] makes the
construction both relatively easy and economical.
Since DLTGs rely on prestress for their integrity, they are inherently col-
lapsible in the unprestressed state. A deployable tensegrity grid can be made
from telescoping bars. The structure can then be deployed by elongating the
bars either mechanically or by means of fluid pressure (e.g., air). Once de-
ployed, additional prestress may be introduced as required. Deployable models
have been produced and demonstrated by the author, but they do not form
part of the present study.
The analytical part of this study employed a flexibility-based linear (small-
deflections) model to obtain the structural response to static load. The results
indicate very low flexibility with displacements of about 1/20 of the span
in the state of full prestress (i.e., when the first cable becomes slack). The
'Res. Fellow, Nat. Building Res. Inst., The Technion-Israel Inst, of Tech., Haifa
32000, Israel; formerly, Asst. Prof., Dept. of Civ. Engrg., Rutgers Univ., Pisca-
taway, NJ 08855.
Note. Discussion open until November 1, 1991. Separate discussions should be
submitted for the individual papers in this symposium. To extend the closing date
one month, a written request must be filed with the ASCE Manager of Journals. The
manuscript for this paper was submitted for review and possible publication on Jan-
uary 22, 1990. This paper is part of the Journal of Structural Engineering, Vol.
117, No. 6, June, 1991. ©ASCE, ISSN 0733-9445/91/0006-1675/$1.00 + $.15 per
page. Paper No. 25904.

1675

J. Struct. Eng. 1991.117:1675-1684.


Critical cable
s - Slack cable
Downloaded from ascelibrary.org by Universidad De Sevilla on 06/15/15. Copyright ASCE. For personal use only; all rights reserved.

b - Critical bar
1-18 - Caged bars
D- Displacement "transducer
P-Load
R - Vertical support

FIG. 1. Test Mode!

analytical model also suggests that to maintain a state of full prestress in


grids of seven units or more across the span, nearly 80% of the critical bar
capacity is taken by prestress—bar force efficiency of 20%. The bar force
efficiency increases to over 40% for the smallest grid—three units across.
In a state of underprestress bar force efficiency increases but stiffness de-
creases (i.e., deflections increase). Some cables may become slack in this
state.
A previous small-scale model test (Hanaor 1987, 1990) indicated that overall
stiffness and load-bearing capacity were in reasonable agreement with the
linear analytical model, but the test also suggested that member force dis-
tribution deviated substantially from that predicted by the model. This test
was on a very small model, comprising only three prism units in total, and
it was rather scantily instrumented. The failure mechanism was governed by
member buckling.
The objectives of the test presented in this paper are to compare linear
and nonlinear analytical results with experimental results of a larger model
than the one cited and to monitor in more detail member force distribution
under varying levels of prestress.

EXPERIMENTAL PROCEDURE

Description of Test Model


The test model and its instrumentation are shown in Figs. 1 and 2. The
physical dimensions and support and loading conditions are indicated in Fig.
1676

J. Struct. Eng. 1991.117:1675-1684.


Downloaded from ascelibrary.org by Universidad De Sevilla on 06/15/15. Copyright ASCE. For personal use only; all rights reserved.

jres*«
r* '•'. i
I •^.•ds*--'11'"' -<.-f.

1' !'
rss*

-wWS^&<
I
1

^>wLv:*<-\^

FIG. 2. Test Setup, (a) General View; (b) Connection Detail

1677

J. Struct. Eng. 1991.117:1675-1684.


1. The model is similar in layout to the three-unit span grid of the companion
paper (Hanaor and Liao 1991), consisting of seven prismatic units.
Bars were made of steel tubes of 11/16 in. (17.5 mm) outside diameter
and 0.5 in. (12.6 mm) inside diameter, with a buckling force of approxi-
mately 3 kips (13.3 kN). The cables were 1/4 in. (6.4 mm) diameter stain-
less steel cables with a rupture force of approximately 1.7 kips (7.6 kN).
Downloaded from ascelibrary.org by Universidad De Sevilla on 06/15/15. Copyright ASCE. For personal use only; all rights reserved.

The instrumentation comprised linear variable displacement transducers


(LVDTs) at the center of the structure and at each of the supports, yielding
the net average vertical displacement of the three central (loaded) nodes.
Eighteen of the 21 bars had pairs of strain gages mounted on opposite faces.
The average strain readings were calibrated to yield member forces. The
locations of the eighteen pairs of strain gages are indicated in Fig. 1 by
numbers on the respective bars.
The structure was mounted on an electromechanical actuator, producing
a constant rate of deflection of approximately 1.3 in./hr (33 mm/hr). The
load was transmitted to the three central nodes via a common plate on which
they were supported, and the total load was monitored by a load cell. The
grid was loaded up and anchored vertically down to the reaction floor at the
locations indicated in Fig. 1 (see also Fig. 2).
The support system is somewhat different from the one used in the ana-
lytical study, in order to simplify it and eliminate stresses introduced by
maladjustment of supports. This reduced support necessitated the addition
of some boundary members to prevent large mechanism displacements. The
additional boundary members consisted of bars in the bottom layer, which
is in compression, and cables in the top layer, which is in tension.
All data was logged and processed by a computer-controlled data acqui-
sition system. The program calculates and displays member forces, displace-
ments and loads, based on input calibration data, at each load step. The load
steps can be automatically or manually controlled.

Testing Procedure
In all, four tests were performed. Three tests at three different prestress
levels were aimed at observing the effect of the level of prestress on the
response. The fourth test was carried through to failure. Initial calibration
tests were carried out on isolated gauged bars to derive axial forces directly
from strain-gauge bridge measurements. The nominal elastic moduli and ul-
timate forces in bars and cables were also determined in these preliminary
tests (for calibration of the analytical model).
Prestressing was achieved by extending the bars by means of the threaded
rods inserted in them [see Fig. 2(b)]. As bars were extended, the forces in
them were monitored and the lengths adjusted to give the desired average
prestress forces. Although the prestress forces in all bars are nominally iden-
tical, substantial scatter was actually found due to inaccuracies in the con-
struction (see results in the following).
Prestressing of each test was carried out from the final state of the previous
test, with the boundary members released (to avoid distortion of the ge-
ometry). The boundary members were then attached with zero initial stress.
The prestress levels in the first two tests, measured as the average bar force,
were: 250 lb (1.11 kN), 500 lb (2.22 kN), and 1,000 lb (4.45 kN). The
theoretical full prestress at failure (governed by cable rupture) was 1.5 kips

1678

J. Struct. Eng. 1991.117:1675-1684.


(6.67 kN) and the first-order analysis indicated a failure load of 3 kips (13.3
kN).
Each of the first three tests involved loading the structure at a constant
rate of deflection to less than half the predicted failure load, then unloading
at the same rate. The prestress was then increased to the next level and the
procedure resumed. The last (fourth) test was carried through from the un-
Downloaded from ascelibrary.org by Universidad De Sevilla on 06/15/15. Copyright ASCE. For personal use only; all rights reserved.

loaded condition of the preceding (third) test to failure.

RESULTS

Experimental and analytical results are presented in Figs. 3 through 6.


The experimental results are compared to results of two analytical models:
a linear flexibility-based model (Hanaor 1988; Hanaor and Liao 1991), and
a nonlinear large-displacements model (Argyris and Scharpf 1972). Inter-
pretation of the results and their significance is provided below for the load-
deflection response and for member forces.

Load Deflection
The load-deflection curves of the four tests are presented in Fig. 3. It can
be observed that the first-order analysis (Hanaor and Liao 1991) predicts
well the initial slope of the curve in all tests, but the actual curves are non-
linear, strain hardening. The result is that actual displacements for a given
load are considerably smaller than predicted by the first-order linear analysis.
The nonlinear analytical model, on the other hand, generally predicts the
response very well. The lower the level of prestress, the larger the nonlin-
earity, as would be expected. For instance, the total displacement at 1 kip
(4.45 kN) for test 1 and for test 2 are approximately the same whereas linear
theory would predict displacements of test 2 to be approximately half those
of test 1.
The unloading curve terminates at a slope similar to the initial but with a
substantial residual displacement. This displacement may be due to stretch-
ing of the cables and slippages in the fasteners. Some force relaxation is
also observable. For instance, the initial prestress of test 4, which is the
unloaded state of test 3, was 100 lb (445 N) lower than the initial prestress
of test 3.
The increasing nonlinearity with decreasing prestress, implies that the ul-
timate deflection is relatively insensitive to the level of prestress. The ulti-
mate deflection/span ratio of approximately 1/25 obtained in test 4 is, there-
fore, probably representative of the overall stiffness of this type of structure.
Test 4 [Fig. 3(d)] shows additional important features of DLTG response.
The structure failed by cable rupture as predicted, but the failure load—
1,750 lb (7,784 N)—was some 42% below that predicted by the linear the-
ory, suggesting considerably higher member forces. This suggestion is con-
firmed by observations on member forces (see the following). The failure
of the first cable implies the loss of the tensegrity unit to which it belongs,
as confirmed by observation of the forces in the corresponding bars. The
remaining units, however are intact so that the structure as a whole still
possesses substantial stiffness. This feature was confirmed by the test. Fol-
lowing first cable rupture, which was accompanied by substantial load re-
laxation, the structure resumed loading to a level almost identical with that
of the first cable rupture. It was then followed by rupture of the second

1679

J. Struct. Eng. 1991.117:1675-1684.


50 mm
5KN
Downloaded from ascelibrary.org by Universidad De Sevilla on 06/15/15. Copyright ASCE. For personal use only; all rights reserved.

1 2 in.
Central deflection
100 mm

//
slippage// ^ ^ ^

_,- S .

1 2
Central deflection

FIG. 3. Load-Deflection Curves, (a) Test 1; (b) Test 2; (c) Test 3; (d) Test 4 (Ul-
timate Load)

cable, representing a loss of 29% of the units. Even at this stage the re-
maining structure still possessed substantial stiffness, as indicated by the
slope of the curve following the second rupture. At this point, however,
deflections were excessive (1/10 of the span) and the test was terminated.
Test 4 thus brings out an important feature of this type of DLTG, namely
the high structural redundancy. Being constructed of independent units, the
loss of one unit by no means represents collapse of the structure, but only
1680

J. Struct. Eng. 1991.117:1675-1684.


Downloaded from ascelibrary.org by Universidad De Sevilla on 06/15/15. Copyright ASCE. For personal use only; all rights reserved.

CD
oo

J. Struct. Eng. 1991.117:1675-1684.


Applied load
Applied load Applied load

FIG. 4. Maximum Member Forces (Mem- FIG. 5. Average Member Forces


bers 3, 9, 15) FIG. 6. Bar Forces in Critical Units
local failure. The load-carrying capacity of the structure as a whole is vir-
tually unaffected. The test featured an extreme mode of failure (cable rup-
ture), which is drastic, with no warning signals and no inbuilt ductility. It
is more desirable (and indeed, efficient) to design for bar buckling failure
mode. By providing the bars with substantial slenderness ratio (the bars in
the test model had a slenderness ratio of 130) a relatively ductile failure
mode can be secured, allowing for force redistributions (Hanaor 1987, 1990).
Downloaded from ascelibrary.org by Universidad De Sevilla on 06/15/15. Copyright ASCE. For personal use only; all rights reserved.

Member Forces
Member forces for selected member groups are presented in Figs. 4 and
5 for tests 1-3 and in Fig. 6 for test 4.
Fig. 4 presents member force versus the applied load for the most highly
stressed bars (numbers 3 , 9 , and 15 in Fig. 1). The figure shows the average
force and the range (scatter) of the actual forces. The following observations
can be made:

1. The actual bar forces increase faster with the applied load than predicted
by the first-order model, but slower than predicted by the nonlinear model.
2. The member forces increase practically linearly with the load.
3. The higher the prestress level in the bar, the closer the actual slope of the
curve is to that of the linear analytical model.

The overall mean bar forces are presented in Fig. 5 for tests 1-3. This
mean includes bars that are considerably less stressed than those of Fig. 4
(indeed in members 4, 10, and 13 tension is induced by the load). Here the
discrepancy between the linear model and the test is even larger than for the
more highly stressed bars. The nonlinear model, while still overestimating
the rate of increase of bar forces, predicts the shape of the curve well. An
initially nonlinear segment is followed by essentially linear behavior but with
slope considerably higher than that of the linear model.
The discrepancy between the nonlinear model and test results for bar forces
could be due to force relaxation in the cables, as well as to inaccuracies in
modelling the geometry and the elastic properties of members. The good
agreement of the load-deflection curves suggests an underestimate of cable
forces by the model (see further discussion as follows). The load relaxation
is clearly evident in the curve for test 3, where two nonlinear analytical
curves are provided—one for the initial prestress and one for the final.
The higher rate of increase of member forces compared with the linear
model means that the linear model is not acceptable for design purposes as
it produces an unsafe design. On the other hand, this effect implies higher
stiffness than the linear model predicts, allowing for lower levels of pre-
stress. Indeed, the level of full prestress (at which cables become slack) as
predicted by the nonlinear model is considerably lower than suggested by
the linear model.
Forces in the critical units, i.e., the units containing the most highly stressed
(critical) cables (Fig. 1) are presented in Fig. 6, as an indirect measure of
cable forces (cable forces were not monitored due to the difficulty of at-
taching strain gauges to cables). In each of these units, forces in two of the
three bars were monitored. Forces in each of these two groups are shown
separately in Fig. 6 (group I contains members 1 , 7 , and 12, and group II
contains members 2, 6, 11). In addition, the forces of the two bars in the

1682

J. Struct. Eng. 1991.117:1675-1684.


unit that failed first (bars 11 and 12) are shown separately. The following
observations are made:

1. Relation between analytical and observed bar forces are similar to those
noted for other bars.
2. The forces in the two members of the first critical unit are the highest in
Downloaded from ascelibrary.org by Universidad De Sevilla on 06/15/15. Copyright ASCE. For personal use only; all rights reserved.

their respective groups for most of the loading (their average force is the highest
throughout), indicating that this unit contains the most highly stressed cable.
3. The nonlinear analysis indicates an average critical cable force at the peak
load of 1.2 kips (5.5 kN). If it is assumed that scatter in actual cable forces is
similar to that of bar forces, the actual maximum cable force in the critical unit
can be estimated as 1.3 kips (5.8 kN), which is some 25% below the cable-
tested rupture load [1.7 kips (7.6 kN)]. This result suggests that the nonlinear
model underestimated cable forces, as conjectured, based on comparison of re-
sults for deflections and for forces.

SUMMARY AND CONCLUSIONS

This is the second part of a two-part study of the response of a type of


double-layer tensegrity grid (DLTG) to static load. The first part (Hanaor
and Liao 1991) presented an analytical investigation exploring the effects of
span and structural depth on deflections and member forces. The second part
is an experimental investigation of the response of a model grid.
The main results of the experimental investigation, particularly as related
to the analytical results are:

1. The first-order analytical model underestimates the stiffness of DLTGs.


The actual displacements are considerably smaller than predicted by the model.
The nonlinear model predicts the load-deflection response well.
2. Member forces increase faster with the applied load than the simplified
model predicts, and this effect is larger the lower the force is. The nonlinear
model somewhat overestimated the bar forces and appears to have underesti-
mated the cable forces. The reasons for these minor discrepancies are not clear
at present.
3. DLTGs constructed from individual prismatic units possess a high degree
of structural redundancy. The loss of one unit through failure of a member does
not significantly affect the load-bearing capacity of the whole structure.
4. It is desirable to design DLTGs for bar buckling as the governing failure
mechanism. High bar slenderness ratios can provide ductility and enhance force
redistribution capabilities.
5. A design of tensegrity grids must be based on a nonlinear analytical model
accounting for large deflections. A linear model generally produces an unsafe
design.

ACKNOWLEDGMENTS

The assistance of senior students Paul Hanley (in preparing the data-ac-
quisition program), and Anna Vargas (in constructing the test model) is
gratefully acknowledged. The work has been partially supported by a grant
from the National Science Foundation.

1683

J. Struct. Eng. 1991.117:1675-1684.


APPENDIX. REFERENCES

Argyris, J. H., and Scharpf, D. W. (1972). "Large deflection analysis of prestressed


networks." J. Struct. Div., ASCE, 106(3), 633-654.
Hanaor, A. (1987). "Preliminary investigation of double-layer tensegrities." Non-
Conventional Structures, Proc. Int. Conf., H. U. Topping, ed., Civil Comp. Press,
London, England, Vol. 2, 35-42.
Downloaded from ascelibrary.org by Universidad De Sevilla on 06/15/15. Copyright ASCE. For personal use only; all rights reserved.

Hanaor, A. (1988). "Prestressed pin-jointed structures—Flexibility analysis and pre-


stress design." Cotnput. Struct., 28(6), 757-769.
Hanaor, A. (1990). "Double-layer tensegrity grids: Geometric configuration and be-
haviour." Space structures: Theory and design, H. Nooshin, ed., Multi-Science,
London, England, in press.
Hanaor, A., and Liao, M. K. (1991). "Double-layer tensegrity grids: Static load
response. I: Analytical study." J. Struct. Engrg., ASCE, 117(6), 1660-1674.

1684

J. Struct. Eng. 1991.117:1675-1684.

You might also like