Research ArticleAnalysis of Tensegrity Structures with Redundancies,by Implementing a Comprehensive Equilibrium EquationsMethod with Force Densities
Miltiades Elliotis,1 Petros Christou,2 and Antonis Michael2
1Department of Mathematics and Statistics, University of Cyprus, Nicosia, Cyprus2Department of Civil Engineering, Frederick University, Nicosia, Cyprus
Correspondence should be addressed to Miltiades Elliotis; [email protected]
Received 26 April 2016; Revised 26 July 2016; Accepted 19 October 2016
Academic Editor: Hassan Safouhi
Copyright Β© 2016 Miltiades Elliotis et al. This is an open access article distributed under the Creative Commons AttributionLicense, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properlycited.
A general approach is presented to analyze tensegrity structures by examining their equilibrium. It belongs to the class ofequilibrium equations methods with force densities. The redundancies are treated by employing Castiglianoβs second theorem,which gives the additional required equations. The partial derivatives, which appear in the additional equations, are numericallyreplaced by statically acceptable internal forceswhich are applied on the structure. For both statically determinate and indeterminatetensegrity structures, the properties of the resulting linear system of equations give an indication about structural stability. Thismethod requires a relatively small number of computations, it is direct (there is no iteration procedure and calculation of auxiliaryparameters) and is characterized by its simplicity. It is tested on both 2D and 3D tensegrity structures. Results obtained with themethod compare favorably with those obtained by the Dynamic Relaxation Method or the Adaptive Force Density Method.
1. Introduction
Cable networks [1β3] and tensegrity structures [4] are dif-ferent from conventional structures, such as spatial steelframes or space steel trusses, in that they are lightweightstructures withmembers which transmit only tension (cablesand strings) or elements which transmit compression (barsbefore buckling). In this article, we are studying only thebehavior of tensegrity structures.These structures are usuallydefined as planar or spatial trusses with a discontinuousset of members under compression, inside a continuousnetwork of members under tension. The word tensegrityis an artificial word and it combines the words βtensionβand βintegrity.β This word was coined several decades ago.Professor Fuller, in the United States, was essentially involvedin the invention of this technical word. In one of his lastbooks, Fuller described the compressionmembers as βislandsof compression in a sea of tensionβ [4]. Using the sameconcept, Emmerich [5] presented, in France in 1963, hisown tensegrity patent. Snelson, one of Fullerβs students,
describes this type of structures as βcontinuous tension anddiscontinuous compression structuresβ [6].
Pretension, applied by means of tension members, playsan essential role in the structural behavior of the tensegrities.For the design of such structures, their stability is investigatedunder both static and dynamic loads. During the last decades,many methods have been proposed for the analysis of thetensegrities. One of the most important methods is theDynamic Relaxation Method (DRM) which was used inthis research for checking the numerical results obtainedwith the numerical scheme proposed in the present work.DRM is one of the classical techniques. It belongs to thefamily of methods under the title of three-term recursiveformulae. It is an iterative procedure which is based on thefact that a system undergoing damped vibration, excited bya constant force, ultimately comes to rest in the displacedposition of static equilibrium, obtained under the actionof the constant force. One of the numerous first papers,written by pioneers of this method, is that of Papadrakakis[7] which proposes an automatic procedure for the evaluation
Hindawi Publishing CorporationAdvances in Numerical AnalysisVolume 2016, Article ID 4815289, 12 pageshttp://dx.doi.org/10.1155/2016/4815289
2 Advances in Numerical Analysis
of the iteration parameters and it is mentioned here (withoutunderestimating the importance of other works in thisdomain) as an example of an article which presents in astrict, clear, and academic manner the way of implementingthis numerical procedure. Recent research, performed inthe domain of tensegrities, has given some new importanttechniques. However, a general review of the older and of therecent numerical schemes, developed in this area, is out ofthe scope of the present paper. Juan and Mirats Tur presentan excellent review in their work [8] of the basic issuesabout the statics of tensegrity structures. Among the newmethods, presented in literature during the last few years,is that of Zhang and Ohsaki [9] which is also considered inthe present research for comparison purposes. Their methodis an Adaptive Force Density Method (AFDM). It first findsa set of axial forces compatible with a given structure andthen estimates the corresponding nodal coordinates underequilibrium conditions and constraints.
The technique implemented in the present work is a forcedensity method which belongs to the class of the EquilibriumEquationsMethodswith ForceDensities (EEMFD).With thismethod, the system of equilibrium equations is created byconsidering the equilibrium of forces in all the joints. Theredundancies are treated by employing Castiglianoβs secondtheorem which gives the additional equations required tohave a solution [10]. The partial derivatives, which appearin the additional equations, are numerically replaced bystatically acceptable internal forces acting along themembers.Also, the properties of the matrix of the system of linearequations are exploited because they give a strong indicationof the stability of the structure.
The assumptions adopted in the present research are thefollowing:
(1) Joints are frictionless but their mass is considered incalculations unless otherwise specified.
(2) The self-weight of a member (for structures withinearthβs gravity field) is not neglected unless otherwisespecified. It is equally distributed at its ends.
(3) Live loads and pretensions on a member are trans-ferred to the joints.They are equally distributed at theends of the member.
(4) Displacements on the joints and deformations ofthe members of the structure are relatively smallcompared with the dimensions of the structure.
(5) The axial force carried by a member is constant alongits length.
(6) The materials used, for all members of the structure,obey Hookeβs Law for loadings below the yield stress.
The efficiency of this technique is based on its simplicity andthe small number of calculations required.Thus, the objectiveis (i) to present the general idea and formulation and (ii) totest the method in planar and spatial tensegrity structures ofany type of complexity and to compare it with the DRM orthe AFDM.
The outline of the rest of this paper is as follows: inSection 2 we develop the formulation of a Comprehen-sive Equilibrium Equations Method with Force Densities
yk
i
j
O x
yk
yi
yj
Fi,k
Li,k
Li,j
Fi,j
π
xj xi xk
F(E)i,y
F(E)i,x
Figure 1: Equilibrium of an unconstrained node of a 2D tensegritystructure.
(CEEMFD) and in Section 3 we discuss the treatment ofredundancies and we investigate the stability in this type ofstructures. In Section 4, some applications of the methodare presented on planar and on spatial tensegrity structures.The article ends with Section 5 in which conclusions andproposals for future work are given.
2. A Comprehensive Equilibrium EquationsMethod with Force Densities
An important concept related with tensegrity structures isthe βforce densityβ [1β3]. For a member made of a materialobeying Hookeβs Law, with ends at π and π and with a lengthπΏ π,π and a longitudinal force πΉπ,π, the force density is definedas follows:
ππ,π =πΉπ,ππΏ π,π = (πΈπ΄)π,π
ππ,ππΏ π,π , (1)
where ππ,π is the strain, πΈ is the Young modulus [10], andπ΄ is the effective cross-sectional area of the member. Thisdefinition will be useful in the analysis which is presentedherewith. Force density has a negative sign in compressionand a positive sign in tension.
Figure 1 is used to exemplify the equilibrium of a typicalunconstrained node π, on the Oxy plane, which is connectedto joints π and π, through members which have lengths πΏ π,πand πΏ π,π. For the three-dimensional orthogonal Cartesiancoordinate systemππ₯π¦π§, whereππ joints are connected withjoint π, through ππ members, the equilibrium equations, inthe direction of the axes ππ₯, ππ¦, and ππ§ are given by
ππβπ=1
π₯π β π₯ππΏ π,π πΉπ,π =
ππβπ=1
(π₯π β π₯π) ππ,π
=ππβπ=1
ππ,ππ₯π βππβπ=1
ππ,ππ₯π = πΉ(πΈ)π,π₯ ,(2)
Advances in Numerical Analysis 3
ππβπ=1
π¦π β π¦ππΏ π,π πΉπ,π =
ππβπ=1
(π¦π β π¦π) ππ,π
=ππβπ=1
ππ,ππ¦π βππβπ=1
ππ,ππ¦π = πΉ(πΈ)π,π¦ ,(3)
ππβπ=1
π§π β π§ππΏ π,π πΉπ,π =
ππβπ=1
(π§π β π§π) ππ,π =ππβπ=1
ππ,ππ§π βππβπ=1
ππ,ππ§π
= πΉ(πΈ)π,π§ ,(4)
where π = 1, . . . , π and π is the total number of nodes onthe structure. In equations (2), (3), and (4) forces πΉ(πΈ)π,π₯ , πΉ(πΈ)π,π¦ ,and πΉ(πΈ)π,π§ are the external loads which include gravity loads,live loads, pretension, and even reactions in the case wherethe node is a support. Gravity loads are assumed to act alongthe negative π¦-axis, for 2D structures or along the negativeπ§-axis for 3D structures. External loads, which appear on theright hand side of (2), (3), and (4) have the following generalexpressions:
πΉ(πΈ)π,π₯ = βππβπ=1
(π₯π β π₯π) π‘π,π + π΅π,π₯ + π π,π₯,
πΉ(πΈ)π,π¦ = βππβπ=1
(π¦π β π¦π) π‘π,π + π΅π,π¦ + π π,π¦,
πΉ(πΈ)π,π§ = βπΊπ β 12ππβπ=1
π€π,ππΏ π,π βππβπ=1
(π§π β π§π) π‘π,π + π΅π,π§
+ π π,π§,
(5)
where πΊπ is the self-weight of joint π andπ€π,π is the weight perunit length of a member with nodes at π and π. Also, π‘π,π is thepretension per unit length of a member joining nodes π and π.Forces π΅π,π₯, π΅π,π¦, and π΅π,π§ are the concentrated live loads andπ π,π₯,π π,π¦, andπ π,π§ are the reaction forces (if they exist) on jointπ. Both live loads and reaction forces are assumed to be actingalong the positive directions of axes ππ₯, ππ¦, and ππ§.
The equilibrium of the whole structure is consideredby introducing the connectivity matrix π. This matrix haselements ππ,π. Index π takes values from 1 to ππ (here ππ isthe number of all members) and index π takes values from 1toπ. The elements ππ,π of matrix π take the following values[1β3]:
ππ,π ={{{{{{{{{
+1 if π is the initial node of member πβ1 if π is the final node of member π0 if π does not belong to member π.
(6)
It really does notmatter which node ofmember πwe considerfirst or last but oncewe consider a node π to be the initial nodefor the member, we also consider it as the initial node for anyother member that is connected with this node.
Also, matrix π is introduced which has elements π π,π.Index π takes values from 1 to π and index π takes values
from 1 to ππ (where ππ is the number of supports of thestructure). Elements π π,π of matrix π are then defined asfollows:π π,π
= {{{β1 if node π coincides with the support π0 if node π is a different node from support π.
(7)
By introducing vectors π₯π, π¦π, and π§π which contain the π₯-coordinates, π¦-coordinates, and π§-coordinates, respectively,of all the π nodes of the structure, the set of linear equi-librium equations for all the joints of the structure can beexpressed in block form as shown below:
π΄ β ππ = [[[[
ππ (π β π₯π)π π π π πππ (π β π¦π)π π π π πππ (π β π§π)π π π π π
]]]]β [[[[[[
ππ π₯π π¦π π§
]]]]]]
= Ξ¨(πΈ), (8)
where vector ππ contains all the force densities and reactionforces of the structure. For spatial tensegrities it has dimen-sions (ππ + 3ππ ) Γ 1. Vector π contains all the unknownforce densities ππ,π which are inserted in π according tothe ascending order of numbering of the elements of thestructure and it has dimensions ππ Γ 1. Vectors π π₯, π π¦, andπ π§ contain the reaction forces on the supports and each oneof them has dimensions ππ Γ 1. Vector Ξ¨(πΈ) contains all theknown external loads acting in the π₯, π¦, and π§-directions,respectively, on all the joints of the structure (gravity loads,live loads, pretension, etc.) and for 3D tensegrities it hasdimensions 3πΓ 1. Its componentsΞ¨(πΈ)π,π₯ ,Ξ¨(πΈ)π,π¦ , andΞ¨(πΈ)π,π§ havethe same expression as for πΉ(πΈ)π,π₯ , πΉ(πΈ)π,π¦ , and πΉ(πΈ)π,π§ in equations(2), (3), and (4), respectively, except that they do not containthe reaction forces. Matrix π΄ is the global shape matrix ofthe tensegrity and is very sparse. This means a significantlysmaller number of computations and memory space on acomputer compared with the classical finite element method.For the creation of a computer code, definitions (6) and (7)and formulation (8) are easy to use.
In the case of statically determinate structures, the set oflinear equilibriumequations (8) is the set of equations to solveto directly find the unknown values of the force densities ππ,π.Considering all theπ joints of the structure andwith the helpof matrix π, the set of equilibrium equations (2), (3), and (4),for the whole structure, takes the following block form:
[[[
π· π ππ π· ππ π π·
]]]β [[[
π₯ππ¦ππ§π]]]
= [[[[
πΉ(πΈ)π₯πΉ(πΈ)π¦πΉ(πΈ)π§
]]]]
= [[[[
(π π₯)gl + Ξ¨(πΈ)π₯(π π¦)gl + Ξ¨(πΈ)π¦(π π§)gl + Ξ¨(πΈ)π§
]]]], (9)
where (π π₯)gl, (π π¦)gl, and (π π§)gl are global vectors of dimen-sionsπΓ1which contain the unknown reaction forces. Also,matrixπ·, which is known as the force density matrix (FDM)of the tensegrity, relates the nodal coordinates and the forceswhich are acting on the structure and is given by
π· = ππ diag (π) π. (10)
4 Advances in Numerical Analysis
The elements ππ,π of matrixπ· are as follows [1β3]:
ππ,π
={{{{{{{{{{{{{{{
βππ,π when π ΜΈ= πππβπ=1(π ΜΈ=π)
ππ,π when π = π
0 if nodes π and π are not connected.
(11)
Matrixπ· is a Kirchhoff matrix and it is a symmetric positivesemidefinite matrix. It is also called discrete matrix or com-binatorial Laplacian or admittance matrix and it containselements with a positive or a negative sign [8]. It has dimen-sionsπ Γπ. A more general form of (9) is the following:
(πΌ β π·) β [[[[
π₯π β π·β1 (π π₯)glπ¦π β π·β1 (π π¦)glπ§π β π·β1 (π π§)gl
]]]]
= [[[[
Ξ¨(πΈ)π₯Ξ¨(πΈ)π¦Ξ¨(πΈ)π§
]]]]
= Ξ¨(πΈ), (12)
where πΌ is the 3 Γ 3 unit matrix. So, equations (8) and (12)give the following elegant and important, at the same time,relation:
(πΌ β π·) β [[[[
π₯π β π·β1 (π π₯)glπ¦π β π·β1 (π π¦)glπ§π β π·β1 (π π§)gl
]]]]β π΄ β ππ = 0. (13)
One may observe that equation (13) does not contain anyvalues of the external loads except from the reaction forceson the supports.
3. Redundancies, Numerical Representation ofCastiglianoβs Second Theorem, and Stability
The values of the elements of matrix π΄ in (8) depend directlyon the values of the coordinates of the nodes. For planarproblems it has dimensions 2π Γ (ππ + 2ππ ) and for spatialstructures it is of dimensions 3π Γ (ππ + 3ππ ). For thestatically determinate structures [10] and after inserting inthe set of equations (8) the known values of reactions onthe supports, matrix π΄ is reduced to a square matrix Kπ ofdimensions 2πΓ2π, in two dimensions and 3πΓ3π in threedimensions. Its properties, as a square matrix, give importantinformation about the stability of the form of the structureunder study. So, for statically determinate structures, the setof linear equilibrium equations (8) has a solution if and onlyif det |Kπ| ΜΈ= 0. Other equivalent conditions concerningsquare matrix Kπ, which secure the existence of a solutionare given in many references about numerical analysis (e.g.,[11]). If matrix Kπ has a determinant equal to zero then, mostprobably, the structure is unstable.
For the 2D case, the degree of redundancy is ππ = ππ +πur β 2π and for the 3D problems it is ππ = ππ + πur β3π, where πur is the number of unknown reaction forceson the supports. We say that the system of linear equations
has ππ parametric solutions. Thus, we need an additionalnumber of ππ equations which, in this study, are providedby Castiglianoβs second theorem. According to this theorem,which is well presented in many classical books about solidmechanics (e.g., Zhang and Ohsaki [9]), all the forces inthe bars or strings of the structure are expressed in termsof any ππ, in number, forces π(π), which are consideredas redundancies and are arbitrarily chosen among the ππforces ππ, acting as internal loads along the members of thestructure. It is assumed that no local mechanisms are formed.Then, Castiglianoβs second theorem gives the additional ππequations which are
ππππ(π) =
ππβπ=1
ππ(πΈπ΄)π β
πππππ(π) πΏ π = 0, π = 1, . . . , ππ, (14)
whereπ is the total potential energy of the structure. Internalforces ππ are then expressed as
ππ = π(RL)π +ππβπ=1
π(π)π π(π), π = 1, . . . , π, (15)
where π(π) are the unknown internal redundant forces. In(15) forces π(RL)π represent a set of forces acting along eachmember π and being in equilibrium with all other real forcesacting on the same node. This set of forces is created byremoving the redundancies and solving the resulting stati-cally determinate structure (which is now called fundamentalstructure) under the action of its real loading. Then, solutiongives the values of π(RL)π . Also, in (15) forces π(π)π represent aset of forces in static equilibriumwhich appears when no realloading is applied on the fundamental structure and whenthe redundancy members (one at a time) are replaced witha pair of unit forces, opposing each other and acting alongthe memberβs axis. One such pair of forces is acting eachtime and for each pair the fundamental structure is solved.Forces π(π)π are then considered as quantities without theunit of force. They are used as the coefficients of π(π) andtheir values constitute a group of statically acceptable internalforces, acting along the redundantmembers.Then, the partialderivative of ππ with respect to π(π) in (14) is expressed as
πππππ(π) = π(π)π , π = 1, . . . , ππ, π = 1, . . . , ππ. (16)
Using (15) and (16), equations (14) take the form
ππβπ=1
π(RL)π + βπππ=1 π(π)π π(π)(πΈπ΄)π β π(π)π πΏ π = 0
or πΏ(RL)π +ππβπ=1
βπππ(π) = 0,
π = 1, . . . , ππ,
(17)
Advances in Numerical Analysis 5
where
πΏ(RL)π =ππβπ=1
π(RL)π π(π)π(πΈπ΄)π πΏ π,
βππ =ππβπ=1
π(π)π π(π)π(πΈπ΄)π πΏ π.
(18)
If there is a real relative displacement πΏ(π )π
between one end ofthe redundant member π and the joint at the other end, thenequation (17) does not have a zero on the right-hand side andit takes the form
πΏ(π )π = πΏ(RL)π +ππβπ=1
βπππ(π), π = 1, . . . , ππ. (19)
Equations (19) can be written in matrix form as follows:
[[[[[
πΏ(π )1...
πΏ(π )ππ
]]]]]
=[[[[[
πΏ(RL)1...
πΏ(π )ππ
]]]]]+[[[[[
β1,π β β β β1,ππ... ... ...
β1,ππ β β β βππ ,ππ
]]]]]β [[[[[
π(1)...
π(ππ)
]]]]], (20)
where π(π) is also expressed in terms of force density π(π) asπ(π) = π(π)πΏ(π). So, equilibrium equations (8) or (12) togetherwith (20) make a system of 2π + ππ equations, in the caseof planar problems or 3π + ππ equations, in the case of3D problems, which is solved to give the values of all theunknown forces on the members of the structure. Anotherway to find the unknown internal forces of the structure is bysolving separately the system of linear equations (20) forπ(π).Next, by substituting the known values ofπ(π) in (8) or (12) wemayfind the values of all other forcesππ of themembers of thetensegrity structure. However, it is preferable to consider thecomplete system of 2π+ππ or 3π+ππ equilibrium equationsin order to have the opportunity to investigate the stability ofthe whole structure.
The existence of solution, for the resulting systemof linearequationsKπ β π½ = πΎ, is an indication of the stability conditionof the structure. If the solution exists then it is unique. Onecan be certain about the existence of the solutionwhen squarematrix Kπ has one of the following properties [11]:
(1) Matrix Kπ has a rank π equal to the rank of theaugmented matrix (Kπ/πΎ).
(2) The determinant of the square matrix Kπ is not equalto zero; that is, det |Kπ| ΜΈ= 0.
(3) The inverse Kβ1π of the square matrix Kπ exists andgives Kβ1π Kπ = πΌ.
(4) The number πind of linearly independent rows orcolumns of matrix Kπ is equal to π.
If matrix Kπ does not have one of the above equivalentproperties then the solution does not exist and we say thatthe stability of our tensegrity structure is doubtful.
4. Numerical Examples
4.1. The X-Shape 2D Tensegrity Structure. One of the fun-damental tensegrity configurations, used by Fuller [4] andSnelson [6] to create more complicated stable structures,is the X-shape 2D tensegrity truss with π = 4 nodes(Figure 2(a)). Skelton [12] studied analytically this trusswhich has four steel cables (elements 1, 2, 5, and 6) and tworods made of an aluminum alloy (members 3 and 4). Thetechnical characteristics of the cables and rods are presentedin Table 1. The weight of each joint is πΊ = 0.01 kN. No otherexternal loads are applied. Cable 1 in our example is shorterthan the required length. During construction one end ofthe cable is connected to joint 1 and the other is extendedby πΏ2,π = 0.3mm to meet joint 2, as it is explained in theexaggerated detail βπ΄β (Figure 2(b)), which is not to scale.
The complete set of linear equilibrium equations (8), forthis problem, consists of 2π = 8 equations. It hasππ +πur β2π = 1 parametric solutions. We say that the structure hasa redundancy equal to 1. Rod 1 is chosen as the redundantmember of the structure (Figure 2(a)). Using expression (19)with πΏ(π )1 = 0.3mm and considering that π(1) = πΏ1,2π1,2 weobtain the additional linear equation β0.0346 + 7.3246π1,2 =3.16512 which gives π1,2 = 0.4368 kN/m. Thus, the valueof the force on this element is πΉ1,2 = 0.4368 kN. Also, thisequation is added to equations (12) to give a complete systemof linear equations. Expanded matrix Kπ of the final systemof equations has a nonzero determinant. The results for allforces on the members and the reactions on the supports aretabulated in Table 2 together with those obtained with theDRM. The CEEMFD and the DRM give results which arecomparable. However, the CEEMFD is proved to be fasterthan the DRM (Table 2) because it is a direct method (noiterations are necessary). Also, Table 3 presents the value ofthe vertical displacement πΏtop,π¦ of joint 3, obtained by themethod, the DRM and the analytic approach proposed bySkelton [12] for the same problem. The results for πΏtop,π¦,with all three methods, are the same. By comparing theresults obtained in [2], for an analogous problem of the samegeometry, with those obtained in the present work (Tables2 and 3), one may observe that the numerical results in thecurrent work have values which are almost equal to 10% ofthe corresponding values obtained in solving the analogousproblem proposed in [2]. This difference was expected dueto the linearity existing in both problems and due to the factthat the values of πΏ(π )1 , in the two cases, differ by 90%. Thematerials used, in both cases, had slightly different properties.It is also verified that the structure passes axial yield and Eulerbuckling criteria.
4.2. A Weightless 2D Tensegrity Structure. Zhang and Ohsaki[9] use the Adaptive Force Density Method (AFDM) to solvea weightless two-dimensional tensegrity structure which iswithout any support and out of any gravitational field (Fig-ure 3). This structure has four cables (elements 1, 2, 3, and4) and four struts (members 5, 6, 7, and 8). The steel cablesand the aluminum rods have the same stiffness which in thecurrent work is chosen to be (πΈπ΄)π = (πΈπ΄)π = 12694 kN,
6 Advances in Numerical Analysis
Y
X1
3 4
2
2
1
5
6
3 4
βAβ
1m
1m
(a)
X
2
DETAIL βAβ
35
1
πΏ2,π = 0.3mm
(b)
Figure 2: (a) The X-shape 2D tensegrity truss. (b) Detail βπ΄β of joint 2.
Table 1: Technical characteristics of the members of the X-shape 2D tensegrity.
Type of member Outer diameterππ (mm)
Inner diameterππ (mm)
Cross-sectionalarea π΄ (mm2)
Moment ofinertia πΌ0 (m4)
Specific weight πΎ(kN/m3)
Youngβsmodulus πΈ(GPa)
Yield stressππ¦ (MPa)
Tensionmembers (steelcables)
8 β 50.24 β 78 210 360
Compressionmembers(aluminumrods)
40 37 181.34 3.36 Γ 10β8 26.50 70 260
without changing the generality of the problem. Also, thealuminum rods have a hollow cross section with a momentof inertia, about any diameter, equal to πΌπ = 3.36 Γ 10β8m4.The only load on the structure is the pretension on cable1 which is equal to 1 kN (Figure 3). In order to use theCEEMFD we introduce, temporarily, two auxiliary supportsat joints 2 and 5 on which there are no reactions (it isas if they do not exist). These supports will only help todefine the degree of redundancy. The complete set of linearequilibrium equations (12), for this problem, consists of 2π =10 equations. It hasππ +πur β 2π = 1 parametric solutions.Thus, the structure has a redundancy equal to 1. Cable 1 ischosen as the redundant member. In using the CEEMFDthe additional linear equation, which is required in order tofind the unknowns, is β6.8284 + 6.8284π1,2 = 0, from whichone obtains π1,2 = 1 kN/m. From the form of matrix Kπit is verified that the structure is stable (i.e., |Kπ| ΜΈ= 0).The same problem is also solved with the DRM and theAFDM [9]. The results are presented in Table 4. One can seethat the CEEMFD and the AFDM [9] give results which arecomparable and more accurate than those obtained with theDRM. Exactly the same results were obtained and the sameobservation was made in [2] for an analogous problem of thesame geometry but with different material properties. Thus,for this specific problem the results are independent of the
2 1
5
3
4
1
47
2
3
8
5 6
Figure 3: A weightless two-dimensional tensegrity truss.
material properties. It is also verified that the structure passesaxial yield and Euler buckling criteria.
4.3. A Cantilever 2D Tensegrity Beam. In this subsection, aproblem of a cantilever planar tensegrity beam is investigated(Figure 4). This structure is supported at the two leftmost
Advances in Numerical Analysis 7
Table 2: Values of the forces on the members and of the reaction forces on the supports for the X-shape 2D tensegrity structure.
Forcedensity ππ,π
Value ofππ,π withCEEMFD(kN/m)
Value ofππ,π withDRM(kN/m)
Length πΏ π,πof a
member(m)
Value offorce πΉπ,πwith
CEEMFD(kN)
Value offorce πΉπ,πwith DRM
(kN)
Reactionforce on asupport
Value ofreactionwith
CEEMFD(kN)
Value ofreactionwith DRM
(kN)
CPU timewith
CEEMFD(sec)
CPU timewith DRM
(sec)
π1,2 0.4368 0.4367 1.0000 0.4368 0.4367 π 1,π₯ 0 0 0.02 2.03π1,3 0.4195 0.4196 1.0000 0.4195 0.4196 π 2,π₯ 0 0π1,4 β0.4369 β0.4368 1.4142 β0.6178 β0.6177 π 1,π¦ 0.0346 0.0345π2,3 β0.4369 β0.4368 1.4142 β0.6178 β0.6177 π 2,π¦ 0.0346 0.0345π2,4 0.4195 0.4196 1.0000 0.4195 0.4196π3,4 0.4368 0.4367 1.0000 0.4368 0.4367
Table 3: Comparison between the results of the CEEMFD and thoseof the DRM and Skeltonβs analytic approach.
CEEMFD DRM Skeltonβs analyticapproach [12]
Verticaldisplace-ment ofjoint 3 (inmm)
0.04 0.04 0.04
nodes 1 and 3 and loaded with a unit vertical force acting atthe top-right node 10 (Figure 4). This type of structure waswell investigated by Masic et al. [13] who used a nonlinearlarge displacement model to find its static response and tomake a design with optimal mass-to-stiffness ratio. In thepresent work, no optimization was made. The CEEMFD wassimply implemented on this problem, by considering thesame geometry proposed byMasic et al. [13] and by arbitrarilychoosing the material properties for the rods and the cables.So, this structure has thirteen steel cables (elements 1, 2,3, 4, 5, 8, 9, 12, 13, 16, 17, 20, and 21) and eight aluminumstruts (members 6, 7, 10, 11, 14, 15, 18, and 19). The technicalcharacteristics of the cables and rods are given in Table 5.According to the method, along each one of cables 1, 2, 3,and 4 a pair of opposite axial forces, of value 1 kN, is actingeach time. Each one of these forces is pushing an end node.Also, a gap exists between the right-end on each one of thesemembers and its nearest joint. This gap is the same for all thefourmembers and is equal to 54.2640/(πΈπ΄)π or 10mm (0.4 in.approx.).The truss is a statically indeterminate structure witha degree of redundancyππ = ππ + πur β 2π = 4.
Tensions in cables 1, 2, 3, and 4 are chosen as redun-dancies. Matrix Kπ is well-conditioned. The results withCEEMFD are presented in Table 6 together with those ofthe DRM which was also implemented to this problem. Thetwomethods give results which are comparable. Also, Table 7gives the values of the vertical displacement πΏ(π)π¦ at joints 4,6, 8, and 10 obtained by the method and the DRM. The twosets of results are again comparable. One may visualize thedeformed shape of this structure by observing the graph inFigure 5, which presents the displacements along its upper
side. It is also verified that the structure passes axial yield andEuler buckling criteria.
4.4. A One-Stage 3D Tensegrity Structure. One of the classicaltensegrity structures is the one-stage 3D tensegrity structure(Figure 6) which contains a very basic 2D configuration: theX-shape 2D tensegrity truss. This 3D structure can be mademuch more complicated by adding more bars and cables andmore stages.
In the present work, the X-shape planar truss is formedby bars 5 and 6 and cables 3, 8, and 12, as shown in Figure 6.This stable planar structure is connected with bar 4 with theaid of cables 1, 2, 7, 9, 10, and 11, to create a stable three-dimensional structure. All elements aremade of the samepre-cious aluminum alloy. Linear elastic behavior is consideredfor loads below the yield stress. The technical characteristicsof all members are shown in Table 8. Also, the weight ofeach joint is πΊ = 0.01 kN. Cables 1, 2, 3, 10, 11, and 12have a constant pretension equal to 0.01 kN/m along theirlength. The coordinates of all the 6 nodes of the structureare shown on Table 9. In applying the method, the set ofequilibrium equations (8) is obtained which has 3π = 18independent linear equations. There are ππ + πur = 12 +3 Γ 3 = 21 unknowns indicating a degree of redundancy eq-ual to 3. However, since tensions in cables 1, 2, and 3 have noeffect on the values of force densities of the other members,one has π1,2 = π1,3 = π2,3 = 0.01 kN/m. Then, the number ofthe unknowns is reduced to 18. Thus, the system of 18linear equations is sufficient to give the solution to our pro-blem. Matrix Kπ of the linear system of equations has all theproperties given by statements (1) to (4) of Section 3, veri-fying that our tensegrity structure is stable as expected.
The values of the reaction forces on the supports and ofall the forces and the corresponding force densities on themembers of the structure are presented in Table 10. Also, itis verified that for this 3D problem the structure passes axialyield and Euler buckling criteria.
4.5. A Two-Stage Self-Stressed 3D Tensegrity Structure. Zhangand Ohsaki [9] have also solved the problem of a two-stageself-stressed 3D tensegrity structure (Figure 7) by using theAFDM. In this example, the structure is considered to be
8 Advances in Numerical Analysis
Table 4: Values of the forces on the members for the weightless 2D tensegrity structure.
Force densityππ,π
Value of ππ,π withCEEMFD(kN/m)
Value of ππ,π withAFDM [9](kN/m)
Value of ππ,π withDRM (kN/m)
Length πΏ π,π of amember (m)
Value of πΉπ,πwith methodCEEMFD (kN)
Value of πΉπ,πwith method
AFDM [9] (kN)
Value of forceπΉπ,π with
DRM (kN)π1,2 1.00000000 1.00000000 1.00000072 1.00000000 1.00000000 1.00000000 1.00000072π1,3 1.00000000 1.00000000 1.00000072 1.00000000 1.00000000 1.00000000 1.00000072π1,4 1.00000000 1.00000000 1.00000072 1.00000000 1.00000000 1.00000000 1.00000072π1,5 1.00000000 1.00000000 1.00000072 1.00000000 1.00000000 1.00000000 1.00000072π2,4 β0.50000000 β0.50000000 β0.50000036 1.41421356 β0.70710678 β0.70710678 β0.70710729π2,5 β0.50000000 β0.50000000 β0.50000036 1.41421356 β0.70710678 β0.70710678 β0.70710729π3,4 β0.50000000 β0.50000000 β0.50000036 1.41421356 β0.70710678 β0.70710678 β0.70710729π3,5 β0.50000000 β0.50000000 β0.50000036 1.41421356 β0.70710678 β0.70710678 β0.70710729
Table 5: Technical characteristics of the members of the cantilever 2D tensegrity beam.
Type of member Outer diameterππ (mm)
Inner diameterππ (mm)
Cross sec. areaπ΄ (mm2)
Moment ofinertia πΌ0 (m4)
Specific weight πΎ(kN/m3)
Youngβsmodulus πΈ(GPa)
Yield stressππ¦ (MPa)
Tensionmembers (steelcables)
5.74 β 25.84 β 78 210 480
Compressionmembers (alum.rods)
77 75 238.76 1.72 Γ 10β7 26.50 70 260
1 1 2 3 4
9 13 17 21
5 8 12 16 20
6
7
10
11
14
15
18
19
2
3
4
5
6
7
8
9
10
2m
2.5m 2.5m 2.5m 2.5m
P = 1kN
Figure 4: A cantilever planar tensegrity beam [13].
0 2,5 5 7,5 10
x (m)
u(c
m)
0
2
4
6
8
10
12
14
16
18
Figure 5: Vertical displacements along the upper side of the tensegrity beam.
Advances in Numerical Analysis 9
Table 6: Values of the forces on the members for the cantilever 2D tensegrity beam.
Value of ππ,π withCEEMFD(kN/m)
CPU time =0.07 sec
Value of ππ,π withDRM (kN/m)CPU time =8.25 sec
Length πΏ π,π of amember (m)
Value of forceπΉπ,π with
CEEMFD (kN)
Value of forceπΉπ,π with DRM
(kN)
Value of stressππ,π withCEEMFD(MPa)
Force densityππ,π
π1,2 0.240633 0.240613 2.5000 0.601583 0.601533 23.2811π2,5 0.651541 0.651521 2.5000 1.628852 1.628802 63.0361π5,7 1.196278 1.196278 2.5000 2.990696 2.990696 115.7390π7,9 1.966444 1.966444 2.5000 4.916110 4.916110 190.2519π1,3 2.461283 2.461415 2.0000 4.922566 4.922830 190.5018π2,3 β1.861511 β1.861511 3.2016 β5.959814 β5.959814 β24.9615π1,4 β2.454204 β2.454336 3.2016 β7.857380 β7.857803 β32.9091π2,4 4.147074 4.147074 2.0000 8.294147 8.294147 320.9809π3,4 4.075133 4.075096 2.5000 10.187833 10.187740 394.2660π4,5 β1.705920 β1.705788 3.2016 β5.461675 β5.461252 β22.8752π2,6 β2.272414 β2.272414 3.2016 β7.275360 β7.275360 β30.4714π5,6 3.969719 3.969475 2.0000 7.939438 7.938949 307.2538π4,6 3.326840 3.326539 2.5000 8.317101 8.316348 321.8692π6,7 β1.710358 β1.710114 3.2016 β5.475883 β5.475101 β22.9347π5,8 β2.250652 β2.250539 3.2016 β7.205687 β7.205327 β30.1796π7,8 4.204022 4.203534 2.0000 8.408044 8.407068 325.3887π6,8 2.764778 2.764233 2.5000 6.911946 6.910582 267.4902π8,9 β1.966421 β1.966045 3.2016 β6.295692 β6.294490 β26.3683π7,10 β2.480514 β2.480270 3.2016 β7.941615 β7.940834 β33.2619π9,10 1.973494 1.973118 2.0000 3.946987 3.946236 152.7472π8,10 2.480544 2.479735 2.5000 6.201359 6.199338 239.9907
Reaction (kN)π (1)π» 5.534000 5.534380π (3)π» β5.534000 β5.533907π (3)V 1.213600 1.213864
Table 7: Displacements with the CEEMFD and the DRM along the tensegrity beam.
Joint π 1 4 6 8 10Vertical displacement of joint π withCEEMFD 0.0000 1.0091 cm or 0.40 in. 4.1011 cm or 1.61 in. 9.0438 cm or 3.56 in. 15.7700 cm or 6.21 in.
Vertical displacement of joint π withDRM 0.0000 1.0091 cm or 0.40 in. 4.1012 cm or 1.61 in. 9.0467 cm or 3.56 in. 15.7779 cm or 6.21 in.
Table 8: Technical characteristics of the one-stage 3D tensegrity structure.
Type of member Outer diameterππ (mm)
Inner diameterππ (mm)
Cross-sectionalarea π΄ (mm2)
Moment ofinertia πΌ0 (m4)
Specific weight πΎ(kN/m3)
Youngβsmodulus πΈ(GPa)
Yield stressππ¦ (MPa)
Tension members(aluminium cables) 20 β 314.16 β 31.85 89.17 260
Compressionelements 4 & 6(aluminum rods)
63.9 61.7 214.87 1.07 Γ 10β7 31.85 89.17 260
Compressionelement 5(aluminum rod)
58.4 54.9 311.45 1.25 Γ 10β7 31.85 89.17 260
10 Advances in Numerical Analysis
2
3
6
5
21
3
XZ
1
4
7
8
9
4
5
6
10
1211
Y
4.32m
Figure 6: A one-stage 3D tensegrity structure.
Top
Vertical
d
Saddle
c Diagonal
a Bottom b
3.33
m
Figure 7: A two-stage 3D tensegrity structure (perspective view).
Table 9: Values of the coordinates of the joints of the one-stage 3Dtensegrity structure.
Node π π₯π (m) π¦π (m) π§π (m)1 3.0 β1.73205 0.02 0.0 3.46410 0.03 β3.0 β1.73205 0.04 β4.618802 0.0 4.320495 β0.479058 β0.82975 4.320496 2.309401 4.0 4.32049
weightless and is composed of 12 nodes and 30members; thatis,π = 12 andππ = 30.
The structure contains 24 steel cables and 6 aluminumrods. The 24 cables are divided into four groups: (i) cables
of the top and bottom bases, (ii) saddle cables, (iii) ver-tical cables, and (iv) diagonal cables, as indicated clearlyin Figure 7. Its six rods are divided in two groups: (1)rods of the upper stage; (2) rods of the lower stage. Thus,in total, we have six groups of elements. Since the initialforce densities and independent nodal coordinates can bearbitrarily specified, one can have some control over the geo-metrical and mechanical properties of the structure. Thus,the steel cables and the aluminum rods have equal stiffness(πΈπ΄)π = (πΈπ΄)π = 5426 kN. Also, the aluminum rods have ahollow cross section with a moment of inertia, about any dia-meter, equal to πΌπ = 2.894 Γ 10β9m4. The only loading onthe structure is the initial set of force densities π(π) applied incables of all the six groups; that is, for the two groups of rods itis π(π)π = β1.5 kN/m, for the saddle cables is π(π)π = 2.0 kN/m,and for all the other cables is π(π)π = 1.0 kN/m.
By specifying the coordinates of nodes π, π, and π, whichare shown in Figure 7, to make the bottom base located onthe π₯π¦-plane, and node π in the lower stage, we can have theconfiguration of the tensegrity structure as shown in Figure 7.The coordinates of these nodes are shown in Table 11. Then,the CEEMFDgives the set of final values of force densities, forthe six groups of elements, which is listed in Table 12. Thesevalues are compared with those obtained with the AFDM,after 158 iterations [9]. The two sets of solutions comparefavorably but in CEEMFD the solution is obtained directly.It is verified that the structure passes axial yield and Eulerbuckling criteria.
5. Conclusions
We have presented the CEEMFD to analyze 2D and 3D ten-segrity structures.The resulting final linear system of equilib-rium equations can be directly solved to give a unique solu-tion for the force densities on the elements of the structure.For stable statically determinate structures, the global matrix
Advances in Numerical Analysis 11
Table 10: Values of the forces on the members of the one-stage 3D structure.
Force densityon a member
Value of ππ,πwith
CEEMFD(kN/m)
Value of ππ,πwith DRM(kN/m)
Length πΏ π,π ofa member
(m)
Value of πΉπ,πwith
CEEMFD(kN)
Value of πΉπ,πwith DRM
(kN)
Reactionforce on asupport
Value ofreaction force
withCEEMFD
(kN)
Value ofreaction forcewith DRM
(kN)
π1,2 0.010000 0.010000 6.000000 0.060000 0.060000 π 1,π₯ 0.321956 0.322034π1,3 0.010000 0.010000 6.000000 0.060000 0.060000 π 2,π₯ β0.207729 β0.207787π1,4 β0.093977 β0.093998 8.928200 β0.839048 β0.839231 π 3,π₯ 0.529678 0.529822π1,5 0.077393 0.077414 5.620020 0.434951 0.435068 π 1,π¦ β0.031957 β0.031974π2,3 0.010000 0.010000 6.000000 0.060000 0.060000 π 2,π¦ 0.423712 0.423814π2,5 β0.114623 β0.114644 6.110100 β0.700357 β0.700484 π 3,π¦ 0.391755 0.391837π2,6 0.056172 0.056193 4.928199 0.276828 0.276931 π 1,π§ 0.157097 0.157097π3,4 0.054744 0.054765 4.928199 0.269789 0.269892 π 2,π§ 0.334522 0.334524π3,6 β0.096973 β0.096994 8.928200 β0.865798 β0.865984 π 3,π§ 0.264438 0.264437π4,5 0.125318 0.125340 4.222079 0.529100 0.529194π4,6 0.013337 0.013342 8.000000 0.106699 0.106736π5,6 0.096702 0.096720 5.576919 0.539297 0.539401
Table 11: Specified nodal coordinates in the two-stage 3D tensegrity structure.
Node π π₯π (m) π¦π (m) π§π (m)π β2.6667 0.0000 0.0000π 1.3333 β2.3094 0.0000π 1.3334 2.3094 0.0000π β1.8867 1.6666 3.3333
Table 12: Force densities in the members of the two-stage 3D tensegrity structure.
Groupβ Rods of the upperstage (1)
Rods of the lowerstage (2)
Cables (top &bottom) (i) Saddle cables (ii) Vertical cables (iii) Diagonal cables
(iv)Initial value(kN/m) β1.5000 β1.5000 1.0000 2.0000 1.0000 1.0000
Final value(CEEMFD) β1.8376 β1.8376 0.9282 1.9920 1.1735 0.9957
Final value(AFDM) [9] β1.8376 β1.8376 0.9281 1.9918 1.1737 0.9958
Kπ of the final system of equilibrium equations is squareand has a nonzero determinant. For statically indeterminatestructures, this matrix is not square. It is expanded toa square matrix after the implementation of Castiglianoβstheorem which gives the additional equations required forthe solution. The partial derivatives, which appear in theadditional equations, are replaced by statically acceptableinternal forces which are applied on the structure. For stablestructures, the complete system of equations is then solved togive the values of the force densities on the members of thestructure. Five problems of both planar and spatial tensegritystructures were solved. The results compare favorably withthose obtained with the DRM and the AFDM [9] but theCEEMFD is faster as a direct method. As for future work weconsider the extension of themethod for geodesic domes andtensegrity bridges for pedestrians.
Competing Interests
The authors declare that there is no conflict of interestsregarding the publication of this article.
References
[1] P. Christou, A. Michael, and M. Elliotis, βImplementing slackcables in the force density method,β Engineering Computations,vol. 31, no. 5, pp. 1011β1030, 2014.
[2] M. Elliotis, P. Christou, and A. Michael, βAnalysis of planar andspatial tensegrity structures with redundancies, by implement-ing a comprehensive equilibrium equations method,β Interna-tional Journal of Engineering Science and Technology, vol. 5, pp.40β55, 2016.
12 Advances in Numerical Analysis
[3] H.-J. Schek, βThe force density method for form findingand computation of general networks,β Computer Methods inApplied Mechanics and Engineering, vol. 3, no. 1, pp. 115β134,1974.
[4] R. B. Fuller, Synergetics. Explorations in the Geometry of Think-ing, Macmillan, New York, NY, USA, 1975.
[5] D. Emmerich, βConstructions de reseaux autodendantes,βPatent no. 1, 377, 290, 1963.
[6] K. Snelson, βContinuous tension, discontinuous compressionstructures,β US Patent No. 3, 169, 611, 1965.
[7] M. Papadrakakis, βAmethod for the automatic evaluation of thedynamic relaxation parameters,β Computer Methods in AppliedMechanics and Engineering, vol. 25, no. 1, pp. 35β48, 1981.
[8] S. H. Juan and J. M. Mirats Tur, βTensegrity frameworks: staticanalysis review,βMechanism and Machine Theory, vol. 43, no. 7,pp. 859β881, 2008.
[9] J. Y. Zhang and M. Ohsaki, βAdaptive force density method forform-finding problem of tensegrity structures,β InternationalJournal of Solids and Structures, vol. 43, no. 18-19, pp. 5658β5673,2006.
[10] Y. C. Fung, Foundations of Solid Mechanics, Prentice Hall, NewJersey, NJ, USA, 1977.
[11] T. Sauer, Numerical Analysis, Pearson Education, Upper SaddleRiver, NJ, USA, 2005.
[12] R. E. Skelton, Dynamics and Control of Aerospace Systems, CRCPress, New York, NY, USA, 2002.
[13] M. Masic, R. E. Skelton, and P. E. Gill, βAlgebraic tensegrityform-finding,β International Journal of Solids and Structures, vol.42, no. 16-17, pp. 4833β4858, 2005.
Submit your manuscripts athttp://www.hindawi.com
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttp://www.hindawi.com
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Journal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
CombinatoricsHindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
International Journal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
The Scientific World JournalHindawi Publishing Corporation http://www.hindawi.com Volume 2014
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com
Volume 2014 Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Stochastic AnalysisInternational Journal of