Duke University Department of Civil and Environmental Engineering CEE 421L. Matrix Structural Analysis Henri P.

Gavin Fall, 2012

Review of Strain Energy Methods and Introduction to Stiness Matrix Methods of Structural Analysis
1 Strain Energy
Strain energy is stored within an elastic solid when the solid is deformed under load. In the absence of energy losses, such as from friction, damping or yielding, the strain energy is equal to the work done on the solid by external loads. Strain energy is a type of potential energy. Consider the work done on an elastic solid by a single point force F . When the elastic solid carries the load, F , it deforms with strains ( and ) and the material is stressed ( and ). D is a displacement in the same location and in the same direction as a point force, F . D and F are colocated. The work done by the force F on the elastic solid is the area under the force vs. displacement curve. W = F dD (1)

This work is stored as strain energy U within the elastic solid. U= 1 2 (xx
xx

+ yy

yy

+ zz

zz

+ xy xy + xz xz + yz yz ) dV.

(2)

This is a very general expression for the strain energy, U , and is not very practical for structural elements like bars, beams, trusses, or frames.

2 1.1 Bars

CEE 421L. Matrix Structural Analysis Duke University Fall 2012 H.P. Gavin

For a bar in tension or compression, we have internal axial force, N , only,

1111 0000 1111 0000 1111 0000 1111 0000 1111 0000 1111 0000 1111 0000 1111 0000 1111 0000 1111 0000 1111 0000 1111 0000 1111 0000 1111 0000 1111 0000 1111 0000

Nx

Nx

dl

xx dl

xx

so yy = 0, zz = 0, xy = 0, xz = 0, and yz = 0, and U= where xx = N/A and

xx

1 2

xx

xx

dV ,

= N/EA. Substituting dV = A dx we get U= 1 2 N (x)2 dx , E(x) A(x) (3)

and if N , E, and A are constant 1 N2 L . 2 EA Alternatively, we may express the strain as a function of the displacements along the bar ux (x), xx = ux (x)/x, and xx = E ux (x)/x. Again substituting dV = A dx , U= 1 U= 2 ux (x) E(x) A(x) x L
2

dx ,

(4)

and if E, A and ux /x = (u2 u1 )/L are constants, U= 1 EA (u2 u1 )2 2 L

Strain Energy and Matrix Methods of Structural Analysis

1.2 Beams For a beam in bending we have internal bending moments, M , and internal shear forces, V . For slender beams the eects of shear deformation are usually neglected.

v" dl

M zz

M zz

1111 0000 1111 0000 1111 0000 1111 0000 1111 0000 1111 0000 1111 0000 1111111 0000000 1111111 0000000 111 000 111 000 111 000 111 000 x 111 000 111 000 111 000

xx

dl
As in the axially loaded bar, yy = 0, zz = 0, xy = 0, xz = 0, and yz = 0, and U= For bending, xx = M y/I and
xx

1 2

xx

xx

dV .

= M y/EI. Substituting dV = dA dx, 1 2 M (x)2 y 2 dA dx , E(x) I(x)2 M (x)2 dx . E(x) I(x)

U= where
A

y 2 dA = I, so 1 U= 2 (5)
L

Alternatively, we may express the moment in terms of the curvature of the beam, 2 uy /x2 , 2 uy (x) , M (x) = E(x) I(x) x2 from which xx = E ( 2 uy /x2 ) y and xx = ( 2 uy /x2 ) y, so that 1 U= 2 where, again,
A

2 uy (x) E(x) x2 A

y 2 dA dx

y 2 dA = I, so U= 1 2 E(x) I(x)
L

2 uy (x) x2

dx .

(6)

4 1.3 Summary

CEE 421L. Matrix Structural Analysis Duke University Fall 2012 H.P. Gavin

External work is done by a set of forces, Fi , on a linear elastic solid, producing a set of displacements, Di , in the same locations and directions.
Fi Di Fj Dj Fn Dn
111 000 111 000 111 000 111 000 111 000 111 000 111 000 111 000 111 000 11111 00000 11111 00000 11111 00000 11111 00000

F 1 D1

Dn Fn

F1 D
1
11111 00000 11111 00000 11111 00000

D1
11111 00000 11111 00000 11111 00000 11111 00000

F 1
11111 00000 11111 00000 11111 00000

Di Fi Fi Di

Dj Fj Fj Dj

11111 00000 11111 00000 11111 00000 11111 00000

Fn Dn
11111 00000 11111 00000 11111 00000

The work done by these forces is 1 1 1 W = F1 D1 + F2 D2 + F3 D3 + 2 2 2 The external forces are resisted by internal moments, M , and axial forces, N . The total strain energy stored within the solid is U= 1 2
L

M2 1 dx + EI 2

Nj2 Lj Ej Aj

(7)

where the rst term is the integral over all lengths of all the beams and the second term is the sum over all the bars. If torsion and shear are included, then two additional terms are 1 2
L

T2 dx GJ

and

1 2

V2 dx . G A/

Alternatively, we can think of external forces producing curvatures ( 2 uy /x2 ) by bending, and axial stretches (ux /x). In this case U= 1 2 EI
L

2 uy x2

dx +

1 2

Ej Aj (u2j u1j )2 Lj
2

(8)

If torsion and shear are included, then two additional terms are 1 2 GJ
L

ux x

dx ,

and

1 2

uy G A/ x L

dx ,

where ux is the torsional rotation about the x-axis, ux /x is the torsional shear strain, x , (on the face perpendicular to the x-axis and in the -direction) and uy /x is the shear strain, xy , (on the face perpendicular to the x-axis and in the y-direction). Analyses using expressions of the form of equations (3), (5), or (7) are called force method or exibility method analyses. Analyses using expressions of the form of equations (4), (6), or (8) are called displacement method or stiness method analyses.

Strain Energy and Matrix Methods of Structural Analysis

2 Castiglianos Theorems
2.1 Castiglianos Theorem - Part I U= F dD ... strain energy

U(D)
0

1111111 0000000 1111111 0000000 1111111 0000000 1111111 0000000 1111111 0000000 1111111 0000000 1111111 0000000 1111111 0000000 1111111 0000000 1111111 0000000 1111111 0000000 U 1111111 0000000 1111111 0000000 1111111 0000000
Dj

D+ D j j

U U = Di Di A force, Fi , on an elastic solid is equal to the derivative of the strain energy with respect to the displacement, Di , in the direction and location of the force, Fi . Fi = 2.2 Castiglianos Theorem - Part II U = D dF
Fj

...

complementary strain energy

F+ F j j F j

1111111111111111111 0000000000000000000 1111111111111111111 0000000000000000000 1111111111111111111 0000000000000000000

U*

U*(F)

Dj

U U = Fi Fi A displacement, Di , on an elastic solid is equal to the derivative of the complementary strain energy with respect to the force, Fi , in the direction and location of the displacement, Di . Di = If the solid is linear elastic, then U = U .

CEE 421L. Matrix Structural Analysis Duke University Fall 2012 H.P. Gavin

3 Superposition
Superposition is an extremely powerful idea that helps us solve problems that are statically indeterminate. To use the principle of superposition, the system must behave in a linear elastic fashion. The principle of superposition states: Any response of a system to multiple inputs can be represented as the sum of the responses to the inputs taken individually. By response we can mean a strain, a stress, a deection, an internal force, a rotation, an internal moment, etc. By input we can mean an externally applied load, a temperature change, a support settlement, etc.

4 Detailed Example of Castiglianos Theorem and Superposition

An example of a statically indeterminate system with external loads w(x) and three redundant reaction forces, RB , RC , and RD , is shown below. 111 000 111 000
D EA

111111111111111111111111111111111111111111111111111111111111 000000000000000000000000000000000000000000000000000000000000 w(x) 111111111111111111111111111111111111111111111111111111111111 000000000000000000000000000000000000000000000000000000000000 111111111111111111111111111111111111111111111111111111111111 000000000000000000000000000000000000000000000000000000000000 111111111111111111111111111111111111111111111111111111111111 000000000000000000000000000000000000000000000000000000000000 111 000 111111111111111111111111111111111111111111111111111111111111 000000000000000000000000000000000000000000000000000000000000 111 000 111111111111111111111111111111111111111111111111111111111111 000000000000000000000000000000000000000000000000000000000000 111 000 111 000 111 000 111 000 EI 111 000 111 000 111 000 111 000 111 000
A B L C

In general, the displacements at the locations of the unknown reaction forces are known, and, in this example these displacements will be taken as zero: DB = 0, DC = 0, DD = 0. Invoking the principle of superposition, we may apply the external loads, (w(x)) and the unknown reactions (RB , RC , and RD ) individually, and then sum-up the responses to each individual load. Further, we may represent the response to a reaction force, (e.g., RB ) as the response to a unit force co-located with the reaction force, times the value of the reaction force. Note that all four systems to the right of the equal sign in the following gure are statically determinate. Expressions for M o(x), m1 (x), m2 (x), m3 (x), No (x), n1 (x), n2 (x), and n3 (x) may be found from static equilibrium alone.

Strain Energy and Matrix Methods of Structural Analysis

R D
D D

N(x)
111111111111111111111111111111111111111111111111111111111111 000000000000000000000000000000000000000000000000000000000000 w(x) 111111111111111111111111111111111111111111111111111111111111 000000000000000000000000000000000000000000000000000000000000 111111111111111111111111111111111111111111111111111111111111 000000000000000000000000000000000000000000000000000000000000 111111111111111111111111111111111111111111111111111111111111 000000000000000000000000000000000000000000000000000000000000 111 000 111111111111111111111111111111111111111111111111111111111111 000000000000000000000000000000000000000000000000000000000000 111 000 111111111111111111111111111111111111111111111111111111111111 000000000000000000000000000000000000000000000000000000000000 111 000 111 000 111 000 111 000A C B 111 000

= +

1111111111111111111111111111111111111111111111111111111111111 0000000000000000000000000000000000000000000000000000000000000 w(x) 1111111111111111111111111111111111111111111111111111111111111 0000000000000000000000000000000000000000000000000000000000000 1111111111111111111111111111111111111111111111111111111111111 0000000000000000000000000000000000000000000000000000000000000 1111111111111111111111111111111111111111111111111111111111111 0000000000000000000000000000000000000000000000000000000000000 111 000 1111111111111111111111111111111111111111111111111111111111111 0000000000000000000000000000000000000000000000000000000000000 111 000 1111111111111111111111111111111111111111111111111111111111111 0000000000000000000000000000000000000000000000000000000000000 111 000 111 000 111 000 111 000A B C 111 000

No (x)

M(x)

R B

C
111 000 111 000 111 000 111 000 111 000 111 000 111 000

M o (x) n (x) 1 * R m (x) B 1 1 n (x) 2 B

111 000 111 000 111 000 111 000 111 000 111 000 111 000

* R
C

m (x) 2

1 n (x) 3

1
D

111 000 111 000 111 000 111 000 111 000 111 000 111 000 111 000

* R m (x) 3 D

In equation form, the principle of superposition says: M (x) = Mo (x) + m1 (x)RB + m2 (x)RC + m3 (x)RD N = No + n1 RB + n2 RC + n3 RD (9) (10)

(Note that in this particular example, No (x) = 0, n1 = 0, n2 = 0, n3 = 1, m1 (x) = 0 for x > xB , and m2 (x) = 0 for x > xC .)

CEE 421L. Matrix Structural Analysis Duke University Fall 2012 H.P. Gavin

The total strain energy, U , in systems with bending strain energy and axial strain energy is, 1 L M (x)2 1 N2 H U= dx + (11) 2 0 EI 2 EA We are told that the displacements at points B, C, and D are all zero and we need to assume the structure behaves linear elastically in order to invoke superposition in the rst place. Therefore, from Castiglianos Second Theorem, Di = U U = , Fi Fi

we obtain three expressions for the facts that DB = 0, DC = 0, and DD = 0. DB = 0 = DC = 0 = DD = 0 = U RB U RC U RD

Inserting equation (11) into the three expressions for zero displacement at the xed reactions, noting that EI and EA are constants in this problem, and noting that the strain energy, U , depends on the reactions R, only through the internal forces, M and N , we obtain DB = 0 = DC = 0 = DD = 0 = 1 EI 1 EI 1 EI
L

M (x)
0 L

M (x) H N dx + N RB EA RB M (x) N H N dx + RC EA RC M (x) H N dx + N RD EA RD

M (x)
0 L

M (x)
0

Now, from the superposition equations (9) and (10), M (x)/RB = m1 (x), M (x)/RC = m2 (x), M (x)/RD = m3 (x), N (x)/RB = n1 , N (x)/RC = n2 , and N (x)/RD = n3 . Inserting these expressions and the superposition equations (9) and (10) into the above equations for DB , DC , and DD , DB = 0 = DC = 0 = 1 EI
L 0

[Mo + m1 RB + m2 RC + m3 RD ] m1 dx +

H [No + n1 RB + n2 RC + n3 RD ] n1 EA

L 1 H [Mo + m1 RB + m2 RC + m3 RD ] m2 dx + [No + n1 RB + n2 RC + n3 RD ] n2 EI 0 EA L 1 H DD = 0 = [Mo + m1 RB + m2 RC + m3 RD ] m3 dx + [No + n1 RB + n2 RC + n3 RD ] n3 EI 0 EA These three expressions contain the three unknown reactions RB , RC , and RD . Everything else in these equations (m1 (x), m2 (x) ... n3 ) can be found without knowing the unknown

Strain Energy and Matrix Methods of Structural Analysis

reactions. By taking the unknown reactions out of the integrals (they are constants), we can write these three equations in matrix form.

L m1 m1 dx 0 EI L m2 m1 dx 0 EI L m3 m1 dx 0 EI

+ + +

n1 n1 H EA n2 n1 H EA n3 n1 H EA

L m1 m2 dx 0 EI L m2 m2 dx 0 EI L m3 m2 dx 0 EI

+ + +

n1 n2 H EA n2 n2 H EA n3 n2 H EA

L m1 m3 dx 0 EI L m2 m3 dx 0 EI L m3 m3 dx 0 EI

+ + +

n1 n3 H EA n2 n3 H EA n3 n3 H EA

RB RC = RD

L Mo m 1 dx 0 EI L Mo m 2 dx 0 EI L Mo m 3 dx 0 EI

+ + +

No n1 H EA No n2 H EA No n3 H EA

(12) This 3-by-3 matrix is called a exibility matrix, F. The values of the terms in the exibility matrix depend only on the responses of the structure to unit loads placed at various points in the structure. The exibility matrix is therefore a property of the structure alone, and does not depend upon the loads on the structure1 . The vector on the right-hand-side depends on the loads on the structure. Recall that this matrix looks a lot like the matrix from the three-moment equation. All exibility matrices share several properties: All exibility matrices are symmetric. No diagonal terms are negative. Flexibility matrices for structures which can not move or rotate without deforming are positive denite. This means that all of the eigenvalues of a exibility matrix describing a xed structure are positive. The unknowns in a exibility matrix equation are forces (or moments). The number of equations (rows of the exibility matrix) equals the number of unknown forces (or moments).

1 There are some fascinating cases in which the behavior does depend upon the loads, but that is a story for another day!

10

CEE 421L. Matrix Structural Analysis Duke University Fall 2012 H.P. Gavin

It is instructive to now examine the meaning of the terms in the matrix, F

D
L

F11 =

m1 m1 n1 n1 H dx + = 11 EI EA

111 displacement at 1 due to unit force at 1 000 111 000

A

111 000 111 000 111 000 111 000 111 000

11

D
L

F21 =

m2 m1 n2 n1 H dx + = 21 EI EA

111 displacement at 2 due to unit force at 1 000 111 000

A B

111 000 111 000 111 000 111 000 111 000

F 21

F12 =

m1 m2 n1 n2 H dx + = 12 EI EA

displacement at 1 due to unit force at 2

111 000 111 000 111 000 111 000 111 000 111 000 111 000
A

F 12

31

F31 =

n3 n1 H m3 m1 dx + = 31 EI EA

111 displacement at 3 due to unit force at 1 000 111 000

1
B C A

111 000 111 000 111 000 111 000 111 000

The fact that F12 = F21 is called Maxwells Reciprocity Theorem.

Strain Energy and Matrix Methods of Structural Analysis

11

5 Introductory Example of the Stiness Matrix Method

In this simple example, elements are springs with stiness k. A spring with stiness k > 0 connecting point i to point j, will have a force f = k(dj di ) where di is the displacement of point i and dj is the displacement of point j. (Tension is positive so di points into the spring and dj points away from the spring.)

The stiness matrix for this structure can be found using equilibrium and force-deection relationships (f = kd) for the springs.

#1: #2: #3:

Fx = 0 : Fx = 0 : Fx = 0 :

f1 k1 d1 + k2 (d2 d1 ) = 0 f2 k2 (d2 d1 ) k4 d2 + k3 (d3 d2 ) = 0 f3 k3 (d3 d2 ) k5 d3 = 0

In matrix form these three equations may be written:

k1 + k2 k2 0 d1 f1 k2 + k3 + k4 k3 d2 = f2 k2 0 k3 k3 + k5 d3 f3 The displacements are found by solving the stiness matrix equation for d, d = K1 f . The matrix K is called a stiness matrix. All stiness matrices are symmetric.

12

CEE 421L. Matrix Structural Analysis Duke University Fall 2012 H.P. Gavin

All diagonal terms of all stiness matrices are positive. Stiness matrices are diagonally dominant. This means that the diagonal terms are usually larger than the o-diagonal term. If the structure is not free to translate or rotate without deforming, then the stiness matrix is positive denite. This mathematical property guarantees that the stiness matrix is invertible, and a unique set of displacements, d, can be found by solving K d = f. The total potential energy, U , in this system of springs is 1 1 1 1 1 U = k1 d2 + k2 (d2 d1 )2 + k3 (d3 d2 )2 + k4 d2 + k5 d2 . 1 2 3 2 2 2 2 2 You should be able to conrm that this is equal to U= 1 T d Kd 2

Also, note that no matter what the values of the displacements, d, may be, the energy U is always positive. The statement 1 dT Kd > 0 d = 0 is another way of saying that 2 K is positive-denite. The set of forces required to deect coordinate i by a deection of 1 unit equals the i-th column of the stiness matrix. For example consider the case in which d1 = 1, d2 = 0, and d3 = 0,

k1 + k2 k2 0 1 k1 + k2 k2 + k3 + k4 k3 0 = k2 , k2 0 k3 k3 + k5 0 0 which is equal to the rst column of the stiness matrix.

Strain Energy and Matrix Methods of Structural Analysis

13

This fact may be used to derive the stiness matrix: d1 = 1, d2 = 0, d3 = 0

f1 = k1 + k2 ,

f2 = k2 ,

f3 = 0 ... 1st column

d1 = 0, d2 = 1, d3 = 0

f1 = k2 ,

f2 = k2 + k3 + k4 ,

f3 = k3 ... 2nd column

d1 = 0, d2 = 0, d3 = 1

f1 = 0 ,

f2 = k3 ,

f3 = k3 + k5 ... 3rd column

14

CEE 421L. Matrix Structural Analysis Duke University Fall 2012 H.P. Gavin

6 Basic Concepts of the Stiness Matrix Method

The previous example illustrates some of the basic concepts needed to apply the stiness matrix method to structures made out of bars and beams. There are, however, a few additional complications. Displacements in structures can be vertical, horizontal, or rotational, and structural bars and beams have a more complicated force-displacement relationships than those of simple springs. In applying the matrix stiness method of structural analysis, structures are described in terms of elements that connect nodes which can move in certain coordinate directions.

6.1 Elements In the stiness matrix method, structures are modeled as assemblies of elements such as bars, beams, cables, shafts, plates, and walls. Elements connect the nodes of the structural model. Like the simple springs in the previous example, structural elements have clearly dened, albeit more complicated, stiness relationships. The stiness properties of structural elements can be determined from equilibrium equations, Castiglianos Theorems, and/or the principle of virtual work. Structural elements can be mathematically assembled with one another (like making a structural system using a set of tinker-toys), into a system of equations for the entire structure.

6.2 Nodes The force-displacement relationship of a structural element is dened in terms of the forces and displacements at the nodes of the element. Nodes dene the points where elements meet. The nodes in the model of a truss are at the joints between the truss bars. The nodes in the model of a beam or a frame are at the reaction locations, at locations at which elements connect to each other, and possibly at other intermediate locations.

6.3 Coordinates Coordinates describe the location and direction at which forces and displacements act on an element or on a structure. Trusses are loaded with vertical and horizontal forces at the joints. The joints of a 2D truss can move vertically and horizontally; so there are two coordinates per node in a 2D truss. Beams and frames carry vertical and horizontal loads as well as bending moments. The nodes of a 2D frame can move vertically, horizontally, and can rotate; so there are three coordinates per node in a 2D frame.

Strain Energy and Matrix Methods of Structural Analysis

15

6.4 Elements, Nodes and Coordinates

Planar (2D) truss bar elements have two nodes and four coordinates, two at each end.

Space (3D) truss bar elements have two nodes and six coordinates, three at each end.

Planar (2D) frame elements have two nodes and six coordinates, three at each end.

Space (3D) frame elements have two nodes and twelve coordinates, six at each end.

16

CEE 421L. Matrix Structural Analysis Duke University Fall 2012 H.P. Gavin

6.5 Structural Nodes and Coordinates

Planar (2D) truss nodes and coordinates

Planar (2D) frame nodes and coordinates

Strain Energy and Matrix Methods of Structural Analysis

17

7 Relate the Flexibility Matrix to the Stiness Matrix

Column j of the stiness matrix: The set of forces at all coordinates required to produce a unit displacement at coordinate j

Column j of the exibility matrix: The set of displacements at all coordinates resulting from a unit force at coordinate j

Stiness matrix equation:

K11 K12 K21 K22 K31 K32 . . . . . . Kn1 Kn2

K13 K1n K23 K2n K33 K3n . . .. . . . . . Kn3 Knn

d1 d2 d3 . . . dn

f1 f2 f3 . . . fn

Flexibility matrix equation:

F11 F12 F21 F22 F31 F32 . . . . . . Fn1 Fn2

F13 F1n F23 F2n F33 F3n . . ... . . . . Fn3 Fnn

f1 f2 f3 . . . fn

d1 d2 d3 . . . dn

K = F1 and K1 = F A useful fact for 2-by-2 matrices ... a b c d

1

1 ad bc

d b c a

... you should be able to prove this fact to yourself.