This document is part of the notes written by Terje Haukaas and posted at www.inrisk.ubc.ca.

The notes are revised without notice and they are provided as is without warranty of any kind. You are encouraged to submit comments, suggestions, and questions to terje@civil.ubc.ca. It is unnecessary to print these notes because they will remain available online.

Warping Torsion

In addition to shear stresses, some members carry torque by axial stresses. This is called warping torsion. This happens when the cross-section wants to warp, i.e., displace axially, but is prevented from doing so during twisting of the beam. Not all cross-sections warp, and even those that warp do not carry torque by axial stresses unless they are axially restrained at some location(s) along the member. Cross- sections that do NOT warp include axisymmetric cross-sections and thin-walled cross-sections with straight parts that intersect at one point the cross-section, such as X-shaped, T-shaped, and L-shaped cross-sections. For these cross-sections all torque is carried by shear stresses, i.e., St. Venant torsion, regardless of the boundary conditions.

Warping of I-sections
As a pedagogical introduction to warping torsion, consider a beam with an I-section, such as a wide-flange steel beam. When torsion is applied to the beam then the flanges of this cross-section experiences bending in the flange-planes. In other words, torsion induces bending about the strong axis of the flanges. When the flanges are fixed at some point, such as in a cantilevered beam with a fully

Terje Haukaas

University of British Columbia

www.inrisk.ubc.ca

clamped end, some of the torque is carried by axial stresses. To understand this, denote the bending moment and shear force in each flange by M and V, respectively, as shown in Figure 1.

z, w
! !
h

z, w

Figure 1: Warping of I-section (z is the local axis for bending of flange, not the global z-axis of the cross-section).

The torque that the cross-section carries by bending in the flanges is:
T = h ! V

(1)

where h is the distance between the flanges and V is positive shear force in accordance with the document on Euler-Bernoulli beam theory. That document also provides the equilibrium equation that relates shear force to bending moment, which yields:
V= dM dx ! T = h" dM dx

(2)

The beam theory also provide the relationship between bending moment and flange displacement, w:

d 2w M = EI ! 2 dx

d 3w " T = h ! EI ! 3 dx

(3)

where I is the moment of inertia of one flange about its local strong axis. Next, Figure 1 is reviewed to determine the relationship between w and :

w = !" #

h 2

$ T =!

h2 d 3" # EI # 3 2 dx

(4)

which resulted in the differential equation for warping torsion of an I-section. However, this equation is generally written in this format:

T = ! ECw "

d 3# dx 3

(5)

which implies that, for I-sections, the cross-sectional constant for warping is:

Cw = I !

h2 2

(6)

Warping Torsion

Page 2

Terje Haukaas

University of British Columbia

www.inrisk.ubc.ca

where it is reiterated that I is the moment of inertia of one flange about its local strong axis.

Complete Differential Equation for Torsion

As mentioned earlier, when warping is restrained the torque is carried by both shear stresses, i.e., St. Venant torsion and axial stresses, i.e., warping torsion. Specifically, the torque from shear and axial stresses are superimposed, which leads to the following complete differential equation for torsion:

T = GJ !

d" d 3" # ECw ! 3 dx dx

(7)

When equilibrium with distributed torque along the beam, mx, is included, i.e., mx= dT/dx, then the full differential equation reads

ECw !

d 4" d 2" # GJ ! = mx dx 4 dx 2

(8)

Solution
The characteristic equation to obtain the homogeneous solution for the differential equation in Eq. (8) reads

!4"

GJ # ! 2 = 0 ECw

(9)

The roots are 0, 0, (GJ/ECw), and (GJ/ECw). Accordingly, the homogeneous solution is

! ( x ) = C1 " e

GJ ECw "x

+ C2 " e#

GJ ECw x

+ C3 " x + C4

(10)

which guides the selection of shape functions if an exact stiffness matrix with both St. Venant and warping torsion is sought. Another way of expressing the solution is:

! ( x ) = C1 " sinh

GJ ECw " x + C2 " cosh

GJ ECw " x + C3 " x + C4

(11)

where the coefficients, Ci, in Eq. (10) are different from those in Eq. (11). For example, the homogeneous solution for a cantilevered beam that is fully fixed at x=0 and subjected to a torque, To, at x=L has the solution

( $ &+ 1 To * tanh GJ ECw " L " % cosh GJ ECw " x # 1' !(x) = " " GJ ECw GJ * # sinh GJ EC " x + GJ EC " x w w ) ,

(12)

From this solution the torque carried by St. Venant torsion is computed by:
TSt .V . ( x ) = GJ ! d" dx

(13)

and the torque carried by warping torsion is computed by:

Warping Torsion

Page 3

Terje Haukaas

University of British Columbia

www.inrisk.ubc.ca

d 3# Twarping ( x ) = ! ECw " 3 dx

(14)

where TSt.V.(x)+Twarping(x)=To for all 0<x<L.

Bi-moment
In the theory of warping torsion the bi-moment, B, is defined as an auxiliary quantity. This has two primary objectives. The first is to introduce a degree of freedom for beam elements that carry torque by restrained warping. The other objective stems from our desire to formulate a theory with a quantity that is tantamount to the ordinary bending moment in beam theory. In other words, the objective is to establish an equation of the form B=ECw, which is analogous to the equation M=EIwfrom beam theory. To this end, let the bi-moment for I-sections be defined by
B ! M " h

(15)

where M is again the bending moment in the flange about its strong axis. Substitution of the relationship between bending moment in the flange, w, and the flange displacement, w, from Euler-Bernoulli beam theory yields

B = EI !

d 2w ! h dx 2 d 2# h 2 " dx 2 2 d 2# dx 2

(16)

and substitution of the relationship between w and from Eq. (4) yields

B = ! EI "

(17)

which in light of Eq. (6) is written

B = ! ECw "

(18)

This is the desired result, which shows that Eq. (15) is the appropriate definition of the bi-moment for I-sections. It is emphasized that the bi-moment in itself is not measurable, but it serves as a convenient auxiliary quantity in the theory of warping torsion. When warping degrees of freedom are included in beam elements then the bi-moment in the force vector corresponds to the derivative of the rotation, i.e., , in the displacement vector.

Unified Bending and Torsion of Thin-walled Cross-sections

The following theory, named after Vlasov, is developed for warping torsion of thin- walled cross-sections. Because warping torsion and beam bending are both formulated in terms of axial stresses it is possible to combine the two theories. In fact, the omission of shear deformation in Euler-Bernoulli beam theory is carried over to the warping theory that is presented in the following. It is noted that no theory of warping for general thick-walled cross-sections is currently provided in

Warping Torsion

Page 4

Terje Haukaas

University of British Columbia

www.inrisk.ubc.ca

these documents. Although this is a shortcoming, the presented theory is sufficient for many practical applications. This is because many thick-walled cross-section types are difficult to fully restrain axially. In contrast, it is easier to imagine connection designs for thin-walled cross-sections that provide sufficient axial restraint to develop torque due to axial stresses.

z, w

Centroid (0, 0)

r
h

y, v

! s, v

x, u T, !

Shear centre / centre of twist (ysc, zsc)

Figure 2: Beam axis system and corresponding displacements.

Kinematics

The objective in this section is to establish a relationship between axial strain, x, and axial displacement, u. For this purpose, let the coordinate s follow the centre line of the contour of the cross-section and let h denote the distance from the centre of rotation (ysc, zsc) to the tangent of the coordinate line s, as shown in Figure 2. The y-z-axes originate in the centroid of the cross-section, while the shear centre coordinates ysc and zsc are presently unknown. Analogous to the omission of shear deformation in Euler-Bernoulli beam theory, the shear the shear strain xs is neglected in the combined theory of warping torsion and bending. In turn, this precludes shear stresses, but as in the beam theory they are recovered later by equilibrium. Furthermore, the shear stresses from St. Venant torsion are superimposed so that the total shear strain xs is equal to that from St. Venant torsion. More precisely, xs is equal to the shear strain at r=0 from St. Venant theory, i.e., at the mid-plane of the cross-section profile. In passing it is noted that according to the St. Venant theory the shear strain at r=0 for open cross-sections is zero. Furthermore, as a fundamental kinematics postulation it is assumed that the cross- section retains its shape. This implies that s=s=0 and that the displacement in the s-direction is:

! = ! v " cos(# ) + w " sin(# ) + $ " h v

(19)

Warping Torsion

Page 5

Terje Haukaas

University of British Columbia

www.inrisk.ubc.ca

where v and w are the displacements of the cross-section and is the angle between the s-axis and the y-axes. The contributions to Eq. (19) are illustrated in Figure 3.

! "h

w ! sin(" )

-v

dz
! v " cos(# )

ds dy
!

!
(ysc, zsc)

Figure 3: Contributions to the displacement along the s-axis.

Given the previous assumptions the relevant kinematics equations are

! x = u, x
St .V . !, x + u,s = ! xs ! xs = v

(20) (21)

StV = 0 for open cross-sections. Substitution of Eq. where it is reemphasized that ! xs (19) into Eq. (21) yields

dv dw d$ du St .V . " cos(# ) + " sin(# ) + "h + = % xs dx dx dx ds

(22)

Next, the relationships between the differentials ds, dy, and dz are established. With reference to Figure 3 they are:

dy = ! cos(" ) ds dz = sin(" ) ds
dv dy dw dz d" du St .V . ! + ! + !h + = # xs dx ds dx ds dx ds

(23)

Substitution of Eq. (23) into Eq. (22) yields Solving for du yields (24)

du = !

dv dw % d# St .V . ( " dy ! " dz ! ' " h ! $ xs " ds * & dx ) dx dx

(25)

St .V . It is desirable to express ! xs in terms of d/dx so that the latter can be pulled outside the parenthesis. This is achieved by utilizing equations from the theory of St.

Warping Torsion

Page 6

Terje Haukaas

University of British Columbia

www.inrisk.ubc.ca

St .V . Venant torsion. For open cross-sections ! xs vanishes, but in general material law yields

St .V . ! xs =

St .V . " xs G

(26)

St .V . and ! xs = d" dr where is Prandtls stress function, which is cross-section dependent. For thin-walled cross-sections with one cell a good stress function is

As a result, the shear stress is

# 1 r& ! ( s, r ) = K " % + ( $2 t'

K t

(27)

St .V . ! xs = " ,r =

(28)

To express the stress in terms of d/dx it is first noted that

T = GJ

d! dx

(29)

and that the torque, T, can be related to the stress by the stress-resultant equation

T = 2 ! # " dA
A

(30)

For the stress function in Eq. (27) the integral evaluates

T = 2 ! K ! Am

(31)

Finally, Eqs. (31) and (29) substituted into Eq. (28) yields the desired expression:
St .V . ! xs =

K T GJ d# = = " t t " 2 " Am 2 " t " Am dx

(32)

Substitution into Eq. (25) yields Integration yields

u ( y, z ) = u ! dv dw d# "y! "z ! "$ dx dx dx

du = !

dv dw d# $ J ' " dy ! " dz ! " & h ! " ds dx dx dx % 2 " t " Am ) (

(33)

(34)

where u is the integration constant, which represents a uniform axial displacement of the cross-section and has been defined as:

% J ( ! " +'h# ds 2 $ t $ Am * & )

(35)

Warping Torsion

Page 7

Terje Haukaas

University of British Columbia

www.inrisk.ubc.ca

It is reemphasized that this expression is valid for cross-sections with one cell. For open cross-sections the shear strain in the mid-plane of the cross-section profile is zero, which implies that

! " # h ds

(36)

Finally, combining Eq. (20) with Eq. (34) yields the final kinematics equation:

!x =

du d 2 v d2w d 2$ " 2 # y " 2 # z " 2 #% dx dx dx dx

(37)

Material Law
Hookes law provides the relationship between axial stress and axial strain:

! x = E " # x

(38)

Section Integration

Integration of axial stress over the cross-section yields the axial force:

N = " ! x dA
A

(39)

Integration of axial stress multiplied by distance from the centroid yields bending moment:

M z = ! $ " x # y dA
A

(40)

Integration of axial stress multiplied by distance from the centroid yields bending moment:

M y = ! $ " x # z dA
A

(41)

Integration of axial stress multiplied by the previously defined quantity is defined as the bi-moment:

B ! " & # x $% dA
A

(42)

Equilibrium
Distributed axial load along the beam is related to the axial force by the following equilibrium equation:
qx = ! dN dx

(43)

Distributed load in z-direction, which is assumed to act through the shear centre or it will contribute to mx, is related to the shear force by the following equilibrium equation:

Warping Torsion

Page 8

Terje Haukaas

University of British Columbia

www.inrisk.ubc.ca

dVz (44) dx Equilibrium also provides the relationship between shear force and bending moment: qz = !
dx The corresponding equilibrium equations in the other direction are:

Vz =

dM y

(45)

and

qy =

dVy dx

(46)

dM z dx Finally, equilibrium for distributed torque along the beam yields: Vy =

mx = ! dT dx

(47)

(48)

Differential Equations
Substitution of the kinematics equation in Eq. (37) into the material law in Eq. (38) yields:

!x = E"

du d2v d2w d 2$ # E " 2 " y # E " 2 " z # E " 2 "% dx dx dx dx

(49)

Substitution of Eq. (49) into the section integration Eqs. (39) to (42) yields the following set of equations:

, . ) dA . A ! N % . * y dA # # . ) M z # # A " & = E(. . * ) z dA # My # . A # B # $ ' . . * ) + dA . - A

! * ) + dA / 1# A A A 1# 2 y dA y ( z dA ) y (+ dA 1 # ) ) 1# # A A A 1" y ( z dA z 2 dA ) z (+ dA 1 # ) ) 1# A A A 1# y (+ dA ) z (+ dA ) +2 dA 1 # ) A A A 1 0# $ * ) y dA * ) z dA

du dx d 2w dx 2 d 2v dx 2 d 22 dx 2

% # # # # # & # # # # # '

(50)

where symmetry is observed. Under certain condition described shortly, the equations become decoupled and reduces to:

N = EA !

du dx

(51)

Warping Torsion

Page 9

Terje Haukaas

University of British Columbia

www.inrisk.ubc.ca

d 2w M y = EI y 2 dx M z = EI z d 2v dx 2

(52) (53) (54)

d 2" B = ! ECw 2 dx

where the diagonal components of the matrix in Eq. (50) have been named as follows:

A = ! d A
A

(55) (56) (57) (58)

I z = ! y 2 dA
A

I y = ! z 2 dA
A

Cw = " !2 dA
A

For the system of equations in Eq. (50) to be decoupled, the six off-diagonal elements of the coefficient matrix must be zero. These six conditions form an important part of the cross-section analysis. In fact, they determine the following six unknowns of the cross-section: 1. 2. 3. 4. 5. 6. yo = y-coordinate of the centroid zo = z-coordinate of the centroid o = orientation of the principal axes C = normalizing constant for the -diagram ysc = y-coordinate of the shear centre zsc = z-coordinate of the shear centre

Specifically, the coordinates of the centroid of the cross-section are determined by the two equations:

! y d A = ! z d A = 0
A A

(59)

The orientation of the principal axes are determined by the condition:

" y ! z d A = 0
A

(60)

The normalization of the -diagram is carried out to satisfy the condition:

" ! d A = 0
A

(61)

Warping Torsion

Page 10

Terje Haukaas

University of British Columbia

www.inrisk.ubc.ca

The following equations are satisfied if the -diagram is established with respect to the shear centre, which implies that the shear centre coordinates are determined by these equations:

# y !" d A = # z !" d A = 0
A A

(62)

Adding equilibrium with external forces yields the final differential equations for axial deformation and bending

qx = ! EA " qz = EI y qy = EI z

d 2 uo dx 2

(63) (64) (65)

d 4 wD dx 4 d 4 vD dx 4

The derivation of the differential equation that combines St. Venant torsion and warping torsion starts with the definition of the stress resultant:

T = # ! xs " t " h dA = # qs " h ds = # qs d $ = [ qs "$ ]% & # $ dqs

A A A A

(66)

where qs is the shear flow and the boundary term [qs] from integration by parts is zero. Because shear strains are omitted from the warping theory it is necessary to employ equilibrium to recover the shear flow. With reference to Figure 4, equilibrium yields:
d! x " ds " t + d# xs " dx " t = 0 $ d! x d# " t + xs " t = 0 dx ds $ dqs d! = % x " t (67) ds dx

!x
ds

! xs

dx

! x + d! x ! xs + d! xs
s
Figure 4: Equilibrium to recover shear stresses.

Warping Torsion

Page 11

Terje Haukaas

University of British Columbia

www.inrisk.ubc.ca

Substitution of Eq. (67) into Eq. (66) yields:

T = ! # " dqs = # " $

A A

d% x d% d dB $ t ds = # " $ x dA = " $ % x dA = # dx dx dx A dx A

(68)

Adding the torque carried by shear stresses, i.e., T=GJ(d/dx) and employing Eq. (54) yields:

d" d 3" T = GJ ! # ECw 3 dx dx

(69)

Adding equilibrium with distributed torque from Eq. (48) yields the complete differential equation for St. Venant torsion and warping torsion:

mx = ECw

d 4! d 2! " GJ # dx 4 dx 2

(70)

The solution to this differential equation was presented in Eq. (10).

Final -diagram

Eqs. (61) and (62) provide the conditions that the final -diagram must satisfy. In particular, the final -diagram must be normalized and it must be drawn about the point (ysc, zsc), i.e., the shear centre of the cross-section. In order to arrive at the final diagram, a trial diagram is first drawn. Thereafter, the trial diagram is modified to obtain the final one. For this purpose it is useful to establish an expression that relates the -diagram drawn about an arbitrary point Q to another diagram drawn about another point SC. Since the -values consist of accumulated double sector areas it is of interest to study a generic infinitesimal contribution. To this end, Figure 5 identifies by gray-shading the -diagram contributions for an infinitesimal length ds of the cross-section. The double area of the sector that originates in (ysc, zsc) is
d! sc = dy " ( z # zsc ) # dz " ( y # ysc )

(71)

The double area of the sector that originates in (yQ, zQ) is

d!Q = dy " z # zQ # dz " y # yQ

(72)

The difference between the two -diagram contributions in Figure 5 is

d! sc " d!Q = ysc " yQ # dz " zsc " zQ # dy

(73)

Integration of infinitesimal contributions yield the expression for the difference in -diagrams:

! sc " !Q = ysc " yQ # z " zsc " zQ # y + C

(74)

where C is the integration constant. Rearranging and renaming sc to yields:

! = !Q + ysc " yQ # z " zsc " zQ # y + C

(75)

Warping Torsion

Page 12

Terje Haukaas

University of British Columbia

www.inrisk.ubc.ca

which is the expression for the final -diagram once C, ysc, and zsc are known.

dy (y, z) dz

(yQ, zQ)

ds

(ysc, zsc)
Figure 5: Relationship between two -diagrams.

The integration constant, C, normalizes the -diagram and is determined by requiring the integral of the -diagram in Eq. (75) to be zero according to Eq. (61):

" ! dA = " !
A A

dA + " C dA = 0
A

# C=$

"!
A

dA
(76)

where terms cancel in the integration because y and z originate at the centroid of the cross-section. The shear centre coordinates ysc and zsc are determined by requiring the integrals in Eq. (62) of the -diagram in Eq. (75) to vanish:

# y !" dA = # y !"
A A

dA $ zsc $ zQ ! # y dA = 0
2 A

)
)

% zsc = zQ +

# y !"
A

dA
(77)

Iz

# z !" dA = # z !"
A A

dA + ysc $ yQ ! # z dA = 0
2 A

% ysc = yQ $

# z !"
A

dA
(78)

Iy

where terms cancel in the integration because y and z are principal axes of the cross-section. In summary, the following procedure is suggested for the determination of C, ysc, zsc, and ultimately the -diagram: 1. Select an arbitrary point, Q, in the cross-section

Warping Torsion

Page 13

Terje Haukaas

University of British Columbia

www.inrisk.ubc.ca

2. Draw the trial Q-diagram about Q, i.e., gather contributions according to Eq. (35) along cross-section parts by a clockwise radar sweep about Q, while ensuring that the Q-diagram is continuous (contributions that are made clockwise are positive, while those that are made by a counter-clockwise sweep to ensure continuous diagram are negative) 3. Determine C by Eq. (76) 4. Determine ysc by Eq. (77) 5. Determine zsc by Eq. (78) 6. Determine final -diagram by Eq. (75) Thereafter, the cross-sectional constant for warping torsion, Cw, is computed by Eq. (58) and the stresses are determined according to the following equations.

Stresses
Analogous to the Euler-Bernoulli beam theory, axial stresses are obtained by first combining kinematics from Eq. (37) with material law from Eq. (38), followed by substitution of the full differential equations without external equilibrium, i.e., Eqs. (51) to (54), which yields:

!x =

M N Mz B " #y+ y #z + #$ A Iz Iy Cw

(79)

It is observed that the omega diagram shows the distribution of axial stresses in the cross-section due to torsion when warping is restrained. The shear flow is recovered by equilibrium, which was presented in Eq. (67):

qs = t ! " xs = # t ! d" xs = $ # t !
0 0 s

d% x ds dx
(80)

' dN 1 dM z 1 * dM y 1 dB 1 = $# t !) ! $ ! !y+ ! !z + ! !&, dA dx I y dx Cw + ( dx A dx I z 0 =$ V V dN 1 dB 1 ! ! As + y ! Sy $ z ! Sz $ ! ! S& dx A Iz Iy dx Cw

where the following definitions are made:

As = ! dA = ! t ds
0 0 s s

(81) (82) (83)

Sy = ! y dA = ! t " y ds
0 0 s

Sz = ! z dA = ! t " z ds
0 0

Warping Torsion

Page 14

Terje Haukaas

University of British Columbia

www.inrisk.ubc.ca

S! = " ! dA = " t #! ds
0 0

(84)

Warping Torsion

Page 15