International Journal of Solids and Structures 78-79 (2016) 125–130

Contents lists available at ScienceDirect

International Journal of Solids and Structures

journal homepage: www.elsevier.com/locate/ijsolstr

Exact theory for a linearly elastic interior beam

Anssi T. Karttunen∗, Raimo von Hertzen
Department of Applied Mechanics, School of Engineering, Aalto University, FI-00076 Aalto, Finland

a r t i c l e i n f o a b s t r a c t

Article history: In this paper, an elasticity solution for a two-dimensional (2D) plane beam is derived and it is shown that
Received 24 November 2014 the solution provides a complete framework for exact one-dimensional (1D) presentations of plane beams.
Revised 7 August 2015
First, an interior solution representing a general state of any 2D linearly elastic isotropic plane beam under
Available online 19 October 2015
a uniform distributed load is obtained by employing a stress function approach. The solution excludes the
Keywords: end effects of the beam and is valid sufficiently far away from the beam boundaries. Then, three kinematic
Plane beam variables defined at the central axis of the plane beam are formed from the 2D displacement field. Using these
Elasticity solution central axis variables, the 2D interior elasticity solution is presented in a novel manner in the form of a 1D
Airy stress function beam theory. By applying the Clapeyron’s theorem, it is shown that the stresses acting as surface tractions on
Kinematic variables the lateral end surfaces of the interior beam need to be taken into account in all energy-based considerations
Clapeyron’s theorem related to the interior beam. Finally, exact 1D rod and beam finite elements are developed by the aid of the
Finite element
axis variables from the 2D solution.
© 2015 Elsevier Ltd. All rights reserved.

1. Introduction
Elasticity solutions for plane beams are of fundamental interest
in mechanical sciences. An important application of such solutions is
the benchmarking of beam theories based on various kinematic as-
sumptions. Two-dimensional (2D) interior elasticity solutions can be
easily obtained, for example, for an end-loaded cantilever and a uni-
formly loaded simply-supported beam by employing the Airy stress
function (e.g., Timoshenko and Goodier, 1970).
An interior solution excludes, by virtue of the Saint Venant’s prin-
ciple, the end effects that decay with distance from the ends of a
beam. In the calculation of displacements, constraint conditions are
applied at the beam supports to prevent it from moving as a rigid
body. These constraints for the 2D elasticity solution can be chosen so
that they correspond to the boundary conditions of, for example, the
Timoshenko beam theory (Timoshenko, 1921). Due to the foregoing, a
2D interior plane stress solution for a plane beam acts as an ideal ref-
erence solution for narrow one-dimensional (1D) shear-deformable
beam models that do not include end effects.
Many beam and plate theories are based on an assumed dis-
placement field similar to the one first used by Vlasov (1957). These
theories are commonly referred to as third-order theories because
third-order polynomials are used in the expansion of the displace-
ment components. For surveys on third-order kinematics and plate
∗
Corresponding author. Tel.: +358405272713.
E-mail addresses: anssi.karttunen@aalto.fi (A.T. Karttunen), raimo.von.hertzen@
aalto.fi (R. von Hertzen).
theories, see the works by Jemielita (1990) and Reddy (1990, 2003).
Two examples of third-order beam theories are the Levinson and the
Reddy–Bickford beams for which the assumed displacement field is
exactly the same (Bickford, 1982; Heyliger and Reddy, 1988; Levinson,
1981; Reddy, 1984). As first shown by Bickford (1982), the Reddy–
Bickford beam exhibits a boundary layer character, that is, the decay-
ing end effects are present in the beam. The Reddy–Bickford theory
is obtained through an energy-based variational formulation, which
results in additional higher-order load resultants in comparison to an
interior elasticity solution. If the higher-order load resultants are ne-
glected, the Levinson theory is obtained.
In this study, a general interior elasticity solution is derived for
a uniformly loaded linearly elastic homogeneous isotropic 2D plane
beam. As the main novelties of the study we find that
• The 2D solution provides the exact third-order kinematics for the
beam and can be presented directly in the form of a conventional
1D beam theory.
• By applying the Clapeyron’s theorem, it is shown that the stresses
acting as surface tractions on the lateral end surfaces of the inte-
rior beam are an intrinsic part of all energy-based considerations.
• The 2D solution can be discretized in order to obtain 1D rod and
beam finite elements, which provide exact 2D interior displace-
ment and stress distributions.
In more detail, the paper is organized as follows. In the introduc-
tory Section 2, a polynomial Airy stress function is used to derive
the interior stress field for a 2D plane beam under a uniform dis-
tributed load. The strains are calculated from the stresses according
to the plane stress condition and the displacements are integrated

http://dx.doi.org/10.1016/j.ijsolstr.2015.09.010
0020-7683/© 2015 Elsevier Ltd. All rights reserved.
126 A.T. Karttunen, R. von Hertzen / International Journal of Solids and Structures 78-79 (2016) 125–130

p 2.2. Interior stress field of a plane beam

By adapting a general solution procedure outlined by Barber

h/2 y
(2010, chap. 5), we find that the polynomial stress function for the
N x N interior problem of any plane beam under a uniform pressure p (see
h/2 Fig. 1) is
M Q Q M
 
L/2 L/2 4y2
 (x, y) = c1 y2 + c2 y3 + c3 xy 1 − 2
3h
q  3 2  
Fig. 1. 2D homogeneous isotropic plane beam with a rectangular cross-section under
a uniform pressure. The load resultants act at an arbitrary cross-section of the beam. − 5h x + 15h x y + 4y3 y2 − 5x2 ,
2 2
(5)
240I
from the strains. In Section 3, three kinematic variables defined at where q = pt is the uniform load, I = th3 /12 is the second moment of
the central axis of the plane beam are formed from the 2D interior the cross-sectional area and c1 , c2 and c3 are to be solved by the aid
displacement field. Using these new central axis variables, 1D beam of Eqs. (1). The stresses calculated from Eqs. (3) are
equations are developed. The total potential energy of the interior 8c3 xy q(3x2 y − 2y3 )
beam and Clapeyron’s theorem are considered. Finally, exact 1D in- σx = 2c1 + 6c2 y − + , (6)
h2 6I
terior rod and flexural beam finite elements are developed from the
q
2D interior elasticity solution. Conclusions are presented in Section 4. σy = − (h3 + 3h2 y − 4y3 ), (7)
24I
2. Stress function solution for a plane beam    
4y2 qx h2
τxy = c3 −1 + − y2 . (8)
2.1. Plane beam problem and Airy stress function h2 2I 4
Note that the above interior stress distributions are universal, that is,
Fig. 1 presents a 2D homogeneous isotropic plane beam under a they are valid for any plane beam under a uniform load since they
uniform pressure p. The beam has a rectangular cross-section of con- are not associated with any particular constraint conditions at the
stant thickness t and the length and height of the beam are L and h, beam ends. Using Eqs. (6) and (8), the load resultants calculated from
respectively. The load resultants N, M and Q stand for the axial force, Eqs. (1) are
bending moment and shear force, respectively. These cross-sectional  
load resultants are calculated from the equations 2 q 2 h2 2
  N = 2Ac1 , M = 6Ic2 − Ac3 x + x − , Q = qx − Ac3 ,
h/2 h/2 3 2 10 3
N(x) = t σx (x, y)dy, M(x) = t σx (x, y)ydy,
−h/2 −h/2 (9)
 h/2
where A = ht is the area of the cross-section. As a first step towards
Q (x) = t τxy (x, y)dy, (1)
−h/2 presenting the solution in the form of a 1D beam theory, it can be
which can be used to impose the force and moment boundary condi- easily verified that the following global equilibrium equations, which
tions at x = ±L/2. The boundary conditions on the upper and lower can also be obtained by integrating the 2D stress equilibrium equa-
surfaces of the beam are tions, hold for the load resultants (9)

σy (x, h/2) = −p, σy (x, −h/2) = 0, τxy (x, ±h/2) = 0. (2) ∂N ∂M ∂Q

= 0, = Q, = q. (10)
∂x ∂x ∂x
Note that the boundary conditions are satisfied in a strong (pointwise)
sense on the upper and lower surfaces, whereas at the beam ends the We note that Schneider and Kienzler (2015) arrived at the same equi-
tractions are not specified at each point but only through the load librium equations (10) through their recent exact 3D representation
resultants and, thus, the boundary conditions are imposed only in a of linear elasticity. When c1 , c2 and c3 are solved from Eqs. (9) and
weak sense (Barber, 2010). In the case of Fig. 1, the replacement of the substituted into Eqs. (6) and (8), we obtain
true boundary conditions at the beam ends by the statically equiva- N My 3qy qy3
lent weak boundary conditions (load resultants) implies that the ex- σx = + + − , (11)
A I 5A 3I
ponentially decaying end effects of the plane beam are neglected by
Q
virtue of the Saint Venant’s principle and only the interior solution τxy = (h2 − 4y2 ). (12)
of the beam is under consideration. The interior solution represents
8I
essentially a beam section which has been cut off from a complete The stress distribution of Eq. (11) has been called by Rehfield and
beam far enough from the real lateral boundaries at which the true Murthy (1982) the refined (nonclassical) axial stress distribution in
boundary conditions could be set. The stresses of the plane beam are the context of their beam theory. More complicated distributed loads
obtained from the equations lead to different additional load terms in the stresses. By setting
q = 0, Eqs. (11) and (12) give the stress distribution of the classical
∂ 2 ∂ 2 ∂ 2
σx = , σy = , τxy = − , (3) Euler–Bernoulli beam.
∂y 2 ∂x 2 ∂ x∂ y
where  (x, y) is the Airy stress function. Eqs. (3) satisfy the two- 2.3. Example – Simply-supported beam
dimensional equilibrium equations. To ensure compatibility, it is re-
quired that the stress function satisfies the biharmonic equation As an example, let us consider a simply-supported beam under
(Barber, 2010) a constant uniform load q. In a setting according to Fig. 1, the axial
∂ 4 ∂ 4 ∂ 4 force, bending moment and shear force along the beam are given by
+ 2 + = 0. (4)
∂ x4 ∂ x2 ∂ y2 ∂ y4
N(x) = 0, M(x) = q(x2 /2 − L2 /8), Q (x) = qx, (13)
The solution to the plane beam problem is obtained by finding a solu-
tion of Eq. (4) that satisfies the stress boundary conditions (2) of the respectively. We find that the interior stress state in the beam calcu-
beam. lated from Eqs. (7) and (11)–(13) is the same as the one found in any
 A.T. Karttunen, R. von Hertzen / International Journal of Solids and Structures 78-79 (2016) 125–130 127

elasticity textbook for the particular problem at hand (e.g., Barber, cross-section at the central axis the expressions
2010). The benefit of using Eqs. (7), (11) and (12) is that they provide 1 qν x
the interior stress field also for different support conditions, granted ux (x) = Ux (x, 0) = (2c1 x + D1 ) + , (18)
E 2Et
that the load resultants along the beam are known. In other words,
instead of seeking a particular stress function suitable for the case  
at hand, one can take use of the systematic approach provided by 1 4c3 x3
uy (x) = Uy (x, 0) = −3c2 x2 + + Cx + D2
Eqs. (7), (11) and (12). The procedure of creating a general interior E 3h2
stress field for a uniform distributed load presented above may be
qx2  2 
extended to more complicated loads by updating the load dependent + 3h (2 + ν) − 2x2 , (19)
48EI
part of the stress function (5). As a rule of thumb in isotropic cases,
  2 
the interior solution is typically considered to be a good approxima- ∂ Ux 1 4x
tion when the axial distance to an end of a beam is at least equal to φ(x) = (x, 0) = 6c2 x − C − c3 + 2(1 + ν)
∂y E h2
the height of the beam.
qx  2 
+ 4x + 3ν h2 , (20)
24EI
2.4. Interior strains and displacements of a plane beam
respectively. Using these kinematic central axis variables, we can
present the displacements (16) and (17) in the form
Under plane stress, the stress-strain relations read
 
σx − νσy σy − νσx τxy 2(1 + ν) 4y3 ∂ uy ν y3 ∂ 2 φ
x = , y = , γxy = = τxy , (14) Ux (x, y) = ux + yφ − φ+ + , (21)
E E G E 3h2 ∂x 6 ∂ x2

where  x ,  y and γ xy are the axial normal strain, transverse normal

strain and transverse shear strain, respectively, and E, G and ν are the
∂ ux ν y2 ∂φ
Uy (x, y) = uy − ν y −
Young’s modulus, shear modulus and the Poisson ratio, respectively. ∂x 2 ∂x
qy  
The strain-displacement relations are + (2h + 3y)(ν 2 − 1)h2 + 2y3 (1 + 2ν) . (22)
48EI
∂U ∂U ∂U ∂U
x = x , y = y , γxy = x + y , (15) These displacements represent the exact third-order interior kine-
∂x ∂y ∂y ∂x matics of a linear homogeneous isotropic beam having a narrow rect-
where Ux (x, y) and Uy (x, y) are the displacements in the directions of angular cross-section. If we neglect the load term and the Poisson
x and y, respectively. The displacements are integrated from the first effect in Eqs. (21) and (22), the 2D displacement field is exactly of
two of Eqs. (15), and the resulting arbitrary functions f(y) and g(x) are the same form as in the Levinson and Reddy–Bickford beam theories.
resolved by substituting the calculated displacements into the last of Note that instead of the Airy stress function, one may also consider
Eqs. (15). As the final result, we get for the interior displacements the the Marguerre function for the case of plane stress and the Love strain
expressions function for plane strain in order to arrive at the displacements (21)

1
2c3 y
and (22) (e.g., Soutas-Little, 1973). In terms of the central axis vari-
Ux (x, y) = 2c1 x + 6c2 xy − ables, the strains (15) may be readily written in the form
E 3h2
 
∂φ ∂ ux qy3
x = y + − (2 + ν), (23)
× 3h2 (1 + ν) + 6x2 − 2y2 (2 + ν) − Cy + D1 ∂x ∂ x 6EI
qx  3   
+ ν h + 3ν h2 y + 4x2 y − 4y3 (2 + ν) , (16) ∂φ ∂ ux
24EI y = −ν y +
∂x ∂x
Uy (x, y) q  
  + (3y + h)(ν 2 − 1)h2 + 4y3 (1 + 2ν) , (24)
1 4c3 (x3 + 3ν xy2 ) 24EI
= −2c1 ν y − 3c2 (x2 + ν y2 ) + + Cx + D2
E 3h2   
q   4y2 ∂u
+ −2h3 y + 3h2 x2 (2 + ν) − y2 γxy = 1 − 2 φ+ y . (25)
48EI
  h ∂x
+ 2 y4 (1 + 2ν) − x4 − 6ν x2 y2 , (17) Note that if the strains are calculated directly from the expressions
where the constants C, D1 and D2 define the degrees of freedom of (21) and (22) using the definitions (15), the results simplify to Eqs.
the plane beam as a rigid body. The constant C relates to anticlock- (23)–(25) due to the specific low-order polynomial form of the cen-
wise rotation about the origin and D1 and D2 correspond to transla- tral axis variables.
tions in the directions of x and y, respectively. Note that C, D1 and D2
are determined by the boundary region displacements of the beam, 3.2. Load resultants and beam equilibrium equations
whereas the constant coefficients c1 , c2 and c3 can be calculated from
the axial force, bending moment and shear force according to Eqs. (9). Using the strains (23)–(25) and applying the plane stress constitu-
tive relations (14), the load resultants (1) in terms of the central axis
variables can be written as
3. 2D solution in the form of a 1D beam theory
∂ ux qhν
N = EA − , (26)
3.1. Displacement and strain fields by central axis variables ∂x 2

∂φ qh2
The general interior solution of Section 2 can be presented in M = EI − (2 + 5ν), (27)
terms of three kinematic variables derived from the displacements
∂x 40
 
(16) and (17) at the central axis so that the variables depend only on 2 ∂ uy
x. We obtain for the axial displacement and the transverse deflec- Q = GA φ + . (28)
3 ∂x
tion of the central axis, and for the clockwise positive rotation of the
128 A.T. Karttunen, R. von Hertzen / International Journal of Solids and Structures 78-79 (2016) 125–130

Substitution of the load resultants into the equilibrium equations (10) p

leads to
uy,1 uy,2
∂ 2 ux φ1 h/2 y φ2
EA = 0, (29)
∂ x2 N1 ux,1 x ux,2 N2
  h/2
∂ 2φ 2 ∂ uy M1 Q1 Q2 M2
EI 2 − GA φ + = 0, (30)
∂x 3 ∂x L/2 L/2
 
2 ∂φ ∂ 2 uy Fig. 2. Set-up according to which exact rod and flexural beam finite elements are
GA + = q. (31) developed.
3 ∂x ∂ x2

By uncoupling Eqs. (30) and (31), the axis variables uy and φ can be 3.4. Applying Clapeyron’s theorem
solved from
Let us briefly study the energetic aspects of the interior beam. The
∂ 3φ
EI = q, (32) strain energy of the 2D beam and the external work due to the uni-
∂ x3 form load are given by
∂ 4 uy   L/2
EI = −q. (33) 1
∂ x4 U= (σx x + σy y + τxy γxy )dV, Wq = − qUy (x, h/2)dx,
2 V −L/2
In summary, the general solution to Eqs. (29)–(31) is given by (34)
Eqs. (18)–(20), and the solution is expanded into the whole interior
region of the 2D plane beam through Eqs. (21) and (22). These equa- respectively. While the contributions (34) to the total potential en-
tions constitute an alternative representation of the elasticity solu- ergy of the beam are apparent, the following consideration may not
tion presented in Section 2. We can see, for example, that the ob- be that obvious at first. We recall from the problem definition in
tained beam equations (32) and (33) are exactly the same as those Section 2.1 that the interior solution part represents essentially a
in the Timoshenko beam theory with a constant uniform distributed beam section with fully-developed interior stresses which has been
load (e.g., Dym and Shames, 2013). Eqs. (28) and (31) are the same cut off from a complete beam so that the boundary layer is not mod-
as in the static Levinson beam theory. The provided exact 1D interior eled. Thus, we obtain for the work due to the stresses on the lateral
presentation enables a more thorough study of the pros and cons of end surfaces of the interior beam
 h/2
approximate interior beam theories, such as those of Levinson and
Timoshenko, on the level of governing differential equations instead
Ws = t σx (L/2, y)Ux (L/2, y)dy
−h/2
of simplistic (numerical) comparisons between specific 2D elasticity  h/2
and 1D beam solutions. Following the methodology presented above −t σx ( − L/2, y)Ux ( − L/2, y)dy
for the uniform load, 1D beam presentations may be obtained for −h/2
 h/2
more complicated loads by updating the load-dependent part of the
stress function (5). +t τxy (L/2, y)Uy (L/2, y)dy
−h/2
 h/2

3.3. Example – Simply-supported beam revisited

−t τxy ( − L/2, y)Uy ( − L/2, y)dy. (35)
−h/2

By substituting the polynomial expressions according to the general

We are now able to solve a 2D interior plane beam problem in
solution presented in Section 2 for σ x ,  x , σ y ,  y , τ xy , γ xy , Ux and Uy
terms of the 1D beam presentation. Using the load resultants given by
into (34) and (35) we find that
Eqs. (13) and the interior beam conditions uy ( ± L/2) = 0 and ux (0) =
0, integration of Eqs. (26)–(28) yields for the central axis variables the 2U − Wq − Ws = 0. (36)
expressions
The above calculation shows that in static equilibrium the strain en-
qν h3 x ergy of the beam is equal to one-half of the work done by the surface
ux = , tractions if they were to move (slowly) through their respective dis-
24EI
q   placements. This is exactly according to the Clapeyron’s theorem (e.g.,
uy = − (L2 − 4x2 ) 5(5L2 − 4x2 ) + 6h2 (8 + 5ν) , Sadd, 2014). In a recent paper (Karttunen and von Hertzen, 2015), in
1920EI
qx   contrast to the long-standing belief that the interior beam theory by
φ=− 15L2 − 20x2 − 3h2 (2 + 5ν) . Levinson (1981) is “variationally inconsistent”, we provided a consis-
120EI
tent variational formulation for the Levinson beam by accounting for
Substitution of the central axis variables into the 2D displacements the external (virtual) work (35). The resulting beam equations were
(21) and (22) results in the exact 2D interior displacement field. Al- exactly the same as those in the case of the vectorial derivation by
ternatively, to obtain the 2D displacements, we can solve uy and φ Levinson (1981). The work (35) is an integral part of all energy-based
from Eqs. (32) and (33) with the interior boundary conditions considerations related to interior beams.
∂φ qh2
uy ( ± L/2) = 0, M( ± L/2) = 0 → ( ± L/2) = (2 + 5ν). 3.5. Exact rod and beam finite elements
∂x 40EI
Detailed end effects are out of the scope of this study, but one should The polynomial 2D interior elasticity solution presented in
keep in mind that the moment-related boundary conditions at the Section 2 can be used to derive exact 1D interior rod and beam finite
beam ends might take a different form in a solution which includes elements. Fig. 2 presents the setting according to which the elements
the boundary layer behavior. Note also that, unlike in the Timoshenko are developed. Both nodes in Fig. 2 have three degrees of freedom.
beam theory, the distributed load q has an effect on the moment- For nodes i = 1, 2, we have axial displacements ux, i , transverse dis-
related boundary condition. placements uy, i and rotations of the cross-section φ i . By the aid of
 A.T. Karttunen, R. von Hertzen / International Journal of Solids and Structures 78-79 (2016) 125–130 129

the central axis variables (18)–(20), we obtain for nodes 1 and 2 six where the nodal displacement vector is
equations T
u = ux,1 uy,1 φ1 ux,2 uy,2 φ2 . (47)
ux,1 = ux ( − L/2), ux,2 = ux (L/2),
In addition we have
uy,1 = uy ( − L/2), uy,2 = uy (L/2), (37)
xy 2  
φ1 = −φ( − L/2), φ2 = −φ(L/2). L1 = − L − 4 x2 − y2 (2 + ν) , (48)
24EI
We can solve the six unknowns c1 , c2 , c3 , C, D1 and D2 from Eqs. (37). 1  
Explicit expressions for these are given in Appendix A. To obtain the L2 = 16h3 y(ν 2 − 1) − 4L2 L2 − 4x2 − 2y2 (ν − 1)
384EI
finite element equations, we substitute c1 , c2 and c3 into Eqs. (9) to   
calculate the load resultants at nodes i = 1, 2, with the notion that + 8ν y2 L2 + 4(y2 − 3x2 ) − (L2 − 4x2 )2 + 16y4 . (49)
the positive directions are taken to be according to Fig. 1 so that The shape functions Nx and Ny are given in Appendix A. Once the
N1 = −N( − L/2), N2 = N(L/2), nodal displacements have been solved from Eqs. (41) and (42), the
exact interior 2D displacement field can be calculated by substitut-
Q1 = −Q ( − L/2), Q2 = Q (L/2), (38)
ing the nodal displacements into Eqs. (45) and (46), after which the
M1 = M( − L/2), M2 = −M(L/2). calculation of 2D interior strains and stresses is straightforward.
The conventional presentation for the 1D rod and beam elements is
obtained by writing Eqs. (38) in matrix form. Before doing so, we also 4. Conclusions
derive the FE equations from the total potential energy
In this paper, a general interior elasticity solution for a 2D plane

= U − Wq − Ws . (39) beam under a uniform load was derived. It was shown that the solu-
The stresses on the end surfaces in Eq. (35) are written as given by tion can be presented in the form of a conventional 1D beam theory
Eqs. (11) and (12), where the load resultants are expressed as nodal by the aid of kinematic variables defined at the central axis of the 2D
forces according to Eqs. (38). Then, by calculating (34) and (35) and beam. In addition, the solution can be presented in the form of rod
by applying the principle of minimum total potential energy and beam finite elements. The presentation of the 2D interior elas-
ticity solution as a 1D beam theory offers a new point of view on 2D
∂
∂
∂
plane beam solutions and reveals the underlying structure of an ex-
= 0, = 0, = 0 (i = 1, 2) (40)
∂ ux,i ∂ uy,i ∂φi act 1D interior beam. It was also shown that in all energy-based con-
siderations related to the interior beam one has to take into account
we obtain the finite element equilibrium equations. The force-based
the fact that the interior stresses do work on the lateral end surfaces
method and the total potential energy approach result in the same
of the interior beam. Many higher-order beam theories can be found
equations, which can be written in the form
       in the literature which are founded exclusively on interior kinemat-
EA 1 −1 ux,1 N1 qν h 1 ics (e.g. Bickford, 1982). However, these higher-order constructions
= − , (41) are incomplete as interior beam theories because they lack the afore-
L −1 1 ux,2 N2 2 −1
mentioned work due to the stresses at the beam ends.
⎡ ⎤
12 6L −12 6L
Appendix A. Equations for finite element developments
EI ⎢ 6L (4 + )L2 −6L (2 − )L2 ⎥
(1 + )L3 ⎣−12 −6L 12 −6L ⎦
We obtain from Eqs. (37) the following relations
6L (2 − )L2 −6L (4 + )L2
⎧ ⎫ E qν
⎧ ⎫ ⎧ ⎫ ⎪ L ⎪ c1 = (ux,2 − ux,1 ) − ,
⎪ ⎪ ⎪
⎨uy,1 ⎪
⎬ ⎪ ⎨ Q1 ⎪
⎬ ⎪
⎨ 10L2 −3h2 (2+5ν) ⎪
⎬ 2L 4t
φ1 M1 q 60 E q
× = − (42) c2 = (φ1 − φ2 ) − (L2 + 3ν h2 ), (A.1)
⎪uy,2 ⎭
⎩ ⎪ ⎪ ⎩ Q2 ⎪
⎭ 2⎪⎪ L ⎪
⎪ 6L 144I
φ2 M2 ⎪
⎩ ⎪
10L2 −3h2 (2+5ν) ⎭  
− 3Eh2
60 c3 = 3 2(uy,1 − uy,2 ) + L(φ1 + φ2 ) ,
4L (1 + )
where  = 3h2 (1 + ν)/L2 . Eqs. (41) and (42) can be written concisely
E
as C=− 4L2 (uy,1 − uy,2 )
4L3 (1 + )
K r u r = fr − q r , (43)
 
+ L2 6(uy,1 − uy,2 ) + L(φ1 + φ2 ) ,
K b u b = fb − q b , (44) E
D1 = (ux,1 + ux,2 ),
2
E  qL4 (1 + 4)
where Kr and Kb are the rod and beam element stiffness matrices, re-
spectively, and ur and ub are the rod and beam nodal displacement D2 = 4(uy,1 + uy,2 ) + L(φ1 − φ2 ) − , (A.2)
vectors, respectively. The force vectors are fr and qr for the rod ele-
8 384I
ment, and fb and qb for the beam element, respectively. Note that the where  = 3h2 (1 + ν)/L2 . The shape functions in Eqs. (45) and (46)
stiffness matrix Kb is the same as for the Levinson beam (Karttunen are
and von Hertzen, 2015) and also equal to the stiffness matrix of a ⎧ 1 ⎫
⎪ − xL ⎪
Timoshenko beam element when the Timoshenko shear coefficient ⎪
⎪
2
y[3L −12x +4y (2+ν)] ⎪
⎪
⎪ ⎪
2 2 2

has the value of 2/3 (e.g., cf. Kosmatka, 1995). Finally, by substituting ⎪
⎪ 2L3 (1+) ⎪
⎪
⎪
⎪ y{−2L (L−2x)+L[(L−2x)(L+6x)+4y (2+ν)]} ⎪
⎪
⎨ ⎬
2 2
c1 , c2 , c3 , C, D1 and D2 (Appendix A) into the displacements (16) and
3 4L (1+)
(17) we can write the 2D displacements in the form NTx = , (A.3)
⎪
⎪
1
+ xL ⎪
⎪
Ux (x, y) = Nx u + qL1 , (45) ⎪
⎪
2 ⎪
⎪
⎪
⎪ −
y[3L2 −12x2 +4y2 (2+ν)] ⎪
⎪
⎪
⎪ ⎪
⎩ y{−2L2 (L+2x)+L[(L+2x)(L−6x)+4y2 (2+ν)]} ⎪
2L (1+)
3
⎭
Uy (x, y) = Ny u + qL2 , (46)
4L3 (1+)
130 A.T. Karttunen, R. von Hertzen / International Journal of Solids and Structures 78-79 (2016) 125–130

⎧ yν
⎫
⎪
⎪ ⎪
⎪
Kosmatka, J.B., 1995. An improved two–node finite element for stability and natural
⎪
⎪
L
L (L−2x)+12ν xy2 +(L+x)(L−2x)2
2 ⎪
⎪ frequencies of axial–loaded Timoshenko beams. Comput. Struct. 57 (1), 141–149.
⎪
⎪ 2L (1+) ⎪
⎪ Levinson, M., 1981. A new rectangular beam theory. J. Sound Vib. 74 (1), 81–87.
⎪ ⎪
3

⎪
⎨ L2 [L2 −4(x2 +ν y2 )]−4ν L(L−6x)y2 +L(L+2x)(L−2x)2 ⎪
⎬ Reddy, J.N., 1984. A simple higher–order theory for laminated composite plates. J. Appl.
8L3 (1+) Mech. 51 (4), 745–752.
NTy = . (A.4) Reddy, J.N., 1990. A general nonlinear 3rd–order theory of plates with moderate thick-
⎪
⎪ − yLν ⎪
⎪
⎪
⎪  L2 (L+2x)−12ν xy2 +(L−x)(L+2x)2 ⎪
⎪
ness. Int. J. Nonlin. Mech. 25 (6), 677–686.
⎪
⎪ ⎪
⎪ Reddy, J.N., 2003. Mechanics of Laminated Composite Plates and Shells: Theory and
⎪
⎪
2L (1+)
3
⎪
2 ⎪
Analysis, 2nd ed. CRC Press, Boca Raton, Florida.
⎩ [
L2
−L2
+4(x 2
+ ν y2
)] +4ν L(L+6x )y2
−L(L−2x )(L+2x ) ⎭ Rehfield, L.W., Murthy, P.L.N., 1982. Toward a new engineering theory of bending –
8L3 (1+) fundamentals. AIAA J. 20 (5), 693–699.
Sadd, M.H., 2014. Elasticity – Theory, Applications and Numerics, 3rd ed. Academic
References Press, Oxford.
Schneider, P., Kienzler, R., 2015. On exact rod/beam/shaft-theories and the coupling
Barber, J.R., 2010. Elasticity, 3rd ed. Springer, New York. among them due to arbitrary material anisotropies. Int. J. Solids Struct. 56, 265–
Bickford, W.B., 1982. A consistent higher order beam theory. Dev. Theor. Appl. Mech. 279.
11, 137–150. Soutas-Little, R.W., 1973. Elasticity. Dover, New York.
Dym, C.L., Shames, I.H., 2013. Solid Mechanics – A Variational Approach (Augmented Timoshenko, S.P., 1921. On the correction for shear of the differential equation for trans-
Edition). Springer, New York. verse vibration of prismatic bars. Philos. Mag. 41, 744–746.
Heyliger, P.R., Reddy, J.N., 1988. A higher order finite element for bending and vibration Timoshenko, S.P., Goodier, J.N., 1970. Theory of Elasticity, 3rd ed. McGraw-Hill, Singa-
problems. J. Sound Vib. 126 (2), 309–326. pore.
Jemielita, G., 1990. On kinematical assumptions of refined theories of plates – a survey. Vlasov, B.F., 1957. On equations of bending of plates (in Russian). Dokla Ak. Nauk Azer-
J. Appl. Mech. 57 (4), 1088–1091. beijanskoi SSR 3, 955–959.
Karttunen, A.T., von Hertzen, R., 2015. Variational formulation of the static Levinson
beam theory. Mech. Res. Commun. 66, 15–19.