ONE_DIMENSIONAL Elements

Examples of 1D elements and assembly

One-Dimensional Elements 1 / 20
 Introduction
1-D elements include a straight rod loaded axially, a straight beam loaded laterally, an element that conducts
heat or electricity, and so on. In structural terminology, a ROD can resist only axial load, whereas a BAR, in its
most general sense, can resist axial, lateral and twisting loads. In time-independent (static) analysis, a truss of 𝑛𝑛
members can be modeled by 𝑛𝑛 rod elements, and a frame having 𝑛𝑛 straight members, usually requires 𝑛𝑛 bar
elements. Thus, when one-dimensional elements are used for static analysis, the discretization phase of
modeling becomes trivial,a nd for stress analysis the name matrix method of structural mechanics may be used
in preference to FEA. However, rod and bar elements are provided in FEA software and are much used, both as
stand-alone elements and in combination with finite elements of other types. For example, bar elements can be
attached to plate elements to model stiffened plates.

Our attention is now restricted to linear problems, which means that material properties are essentially
unchanged by loading, deformations are small enough that equilibrium equations can be written using original
geometry rather than deformed geometry. Only steady-state problems are considered, which are called static or
(more properly) quasistatic in structural mechanics.

A finite element has a characteristic matrix, which is the stiffness matrix for load-deformation analysis, a
conductivity matrix for heat conduction analysis, and so on.

One-Dimensional Elements 2 / 20
 Introduction
1-D element often incorporates the exact variation of the field quantity. For example, the usual mathematical
model of a uniform beam loaded by concentrated lateral forces displays a cubic variation of lateral
displacement between load points. The standard bar element is also based on a cubic field, so an FE model built
of bar elements, with nodes at load points, provides results in exact agreement with the mathematical model.
Exact agreement is generally not achieved by an FE model of a plane or solid continuum, where element
displacement fields are only approximate.
Regardless of the number or types of elements used, the computational procedure for static FEA is as follows:
1. Generate matrices that describe element behavior
2. Connect elements together, assemblying the element matrices to obtain a structure matrix
3. Provide some nodes with loads (check that instead of nodes, it could be better to use d.o.f)
4. Provide other nodes with boundary conditions
5. The structure matrix and the array of loads are parts of a system of algebraic equations.
Solve these equations to determine nodal values of field quantities
6. Compute gradients (strain and next stress in structural mechanics)

1-D elements will be used as vehicle for the explanation of these procedures

One-Dimensional Elements 3 / 20
 ROD Element
Consider a uniform prismatic elastic rod element of length L and elastic modulus E. Often a rod element is
represented as a line, but the element has a cross-sectional Area A. A node is located at each end.
For now, nodes are allowed to displace only in the axial direction.

Axial displacements at nodes are 𝑢𝑢1 and 𝑢𝑢2 . Internal axial stress σ can be
related to nodal forces 𝐹𝐹1 and 𝐹𝐹2 by free-body diagram. The expression for 𝜖𝜖
is reported below the figure with the consequent relationships which can be
written as (nodal forces and nodal displacements are positive in the same
direction):

𝐹𝐹1 + 𝐴𝐴𝜎𝜎 = 0 𝐹𝐹2 − 𝐴𝐴𝜎𝜎 = 0 𝐴𝐴𝐴𝐴

𝑢𝑢1 − 𝑢𝑢2 = 𝐹𝐹1
𝑢𝑢2 − 𝑢𝑢1 𝐿𝐿
or 𝑘𝑘 −𝑘𝑘 𝑢𝑢1 𝐹𝐹1
𝜎𝜎 = 𝐸𝐸𝐸𝐸 𝜖𝜖 = =
𝐿𝐿 𝐴𝐴𝐴𝐴 −𝑘𝑘 𝑘𝑘 𝑢𝑢2 𝐹𝐹2 1
𝑢𝑢2 − 𝑢𝑢1 𝑢𝑢2 − 𝑢𝑢1 = 𝐹𝐹2
𝐹𝐹1 + 𝐴𝐴𝐴𝐴 =0 𝐿𝐿
𝐿𝐿
𝑢𝑢2 − 𝑢𝑢1 where 𝐸𝐸𝐸𝐸
𝐹𝐹2 − 𝐴𝐴𝐴𝐴 =0 𝑘𝑘 =
𝐿𝐿 𝐿𝐿

One-Dimensional Elements 4 / 20
 ROD Element
Obviously, for this element, equilibrium requires 𝐹𝐹1 = −𝐹𝐹2 . The matrix equation is abbreviated as 𝑘𝑘 𝑑𝑑 = 𝐹𝐹
where 𝑘𝑘 is called the element stiffness matrix.
Note that 𝐸𝐸𝐸𝐸⁄𝐿𝐿 can be regarded as 𝑘𝑘, the stiffness of a linear spring. A rod and a spring have the same behavior
under axial load and are represented by the same stiffness matrix.
In the stiffness matrix just introduced, it is possible to see an instance of the following general rule:

A column of 𝑘𝑘 is the vector of loads that must be applied to an element at its nodes to
maintain a deformation state in which the corresponding nodal d.o.f. has unit value
while all the other nodal d.o.f.are zero

Let 𝑢𝑢1 = 0 and 𝑢𝑢2 = 1 we have

𝐹𝐹1 𝑘𝑘 −𝑘𝑘 0 −1 𝐸𝐸𝐸𝐸 −1
= = 𝑘𝑘 =
𝐹𝐹2 −𝑘𝑘 𝑘𝑘 1 1 𝐿𝐿 1

which, in other words, allows the definition of the stiffness matrix of an element, from the knowledge of the
stress inyternal characteristics, assuming one d.o.f. per time egual to unity.

One-Dimensional Elements 5 / 20
 Structure Equations
Consider a structure built of two uniform elastic rods attached end to end
Only axial displacement are allowed. Stiffnesses of the respective elements are 𝑘𝑘1 and 𝑘𝑘2 . The structure
stiffness equation is
𝐹𝐹1 𝑘𝑘1 −𝑘𝑘1 0 𝑢𝑢1
𝐹𝐹2 = −𝑘𝑘1 𝑘𝑘1 + 𝑘𝑘2 −𝑘𝑘2 𝑢𝑢2 or 𝐹𝐹 = 𝐾𝐾 𝐷𝐷
𝐹𝐹3 0 −𝑘𝑘2 𝑘𝑘2 𝑢𝑢3
𝐾𝐾 is called either the structure stiffness matrix or the global stiffness matrix. It can be easily obtained by
applying the general rule stated above, activating (equating to unity) one d.o.f. in turn and leaving the other two
equal to zero.
An alternative way to obtain 𝐾𝐾 is as follows. Imagine that the two elements are not yet connected, but are
provided with numbered ends: 1 and 2 for element 1 and 2 and 3 for element 2. Separately, when expanded to
“structure size”, stiffness matrices for elements 1 and 2 can be written
𝑢𝑢1 𝑢𝑢2 𝑢𝑢3 𝑢𝑢1 𝑢𝑢2 𝑢𝑢3
𝑘𝑘1 −𝑘𝑘1 0 0 0 0
−𝑘𝑘1 𝑘𝑘1 0 for the element 1, and 0 𝑘𝑘2 −𝑘𝑘2 for the element 2
0 0 0 0 −𝑘𝑘2 𝑘𝑘2

One-Dimensional Elements 6 / 20
 Structure Equations
Added zeros in the matrix for (say) element 1 can be explained as follows: the row of zeros indicates that
because element 1 is not connected to node 3, displacement of any node cannot cause element 1 to produce a
force at node 3; the column of zeros displacement 𝑢𝑢3 does not strain element 1. Clearly summation of thw two
matrices of each element, written in the “structural size” produces the previous global stiffness matrix.
In general, one can imagine a physical space, initially empty except for the numbered nodes in their proper
positions, that becomes populated by addition of stiffness coefficients from elements. This process of building a
structure matrix 𝐾𝐾 from constituent element matrices 𝑘𝑘 is called assembly.
Regarding the support conditions, more generally called boundary conditions, reference should be given to the
fact that 𝑢𝑢1 = 0 and therefore is known an therefore its relative equation does not need to be considered. This
results in the following final stiffness matrix that relates the remaining active d.o.f. to the applied forces:
𝑘𝑘1 + 𝑘𝑘2 −𝑘𝑘2 𝑢𝑢2 𝐹𝐹2
𝑢𝑢3 =
−𝑘𝑘2 𝑘𝑘2 𝐹𝐹3
from which 𝑢𝑢2 e 𝑢𝑢3 can be determined when loads 𝐹𝐹2 and 𝐹𝐹3 are prescribed.
When 𝐾𝐾 is assembled for an FE structure, however complex, computer software uses the summation process
just described, but without formal expansion of element matrices to “structure size”. Dedicated algorithm are
used for the assembly process.

One-Dimensional Elements 7 / 20
 BAR Element
Consider a uniform bar element, along the 𝑥𝑥 axis and into the 𝑥𝑥𝑥𝑥 plane. Such 2D element has a node at each
end. Each node has two d.o.f., namely, lateral translation and rotation.

To begin, lateral displacements are restricted to the 𝑥𝑥𝑥𝑥

plane and bending deformations only are considered,
assuming that transverse shear deformation can be
ignored. These are the basic hypotheses of the
elementary beam theory, which is more formally known
as Euler-Bernoulli beam theory. Transverse shear
deformation is taken into account by the Timoshenko
beam theory.
With axial deformation d.o.f. omitted, the stiffness
matrix of the bar element is 4 by 4. It can be
constructed column by column.
The 𝑗𝑗th column is the vector of nodal loads associated
with unit value of the 𝑗𝑗th d.o.f. and zero values for all
other d.o.f. The load vector contains moments as well
as forces.
One-Dimensional Elements 8 / 20
 BAR Element
To obtain the column associated with 𝑣𝑣1 it is convenient to refer to the following figure for calculating the
nodal loads:

These loads are named 𝑘𝑘11 , 𝑘𝑘21 , 𝑘𝑘31 and 𝑘𝑘41 to indicate that they appear in rows 1, 2, 3, 4 and in column 1 of
element stiffness matrix 𝑘𝑘 . Loads are shown in the positive sense, that is, same direction as their associated
d.o.f. To obtain 𝑘𝑘11 and 𝑘𝑘21 , from handbook formulas of beam elementary theory by regarding the beam as a
cantilever beam fixed at node 2 and loaded at node 1 by force 𝑘𝑘11 and moment 𝑘𝑘21 such that 𝑣𝑣1 = 1 and
𝜃𝜃𝑧𝑧1 = 0. Thus
𝑘𝑘11 𝐿𝐿3 𝑘𝑘21 𝐿𝐿2 𝑘𝑘11 𝐿𝐿2 𝑘𝑘21 𝐿𝐿
𝑣𝑣1 = 1 ⟹ − =1 and 𝜃𝜃𝑧𝑧𝑧 = 0 ⟹ − + =0
3𝐸𝐸𝐼𝐼𝑧𝑧 2𝐸𝐸𝐼𝐼𝑧𝑧 2𝐸𝐸𝐼𝐼𝑧𝑧 𝐸𝐸𝐼𝐼𝑧𝑧
where 𝐸𝐸 is the elastic modulus and 𝐼𝐼𝑧𝑧 is the moment of inertia of the beam cross-sectional area about a
centroidal axis parallel to the 𝑧𝑧 axis. The above equations yield
12𝐸𝐸𝐼𝐼𝑧𝑧 6𝐸𝐸𝐼𝐼𝑧𝑧
𝑘𝑘11 = 𝑘𝑘21 = 2
𝐿𝐿3 𝐿𝐿

One-Dimensional Elements 9 / 20
 BAR Element
Being 𝑘𝑘11 and 𝑘𝑘21 known, 𝑘𝑘31 and 𝑘𝑘41 can be determined from considerations of static equilibrium. Summing
𝑦𝑦 −direction forces and moments about node 2, it results:
𝑘𝑘11 + 𝑘𝑘31 = 0 12𝐸𝐸𝐼𝐼𝑧𝑧 6𝐸𝐸𝐼𝐼𝑧𝑧
⟹ 𝑘𝑘31 =− 𝑘𝑘41 = 2
𝑘𝑘21 + 𝑘𝑘41 − 𝑘𝑘11 𝐿𝐿 = 0 𝐿𝐿3 𝐿𝐿

Similar analysis of the latter three d.o.f. provides terms for the other three columns of element stiffness
matrix 𝑘𝑘 . The complete 2D bar element stiffness is

12𝐸𝐸𝐼𝐼𝑧𝑧 6𝐸𝐸𝐼𝐼𝑧𝑧 −12𝐸𝐸𝐼𝐼𝑧𝑧 6𝐸𝐸𝐼𝐼𝑧𝑧

𝑣𝑣1 The column of symbols on the right is appended
𝐿𝐿3 𝐿𝐿2 𝐿𝐿3 𝐿𝐿2
6𝐸𝐸𝐼𝐼𝑧𝑧 4𝐸𝐸𝐼𝐼𝑧𝑧 −6𝐸𝐸𝐼𝐼𝑧𝑧 2𝐸𝐸𝐼𝐼𝑧𝑧 merely to indicate that 𝑘𝑘 operates on the column
𝜃𝜃𝑧𝑧𝑧
𝐿𝐿2 𝐿𝐿 𝐿𝐿2 𝐿𝐿 vector of element d.o.f. 𝑑𝑑 = 𝑣𝑣1 𝜃𝜃𝑧𝑧1 𝑣𝑣2 𝜃𝜃𝑧𝑧2 𝑇𝑇 .
𝑘𝑘 =
−12𝐸𝐸𝐼𝐼𝑧𝑧 −6𝐸𝐸𝐼𝐼𝑧𝑧 12𝐸𝐸𝐼𝐼𝑧𝑧 −6𝐸𝐸𝐼𝐼𝑧𝑧 𝑣𝑣 A different ordering in 𝑑𝑑 would change the ordering
2
𝐿𝐿3 𝐿𝐿2 𝐿𝐿3 𝐿𝐿2 of coefficients in 𝑘𝑘 but not their numerical values.
6𝐸𝐸𝐼𝐼𝑧𝑧 2𝐸𝐸𝐼𝐼𝑧𝑧 −6𝐸𝐸𝐼𝐼𝑧𝑧 4𝐸𝐸𝐼𝐼𝑧𝑧 𝜃𝜃𝑧𝑧𝑧
𝐿𝐿2 𝐿𝐿 𝐿𝐿2 𝐿𝐿

One-Dimensional Elements 10 / 20
 BAR Element - Generalizations
To allow the bar element to stretch as well as bend, axial translation are added to the array of nodal d.o.f. and
expand 𝑘𝑘 to size 6 by 6 by including axial stiffness coefficients 𝐸𝐸𝐸𝐸⁄𝐿𝐿. Additionally bending stiffness terms
may be modified to account for transverse shear deformation, thus producing a Timoshenko beam element:

𝑋𝑋 0 0 −𝑋𝑋 0 0 𝑢𝑢1 𝐸𝐸𝐸𝐸 12𝐸𝐸𝐼𝐼𝑧𝑧 6𝐸𝐸𝐼𝐼𝑧𝑧

0 𝑌𝑌1 𝑌𝑌2 0 −𝑌𝑌1 𝑌𝑌2 𝑣𝑣1 𝑋𝑋 = 𝑌𝑌1 = 𝑌𝑌2 =
𝐿𝐿 1 + 𝜙𝜙𝑦𝑦 𝐿𝐿3 1 + 𝜙𝜙𝑦𝑦 𝐿𝐿2
0 𝑌𝑌2 𝑌𝑌3 0 −𝑌𝑌2 𝑌𝑌4 𝜃𝜃𝑧𝑧𝑧
𝑘𝑘 = 𝑢𝑢2 where
−𝑋𝑋 0 0 𝑋𝑋 0 0
0 −𝑌𝑌1 −𝑌𝑌2 0 𝑌𝑌1 −𝑌𝑌2 𝑣𝑣2 4 + 𝜙𝜙𝑦𝑦 𝐸𝐸𝐼𝐼𝑧𝑧 2 − 𝜙𝜙𝑦𝑦 𝐸𝐸𝐼𝐼𝑧𝑧 12𝐸𝐸𝐼𝐼𝑧𝑧 𝑘𝑘𝑦𝑦
0 𝑌𝑌2 𝑌𝑌4 0 −𝑌𝑌2 𝑌𝑌3 𝜃𝜃𝑧𝑧𝑧 𝑌𝑌3 = 𝑌𝑌4 = 𝜙𝜙𝑦𝑦 =
1 + 𝜙𝜙𝑦𝑦 𝐿𝐿 1 + 𝜙𝜙𝑦𝑦 𝐿𝐿 𝐴𝐴𝐴𝐴𝐿𝐿2

and 𝐴𝐴⁄𝑘𝑘𝑦𝑦 is the effective shear area for transverse shear deformation in the 𝑦𝑦 direction. Commonly accepted
factors are 𝑘𝑘𝑦𝑦 = 1.2 for a solid rectangular cross section, 𝑘𝑘𝑦𝑦 = 2.0 for a thin-walled tube of circular cross
section, and so on. Note that as an element becomes more and more slender, 𝜙𝜙𝑦𝑦 approaches zero, so that the
flexure coefficients 𝑌𝑌𝑖𝑖 reduce to coefficients of the Euler-Bernoulli beam theory, where transverse shear
deformation is neglected.

One-Dimensional Elements 11 / 20
 BAR Element – Generalizations – 3-D
The 3D bar element allows six d.o.f per node, as depicted in the figure below. Three translations and three
rotations. The 𝑤𝑤 and 𝜃𝜃𝑦𝑦 d.o.f. account for lateral deflection in the 𝑧𝑧𝑧𝑧 plane. The 𝜃𝜃𝑥𝑥 d.o.f. account for twist
about the 𝑥𝑥 axis, for which the stiffness coefficient is 𝐺𝐺𝐺𝐺 ⁄𝐿𝐿, where 𝐾𝐾 is a property of the shape and size of
the cross section. Only for a circular cross section, either solid or a tube, does 𝐾𝐾 become equal to 𝐽𝐽, the polar
moment of inertia of the cross sectional area about its centroid. For thin-walled open cross sections, such as
those of standard I-beams and channels, 𝐾𝐾 is a small fraction of 𝐽𝐽.
Considering the following definitions 𝑆𝑆 = 𝐺𝐺𝐺𝐺 ⁄𝐿𝐿, 𝑋𝑋 and 𝑌𝑌𝑖𝑖
terms as defined previously, and 𝑍𝑍𝑖𝑖 terms defined as for 𝑌𝑌 and
𝜙𝜙, but with interchange of subscripts. For example

12𝐸𝐸𝐼𝐼𝑦𝑦 12𝐸𝐸𝐼𝐼𝑦𝑦 𝑘𝑘𝑧𝑧

𝑍𝑍1 = and 𝜙𝜙𝑧𝑧 =
1 + 𝜙𝜙𝑧𝑧 𝐿𝐿3 𝐴𝐴𝐴𝐴𝐿𝐿2

The 𝑘𝑘 matrix, partitioned by nodes for the sake of clarity, is

One-Dimensional Elements 12 / 20
 BAR Element – Generalizations – 3-D

𝑋𝑋 0 0 0 0 0 −𝑋𝑋 0 0 0 0 0 𝑢𝑢1
𝑌𝑌1 0 0 0 𝑌𝑌2 0 −𝑌𝑌1 0 0 0 𝑌𝑌2 𝑣𝑣1
𝑍𝑍1 0 −𝑍𝑍2 0 0 0 −𝑍𝑍1 0 −𝑍𝑍2 0 𝑤𝑤1
𝑆𝑆 0 0 0 0 0 −𝑆𝑆 0 0 𝜃𝜃𝑥𝑥𝑥
𝑍𝑍3 0 0 0 𝑍𝑍2 0 𝑍𝑍4 0 𝜃𝜃𝑦𝑦1
𝑌𝑌3 0 −𝑌𝑌2 0 0 0 𝑌𝑌4 𝜃𝜃𝑧𝑧1
𝑘𝑘 = 𝑢𝑢2
𝑋𝑋 0 0 0 0 0
𝑌𝑌1 0 0 0 −𝑌𝑌2 𝑣𝑣2
𝑍𝑍1 0 𝑍𝑍2 0 𝑤𝑤2
𝑠𝑠 𝑦𝑦 𝑚𝑚 𝑚𝑚 𝑆𝑆 0 0 𝜃𝜃𝑥𝑥2
𝑍𝑍3 0 𝜃𝜃𝑦𝑦𝑦
𝑌𝑌3 𝜃𝜃𝑧𝑧𝑧

One-Dimensional Elements 13 / 20
 BAR Element - Cautions
It is assumed in the preceding development that the axes 𝑦𝑦 and 𝑧𝑧 are principal centroidal axes of the cross section. This
assumption is true in the elementary case where 𝑦𝑦 = 𝑧𝑧 = 0 at the centroid and the cross section has at least one
symmetry axis that is coincident with either axis 𝑦𝑦 or axis 𝑧𝑧. If the cross section is unsymmetric, principal axis must be
established by calculation. If 𝑦𝑦 nd 𝑧𝑧 were not principal axes, different terms would appear in the stiffness matrix 𝑘𝑘 , and
there would be fewer zero coefficients, because activation of a d.o.f. would produce more nodal forces and moments.

When a member of noncircular cross section is twisted, cross sections warp; that is, cross sections do not remain plane.
Channels, I beams and other thin-walled open members have low torsional stiffness, and their cross sections tend to
warp appreciably when the member is twisted. Restraint of warping may have a large effect. For example, let member
𝐴𝐴𝐴𝐴 of the following figure be an I beam five times as long as it is deep, with a torque 𝑇𝑇 applied at end 𝐵𝐵.
Complete restraint of warping at end 𝐴𝐴 reduces rotation at 𝐵𝐵 by about half and introduces
axial normal stresses much larger than torsional shear stress.
To account for full or partial restraint of warping, another d.o.f. must be added at each
node, so that 𝑘𝑘 for 3D beam element becomes a 14 by 14 matrix. The added d.o.f. are
rates of twist 𝑑𝑑𝜃𝜃𝑥𝑥 ⁄𝑑𝑑𝑑𝑑 at nodes, each associated with a load term called bimoment.
Commercial software typically allow only six displacement d.o.f. per node, so restraint of
warping is ignored in beam elements.

One-Dimensional Elements 14 / 20
 Rod and BAR Elements of Arbitrary Orientation
Elements discussed previously lie along the 𝑥𝑥 axis. For general use, elements must be capable of assuming any orientation in
space. This capability can be provided by simple manipulation of element matrices already derived.
For this purpose, we regard the previous siffness matriuces as stiffness matrices 𝑘𝑘 ′ that lie on the 𝑥𝑥 ′ axis of a local
coordinate system 𝑥𝑥 ′ 𝑦𝑦 ′ 𝑧𝑧 ′ , which is arbitrarily oriented in a global coordinate system 𝑥𝑥𝑥𝑥𝑥𝑥. To obtain an element matrix 𝑘𝑘
that operates on d.o.f. referred to global coordinates 𝑥𝑥𝑥𝑥𝑥𝑥, a rotational coordinate transformation must be applied to 𝑘𝑘 ′ .

Two-dimensional transformation. Starting with a rod element along a local axis 𝑥𝑥 ′ in the 𝑥𝑥𝑥𝑥 plane, we seek the stiffness
matrix of a bar element arbitrarily oriented in the 𝑥𝑥𝑥𝑥 plane, so that it operates on the four nodal d.o.f. as shown in figure:

One-Dimensional Elements 15 / 20
 Rod and BAR Elements of Arbitrary Orientation
The relation between local and global d.o.f. is easily written by resolving global d.o.f. into local component at a node, then
adding collinear components; for example
𝑢𝑢1′ = 𝑢𝑢1 𝑐𝑐𝑐𝑐𝑐𝑐𝛽𝛽 + 𝑣𝑣1 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑢𝑢2′ = 𝑢𝑢2 𝑐𝑐𝑐𝑐𝑐𝑐𝛽𝛽 + 𝑣𝑣2 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠

In matrix format, displacement arrays have the relation 𝑑𝑑 ′ = 𝑇𝑇 𝑑𝑑 in which transformation matrix 𝑇𝑇 is

𝑐𝑐 𝑠𝑠 0 0
𝑇𝑇 = where 𝑐𝑐 = 𝑐𝑐𝑐𝑐𝑐𝑐𝛽𝛽 and 𝑠𝑠 = 𝑠𝑠𝑠𝑠𝑠𝑠𝛽𝛽
0 0 𝑐𝑐 𝑠𝑠

Similarly for the corresponding forces 𝑟𝑟 ′ = 𝑇𝑇 𝑟𝑟

Because 𝑟𝑟 and 𝑟𝑟 ′ describe the same resultant force, work done during a virtual displacement must be independent of
the coordinate system in which the work is computed. Let 𝛿𝛿𝑑𝑑 and 𝛿𝛿𝑑𝑑′ be descriptions of the same virtual displacement,
in the global and local coordinate systems, respectively. Writing the virtual work equality, we obtain:

One-Dimensional Elements 16 / 20
 Rod and BAR Elements of Arbitrary Orientation

𝛿𝛿𝑑𝑑 𝑇𝑇 𝑟𝑟 = 𝛿𝛿𝑑𝑑 ′ 𝑇𝑇 𝑟𝑟 ′ being 𝑑𝑑 ′ = 𝑇𝑇 𝑑𝑑 ⟹ 𝛿𝛿𝛿𝛿′ 𝑇𝑇 = 𝛿𝛿𝛿𝛿 𝑇𝑇 𝑇𝑇 𝑇𝑇 we can write

𝛿𝛿𝑑𝑑 𝑇𝑇
𝑟𝑟 = 𝛿𝛿𝑑𝑑 𝑇𝑇
𝑇𝑇 𝑇𝑇
𝑟𝑟 ′ from which 𝛿𝛿𝑑𝑑 𝑇𝑇
𝑟𝑟 − 𝑇𝑇 𝑇𝑇
𝑟𝑟 ′ =0 and therefore 𝑟𝑟 = 𝑇𝑇 𝑇𝑇
𝑟𝑟 ′

The latter equation results from the consideration that the equation before it must be true for any virtual displacement
𝛿𝛿𝛿𝛿 . The element stiffness relation in local cooordinates is

𝑘𝑘 ′ 𝑑𝑑′ = 𝑟𝑟 ′

Substituting 𝑑𝑑 ′ = 𝑇𝑇 𝑑𝑑 , premultiplying both sides by 𝑇𝑇 𝑇𝑇 , and then, on the right-hand side, substituting for 𝑇𝑇 𝑇𝑇
𝑟𝑟 ′ the
previous equation is transformed to

𝑇𝑇
𝑇𝑇 𝑘𝑘 ′ 𝑇𝑇 𝑑𝑑 = 𝑇𝑇 𝑇𝑇
𝑟𝑟 ′ ⟹ 𝑘𝑘 𝑑𝑑 = 𝑟𝑟 where 𝑘𝑘 = 𝑇𝑇 𝑇𝑇
𝑘𝑘 ′ 𝑇𝑇

One-Dimensional Elements 17 / 20
 Rod and BAR Elements of Arbitrary Orientation
For the rod element, where 𝑘𝑘 ′ is the 2 by 2 matrix, the transformation from local to global coordinates gives:

𝑐𝑐 0 𝑐𝑐 2 𝑐𝑐𝑐𝑐 −𝑐𝑐 2 −𝑐𝑐𝑐𝑐 where

𝑇𝑇 ′ 𝑠𝑠 0 𝑘𝑘 −𝑘𝑘 𝑐𝑐 𝑠𝑠 0 0 𝑐𝑐𝑐𝑐 𝑠𝑠 2 −𝑐𝑐𝑐𝑐 −𝑠𝑠 2
𝑘𝑘 = 𝑇𝑇 𝑘𝑘 𝑇𝑇 = = 𝑘𝑘 𝑐𝑐 = 𝑐𝑐𝑐𝑐𝑐𝑐𝛽𝛽
0 𝑐𝑐 −𝑘𝑘 𝑘𝑘 0 0 𝑐𝑐 𝑠𝑠 −𝑐𝑐 2 −𝑐𝑐𝑐𝑐 𝑐𝑐 2 𝑐𝑐𝑐𝑐
0 𝑠𝑠 −𝑐𝑐𝑐𝑐 −𝑠𝑠 2 𝑐𝑐𝑐𝑐 𝑠𝑠 2 𝑠𝑠 = 𝑠𝑠𝑠𝑠𝑠𝑠𝛽𝛽
A rod element arbitrarily oriented in global coordinates 𝑥𝑥𝑥𝑥𝑥𝑥 has a 6 by 6 stiffness matrix that operates on nodal d.o.f.
𝑑𝑑 = 𝑢𝑢1 𝑣𝑣1 𝑤𝑤1 𝑢𝑢2 𝑣𝑣2 𝑤𝑤2 𝑇𝑇

Where 𝑢𝑢, 𝑣𝑣 and 𝑤𝑤 are displacement components in 𝑥𝑥, 𝑦𝑦 and 𝑧𝑧 directions, respectively. In this case the transformation matrix
becames
𝑙𝑙1 𝑚𝑚1 𝑛𝑛1 0 0 0
𝑇𝑇 =
0 0 0 𝑙𝑙1 𝑚𝑚1 𝑛𝑛1
where 𝑙𝑙1 , 𝑚𝑚1 , and 𝑛𝑛1 are direction cosines of local member axis 𝑥𝑥 ′ with respect to global axes 𝑥𝑥, 𝑦𝑦 and 𝑧𝑧.

One-Dimensional Elements 18 / 20
 Rod and BAR Elements of Arbitrary Orientation
For the bar element components of translation and rotation trnsform from global to local directions in the same way; for
example
𝑢𝑢1′ = 𝑙𝑙1 𝑢𝑢1 + 𝑚𝑚1 𝑣𝑣1 + 𝑛𝑛1 𝑤𝑤1 and ′
𝜃𝜃𝑥𝑥𝑥 = 𝑙𝑙1 𝜃𝜃𝑥𝑥𝑥 + 𝑚𝑚1 𝜃𝜃𝑦𝑦1 + 𝑛𝑛1 𝜃𝜃𝑧𝑧1
Transformation of the stiffness matrix from local to global d.o.f. remains as before, and the transformation matrix is

Λ 3𝑥𝑥𝑥 0 3𝑥𝑥𝑥 0 3𝑥𝑥𝑥 0 3𝑥𝑥𝑥

𝑙𝑙1 𝑚𝑚1 𝑛𝑛1
𝑇𝑇 0 Λ 0 0
𝑘𝑘 = 𝑇𝑇 𝑘𝑘 ′ 𝑇𝑇 𝑇𝑇 12𝑥𝑥𝑥𝑥 = 3𝑥𝑥𝑥 3𝑥𝑥𝑥 3𝑥𝑥𝑥 3𝑥𝑥𝑥 where Λ = 𝑙𝑙2 𝑚𝑚2 𝑛𝑛2
0 3𝑥𝑥𝑥 0 3𝑥𝑥𝑥 Λ 3𝑥𝑥𝑥 0 3𝑥𝑥𝑥
𝑙𝑙3 𝑚𝑚3 𝑛𝑛3
0 3𝑥𝑥𝑥 0 3𝑥𝑥𝑥 0 3𝑥𝑥𝑥 Λ 3𝑥𝑥𝑥

Local direction should be suited to the element geometry so that cross-sectional properties of a bar element can be stated
without confusion. For example, one might define the 𝑥𝑥 ′ 𝑦𝑦 ′ plane as being xcoincident with the web of an I bar. The direction
of local axis 𝑥𝑥 ′ is dfined by global coordinates of the two bar nodes, which may be placed at centroids of end cross sections.
If coordinates are supplied for a third point that is not collinear with the end nodes, there is enough information to establish
the orientation of the 𝑥𝑥 ′ 𝑦𝑦 ′ plane in global coordinates 𝑥𝑥𝑥𝑥𝑥𝑥, and hence to establish direction 𝑧𝑧 ′ . Direction cosines of axis 𝑦𝑦 ′
are provided by the cross product of unit vectors in 𝑧𝑧 ′ and 𝑥𝑥 ′ directions.

One-Dimensional Elements 19 / 20