Virtual Work and Variational Principles
Virtual Work and Variational Principles
www.inrisk.ubc.ca
Weak Form
Variational Form
This form is only possible for conservative systems, i.e., those in which the work
done by the forces is reversible; friction and follower-loads are examples of nonconservative forces, which render the variational form unattainable
This form is sometimes called the energy form because it is based on formulations of
the energy of the system
Page 1
www.inrisk.ubc.ca
In statics, this form expresses the principle of minimum potential energy; in dynamics
it is called Hamiltons principle
This form also requires only average satisfaction of the governing equations, but
depending on the variational principle it is the equilibrium or kinematic compatibility
requirements that are approximated, or both
To illustrate the four forms of the mechanical BVP the problem of beam bending
according to the Euler-Bernoully theory is addressed. The distributed load is denoted qz,
which is positive in the direction of the z-axis. This is different from the notation q in
beam theory, which for pedagogical convenience acts downwards although the z-axis is
positive upward. Consequently, the qz in this document is the negative of the q that
appears in the basic theory:
Strong form:
EI w '''' qz = 0
(1)
( EI w '''' q ) w dx = 0
(2)
Weak form:
EI w '' w '' dx q
w dx = 0
(3)
2
Variational form: (w) = EI ( w '') qz w dx with = 0
2
(4)
Starting at the strong form, the weighted residual form is obtained by multiplying the
differential equation by a weight function, w, which is subsequently interpreted as a
virtual displacement in the weak form, and integrating over the problem domain. Next,
integration by parts yields the weak form. The boundary terms that appear in the
integration by part are
(5)
All of these boundary terms always cancel; each term cancels because of one of two
possible reasons. Either there are kinematic restrictions on the field function, w, so that
w and w vanish. In other words, wherever w or w is prescribed, e.g., as zero, there
cannot be a corresponding virtual deformation. Those are called essential boundary
conditions. The alternative reason that a boundary term in Eq. (5) cancels is that the
Page 2
www.inrisk.ubc.ca
remainder of the expression is zero, i.e., that EIw or EIw is zero. These are called
natural boundary conditions because they are part of the formulation. Notice that in this
case EIw and EIw represent shear force and bending moment, respectively. I.e., here
the natural boundary conditions impose boundary conditions on the forces in the
structure. Next, the variational form is established by equating the potential energy of the
external load to the internal elastic energy. If the variational form is possible then all the
other forms can be derived from it. In fact, the variation of the variational form yields the
weak form, from which the weighted residual form is obtained by integration by parts. In
turn, the differential equation is obtained by letting the weight function be arbitrary.
When the differential equation is derived from a variational principle in this way then it is
called the Euler equation.
Concepts
Vectorial and Analytical Mechanics
Starting in the late 1600s mechanics developed in two directions, called vectorial and
analytical mechanics. Both approaches are still widely used. Vectorial mechanics is
prevalent in fundamental education in mechanics and in engineering practice. Analytical
mechanics is more popular in advanced application of mechanics, and forms the basis for
this document. A summary of the two approaches is provided in Table 1.
Table 1: Overview of vectorial and analytical mechanics.
Vectorial mechanics
Analytical mechanics
Foundation
Newtons laws
Central quantities
Vectors:
Momentum, force
Scalars:
Kinetic energy and the work function
(potential energy)
Coordinate system
Formulation
Equations
Principles
Page 3
Principle
Significance
BVP form
www.inrisk.ubc.ca
Direct Methods
Weak form
Rayleigh-Ritz method
Finite element method
dAlemberts principle
Weak form
Principle of minimum
potential energy
Implies equilibrium;
deformations are unknown
Variational
form
Rayleigh-Ritz method
Finite element method
Hamiltons principle
Variational
form
Variational
form
Principle of complementary
virtual work
Principle of minimum
complementary energy
Implies kinematic
compatibility; forces are
unknown
Variational
form
Hu-Washizus principle
Variational
form
Hellinger-Reissners
principle
Variational
form
N/A
Weighted
residual form
Galerkin method
Least squares method
Finite element method
N/A
Strong form
Analytical solutions
Finite differences
While ordinary calculus deals with functions of one or more variables the calculus of
variations deals with functions of functions. Often, the quantity at hand is an integral of a
function, in which the function that minimizes the integral is sought. Leonard Euler
employed this type of formulation, specifically the potential energy for beam bending
established by his contemporary Daniel Bernoulli, to solve beam-bending problems. Such
functions of functions are called functionals. The calculus of variations, which is outlined
later in this document, establishes rules that have similarities to ordinary calculus.
Page 4
www.inrisk.ubc.ca
w(x) = N q (x)q
(6)
where Nq contains the shape functions that correspond to each generalized coordinate q.
Caution must be exercised to avoid confusion with the notation q, qx, qy, and qz, for
distributed loads on beams. For now it is noted that the generalized shape functions must
represent a complete set of orthogonal functions.
Principles
Principle of Virtual Displacements
In fundamental structural analysis the principle of virtual forces is utilized to derive the
unit virtual load method, as described in another document. However, the principle of
virtual displacements plays an even greater role in advanced structural mechanics. It
states that a deformable body is in equilibrium if the external virtual work equals the
internal virtual work for any virtual deformation that satisfies the kinematic boundary
conditions:
Wint = Wext
(7)
where the expressions for dWint and dWext contain virtual deformations, such as strain,
displacement, and curvature. Expressions for virtual work are provided in another
Page 5
www.inrisk.ubc.ca
document. Those expressions do NOT have a -factor because the virtual displacements
are applied first, and carry out work when the actual displacements are imposed.
dAlemberts Principle
This principle is the principle of virtual work extended with an inertia term. dAlembert
states that according to Newtons second law there will be an inertia force in the opposite
direction compared to the direction of the acceleration. Figure 1 illustrates this for a
generic single-degree-of-freedom system on the left and an infinitesimal beam element
on the right. Consequently, the extension of Eq. (7) for beam problems according to
dAlemberts principle reads:
Wint = Wext
(8)
dV = ( p m u ) u d
z,w
qz
x,u
C
M ! u!!
!!
m!w
dx
Figure 1: dAlemberts principle.
Principle of Minimum Potential Energy
This principle can be derived from the principle of virtual displacements. Thus, it
represents integrated equilibrium. It can also be derived from the laws of
thermodynamics. The principle states that of all displacements that satisfy kinematic
boundary conditions the ones that minimize the potential energy provide stable static
equilibrium. The principle is applied by expressing the potential energy associated, H,
associated with external loads, and the elastic strain energy, U, and adding them to obtain
the total potential energy:
= H +U
(9)
Equilibrium requires a stationary value of the functional, i.e., that the first variation of the
functional is zero:
= 0
(10)
and stable equilibrium requires that the second-variation is greater than zero for all
variables
Virtual Work and Variational Principles
Page 6
2 > 0
www.inrisk.ubc.ca
(11)
i.e., the equilibrium is stable when the potential energy increase in the vicinity of the
solution.
Castiglianos Theorems
Castiglianos first theorem states that if the strain energy, U, can be expressed in terms of
a set of generalized displacements, q, that are associated with the generalized forces, Q,
then any force can be computed by partial integration of the strain energy with respect to
the corresponding displacement:
Qi =
U(q)
dqi
(12)
qi =
U(Q)
dQi
(13)
Hamiltons Principle
This principle is the same as the principle of minimum potential energy, but extended
with an inertia term. In other words, Hamiltons principle is in dynamics what the
principle of minimum potential energy is in statics. Hamiltons principle can be derived
from dAlemberts principle. It can also be established as an expression for the
conservation of mechanical energy, which states that the work equals the change in
energy. To this end, consider the following derivation of the work-energy theorem. The
incremental work carried out by a force F as it moves a particle by an infinitesimal
displacement du is:
dW = F du
(14)
Next, introduce the rate of change of the displacement over an infinitesimal time period
dt:
dW = F
du
dt F u dt
dt
(15)
Furthermore, apply Newtons second law to express the force, F, in terms of the
particles acceleration:
F = m u
(16)
dW = F u dt = m u u dt =
d 1
2
m ( u ) dt
dt 2
(17)
Eq. (17) expresses that the work over an infinitesimal displacement equals the rate of
change of the kinetic energy. By integration over time it is understood that the work
equals the integral of all changes in the kinetic energy, i.e., the total change in the kinetic
energy between two time instants. More generally, the conservation of mechanical energy
holds:
Virtual Work and Variational Principles
Page 7
K + = W
www.inrisk.ubc.ca
(18)
where denotes change between two time instants. Hamiltons principle formulates this
for conservative systems, i.e., in which all work is recoverable and included in K and
by the action functional:
t2
A=
( K ) dt
(19)
t1
where the integrand K is called the Lagrange function. It is required that for dynamic
equilibrium the variation must be zero:
A = 0
(20)
Essentially, Hamiltons principle requires that, for conservative systems, the sum of
kinetic and potential energy remains constant. For non-conservative systems Hamiltons
principle is extended to include the work:
A =
t2
( K + C + W ) dt
(21)
t1
where C represent Lagrange multipliers on the present constraints. The potential energy
does not appear because the forces cannot be derived from it for non-conservative
problems. In circumstances where some forces are conservative and others are not then
the conservative ones can be derived from the potential energy and the non-conservative
ones from the virtual work, W, that they are associated with.
Principle of Virtual Forces: Complementary Virtual Work
The principle of virtual forces, sometimes called the principle of complementary virtual
work, states that the external and internal work carried out by virtual forces that are
placed on the structure before the actual load is applied are equal:
Wint = Wext
(22)
where the expressions for dWint and dWext contain virtual forces, such as loads, stresses,
and moments. Expressions for virtual work are provided in another document. Those
expressions do NOT have a -factor because the virtual load(s) are applied first, and
carry out work when the actual displacements are imposed.
Principle of Minimum Complementary Energy
In this principle the energy is expressed in terms of forces instead of displacements. A bar
denotes the complementary quantities and the principle reads
= H +U
(23)
= H + U = 0
(24)
Page 8
www.inrisk.ubc.ca
Hu-Washizus Principle
Further extend the principle of minimum potential energy by including both kinematic
compatibility and kinematic boundary conditions by application of Lagrange multipliers.
Hellinger-Reissners Principle
The key notion in direct methods is that the unknown field function is approximated by a
set of complete and linearly independent solution functions as conceptualized in Eq. (6).
Galerkin Method
The Galerkin method employs the weighted residual form of the BVP. Because this form
can be established for all BVPs it is the most versatile approach. However, this method is
also the strictest in terms of selection of solution functions because they must satisfy both
natural and essential boundary conditions.
Rayleigh-Ritz Method
The Rayleigh-Ritz method employs the variational form or the virtual work form of the
BVP. Because those forms are not always available it is a more restricted method.
However, the selection of potential solution functions is more flexible. While the weight
function in the Galerkin approach must satisfy all the same boundary conditions as the
actual displacement function, the Rayleigh-Ritz functions need only satisfy the essential
boundary conditions. For the principle of virtual displacements and the principle of
minimum potential energy these are the kinematic boundary conditions, i.e., those on the
displacements and rotations. For the same shape functions the Galerkin method and the
Rayleigh-Ritz method produce the same results.
Page 9