16.

61 Aerospace Dynamics Spring 2003

Lecture #9

Virtual Work

And the

Derivation of Lagrange’s Equations

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 1
16.61 Aerospace Dynamics Spring 2003

Derivation of Lagrangian Equations

Basic Concept: Virtual Work

Consider system of N particles located at ( x1 , x2 , x3 , … x3 N ) with
3 forces per particle ( F1 , F2 , F3 , … F3 N ) , each in the positive
direction.

xk F5

F3 F6
F4

F2 (x4, x5, x6)

F1
(x1, x2, x3)
xj
FN

xi (xN-2, xN-1, xN)

FN-1
FN-2

Assume system given small, arbitrary displacements in all

directions.

Called virtual displacements

- No passage of time
- Applied forces remain constant

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 2
16.61 Aerospace Dynamics Spring 2003

The work done by the forces is termed Virtual Work.

3N
δ W = ∑ F jδ x j
j =1

Note use of δx and not dx.

Note:
• There is no passage of time
• The forces remain constant.

In vector form:
3
δ W = ∑ Fi • δ ri
i =1

Virtual displacements MUST satisfy all constraint relationships,

! Constraint forces do no work.

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 3
16.61 Aerospace Dynamics Spring 2003

Example: Two masses connected by a rod

R1 R2

eˆr

Constraint forces:
R1 = − R 2 = − R2eˆr

Now assume virtual displacements δ r1 , and δ r2 - but the

displacement components along the rigid rod must be equal, so
there is a constraint equation of the form
er • δ r1 = er • δ r2
Virtual Work:

δ W = R1 • δ r1 + R 2 • δ r2
= − R2eˆr • δ r1 + R2eˆr • δ r2
= ( R2 − R2 ) eˆr • δ r1
=0

So the virtual work done of the constraint forces is zero

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 4
16.61 Aerospace Dynamics Spring 2003

This analysis extends to rigid body case

• Rigid body is a collection of masses
• Masses held at a fixed distance.

! Virtual work for the internal constraints of a rigid body

displacement is zero.

Example: Body sliding on rigid surface without friction

Since the surface is rigid and fixed, δ rs = 0, → δ W = 0

For the body, δ W = R1 • δ r1 , but the direction of the virtual

displacement that satisfies the constraints is perpendicular to the
constraint force. Thus δ W = 0 .

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 5
16.61 Aerospace Dynamics Spring 2003

Principle of Virtual Work

mi = mass of particle i
R i = Constraint forces acting on the particle
Fi = External forces acting on the particle

! For static equilibrium (if all particles of the system are

motionless in the inertial frame and if the vector sum of all
forces acting on each particle is zero)
R i + Fi = 0
The virtual work for a system in static equilibrium is
N
δ W = ∑ ( R i + Fi ) • δ ri = 0
i =1

But virtual displacements must be perpendicular to constraint

forces, so
R i • δ ri = 0 ,
which implies that we have
N

∑F •δ r = 0
i =1
i i

Principle of virtual work:

The necessary and sufficient conditions for the static
equilibrium of an initially motionless scleronomic system
which is subject to workless bilateral constraints is that zero
virtual work be done by the applied forces in moving through
an arbitrary virtual displacement satisfying the constraints.

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 6
16.61 Aerospace Dynamics Spring 2003

Example: System shown consists of 2 masses connected by a

massless bar. Determine the coefficient of friction on the floor
necessary for static equilibrium. (Wall is frictionless.)

Virtual Work: δ W = mgδ x1 − µ N 2δ x2

Constraints and force balance: δ x1 = δ x2 , N 2 = 2mg

Substitution: mg (1 − 2 µ ) δ x = 0

Result: µ = 12

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 7
16.61 Aerospace Dynamics Spring 2003

So far we have approached this as a statics problem, but this is a

dynamics course!!

Recall d’Alembert who made dynamics a special case of statics:

N
δ W = ∑ ( R i + Fi − mi""ri ) • δ ri = 0
i =1
N
⇒ ∑ ( Fi − mi""
ri ) • δ ri = 0
i =1

! So we can apply all of the previous results to the dynamics

problem as well.

Comments:

" Virtual work and virtual displacements play an important

role in analytical dynamics, but fade from the picture in the
application of the methods.

" However, this is why we can ignore the calculation of the

constraint forces.

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 8
16.61 Aerospace Dynamics Spring 2003

Generalized Forces
Since we have defined generalized coordinates, we need
generalized forces to work in the same “space.”

Consider the 2 particle problem:

xj
(x3, x4)

l θ

(x1, x2)
xi

Coordinates: x1 , x2 , x3 , x4
( x1 − x3 ) + ( x2 − x4 ) = l 2
2 2
Constraint:
DOF: 4 −1 = 3

• Select n=3 generalized coordinates:

( x1 + x3 ) ( x + x4 ) ( x − x2 )
q1 = q2 = 2 q3 = tan −1 4
2 2 ( x3 − x1 )
• Can also write the inverse mapping:

xi = fi ( q1 , q2 , q3 , … qn , t )

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 9
16.61 Aerospace Dynamics Spring 2003

3N

Virtual Work: δ W = ∑ F jδ x j
j =1

 ∂x j
n 
δx j = ∑  δqi
i =1  ∂qi
Constraint relations:


 ∂x j
3N n 
δW = ∑∑ F j  δqi
Substitution: j =1 i =1  ∂qi 

 ∂x j 
3N

Define Generalized Force: Qi = ∑ Fj  

j =1 ∂q
 i
! Work done for unit displacement of qi by forces acting on the
system when all other generalized coordinates remain constant.

n
⇒ δW = ∑ Qi δqi
i =1
• If qi is an angle, Qi is a torque
• If qi is a length, Qi is a force
• If the qi ’s are independent, then for static equilibrium must
have:
Qi = 0, i = 1, 2, … n

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 10
16.61 Aerospace Dynamics Spring 2003

Derivation of Lagrange’s Equation

• Two approaches

(A) Start with energy expressions

Formulation

Lagrange’s Equations (Greenwood, 6-6)

Interpretation

Newton’s Laws

(B) Start with Newton’s Laws

Formulation

Lagrange’s Equations (Wells, Chapters 3&4)

Interpretation

Energy Expressions

(A) Replicated the application of Lagrange’s equations in

solving problems

(B) Provides more insight and feel for the physics

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 11
16.61 Aerospace Dynamics Spring 2003

Our process
1. Start with Newton
2. Apply virtual work
3. Introduce generalized coordinates
4. Eliminate constraints
5. Using definition of derivatives, eliminate explicit use of
acceleration

• Start with a single particle with a single constraint, e.g.

o Marble rolling on a frictionless sphere,
o Conical pendulum

z
m
r=L

φ F
y
Free body
θ Diagram
x

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 12
16.61 Aerospace Dynamics Spring 2003

1. Newton: F = ma

• For the particle: Fx = mx"", Fy = my"", Fz = mz"" where axes x, y, z

describe an inertial frame

• Note that Fx , Fy , Fz are the vector sum of all forces acting on

the particle (applied and constraint forces)

2. Apply Virtual Work:

• Consider δs, which is an arbitrary displacement for the
system, then the virtual work associated with this
displacement is:

δ W = Fxδ x + Fyδ y + Fzδ z

• Note that δs may violate the applied constraints, because F

contains constraint forces

• Combine Newton and Virtual Work

Fxδ x = mx""δ x
Fyδ y = my""δ y
Fzδ z = mz""δ z
• Add the equations

m ( ""
xδ x + "" zδ z ) = Fxδ x + Fyδ y + Fzδ z
yδ y + ""

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 13
16.61 Aerospace Dynamics Spring 2003

• Called D’Alembert’s equation:

m ( ""
xδ x + "" zδ z ) = Fxδ x + Fyδ y + Fzδ z
yδ y + ""

• Observations:
o Scalar relationship
o LHS ≈ kinetic energy
o RHS ≈ virtual work term

3. Introduce generalized coordinates

o Assumed motion on a sphere ! 1 stationary constraint
o DOF = 3 - 1 = 2 generalized coordinates

x = f1 ( q1 , q2 ) , y = f 2 ( q1 , q2 ) , z = f 3 ( q1 , q2 )

• Define virtual displacements in terms of generalized

coordinates:
∂x ∂x
δx= δ q1 + δ q2
∂q1 ∂q2
n  ∂x j 
δx j = ∑  δqi ∂y ∂y
δy= δ q1 + δ q2
i =1  ∂qi
!
 ∂q1 ∂q2
∂z ∂z
δz = δ q1 + δ q2
∂q1 ∂q1

• Note: these virtual displacements conform to the constraints,

because the mapping of the generalized coordinates conforms
to the surface of the sphere.

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 14
16.61 Aerospace Dynamics Spring 2003

• Substitute virtual displacements into D’Alembert’s equation

∂x ∂x
δx= δ q1 + δ q2
∂q1 ∂q2
∂y ∂y
δy= δ q1 + δ q2
∂q1 ∂q2
∂z ∂z
δz = δ q1 + δ q2
∂q1 ∂q1

m ( ""
xδ x + "" zδ z ) = Fxδ x + Fyδ y + Fzδ z
yδ y + ""

 ∂x ∂y ∂z   ∂x ∂y ∂z 
m  ""
x + ""
y + ""
z δ
 1 q + m  ""
x + ""
y + ""
z  δ q2
 ∂q1 ∂q1 ∂q1   ∂q2 ∂q2 ∂q2 
 ∂x ∂y ∂z   ∂x ∂y ∂z 
=  Fx + Fy + Fz δ q +
 1  x F + F y + F z  δ q2
 ∂q1 ∂ q1 ∂ q 1   ∂q2 ∂q 2 ∂ q 2 

• Facts:
o Virtual displacements δ q1 and δ q2 conform to constraints
o Virtual work δ W is work that conforms to constraints
o δ q1 and δ q2 are independent and can be independently
moved without violating constraints

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 15
16.61 Aerospace Dynamics Spring 2003

• Conclusion:
o Force of the constraint has been eliminated by selecting
generalized coordinates that enforce the constraint
(Reason 2 for Lagrange, pg 24)
o Further, we can split the equation into two equations in two
unknowns due to independence of δ q1 and δ q2 .

 ∂x ∂y ∂z   ∂x ∂y ∂z 
m  ""
x + ""
y + ""
z =
  x F + Fy + Fz 
 ∂q1 ∂q1 ∂q1   ∂q1 ∂q1 ∂q1 

 ∂x ∂y ∂z   ∂x ∂y ∂z 
m  ""
x + ""
y + ""
z =
  x F + Fy + Fz 
 ∂q2 ∂q2 ∂q2   ∂q 2 ∂q 2 ∂q2 

5. Finally, eliminate acceleration terms

• Consider the total derivative

d  ∂x  ∂x d  ∂x 
 "
x  = ""
x + "
x  
dt  ∂q1  ∂q1 dt  ∂q1 

• Rearrange

∂x d  ∂x  d  ∂x 
""
x =  x"  − "
x   (1)
∂q1 dt  ∂q1  dt  ∂q1 

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 16
16.61 Aerospace Dynamics Spring 2003

• Recall
d
x = f1 ( q1 , q2 ) ∴ x" =  f1 ( q1 , q2 ) 
dt 

• Perform the derivative (chain rule):

∂x ∂x
x" = q"1 + q"2 (2)
∂q1 ∂q2

• Partial derivative of (2) with respect to q"1 gives

∂x" ∂x
= (3)
∂q"1 ∂q1

∂x
• Since x = f1 ( q1 , q2 ) , = g1 ( q1 , q2 ) is a ftn of both q1 and q2
∂q1
∂x
the time derivative of gives (chain rule again)
∂q1

d  ∂x  ∂  ∂x  ∂  ∂x 
  =  "
q
 1 +   q"2 (4)
dt  ∂q1  ∂q1  ∂q1  ∂q2  ∂q1 

• Partial derivative of x" (2) with respect to q1 gives

∂x" ∂  ∂x  ∂  ∂x 
=  "
q
 1 +   q"2 (5)
∂q1 ∂q1  ∂q1  ∂q1  ∂q2 

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 17
16.61 Aerospace Dynamics Spring 2003

• Note RHS of 4 and 5 are the same, thus

∂x" d  ∂x 
! =   (6)
∂q1 dt  ∂q1 

• Now, insert (3) and (6) into (1):

∂x d  ∂x  d  ∂x 
""
x =  x"  − "
x   (1)
∂q1 dt  ∂q1  dt  ∂q1 

∂x ∂x" d  ∂x  ∂x"
=  = (3 and 6)
∂q1 ∂q"1 dt  ∂q1  ∂q1

• Results in:
∂x d  ∂x"  ∂x"
""
x =  x"  − "
x (7)
∂q1 dt  ∂q"1  ∂q1

• Note that

 x" 2   x" 2 
∂  ∂ 
∂x" 2 ∂x" 2
x" =   and x" =  
∂q"1 ∂q"1 ∂q1 ∂q1

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 18
16.61 Aerospace Dynamics Spring 2003

• Finally

  x" 2    x" 2 
 ∂   ∂ 
∂x d   2   2
""
x = −   (8)
∂q1 dt  ∂q"1  ∂q1
 
 

• The above process is identical for y and z.

• Recall our virtual work equation for q1 :

 ∂x ∂y ∂z   ∂x ∂y ∂z 
m  ""
x + ""
y + ""
z =
  x F + Fy + Fz 
 ∂q1 ∂q1 ∂q1   ∂q1 ∂q1 ∂q1 

• Insert equations (8) for x, y and z and collect terms to

eliminate acceleration terms. (Reason 3 for Lagrange, pg 24)

d  ∂  x" 2 + y" 2 + z" 2   ∂  x" 2 + y" 2 + z" 2 

  m  −  m 
dt  ∂q"1  2   ∂q1  2 
 ∂x ∂y ∂z 
=  Fx + Fy + Fz 
 ∂q1 ∂q1 ∂q1 

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 19
16.61 Aerospace Dynamics Spring 2003

• Observe that:

1 2 1
T = mv = m ( x" 2 + y" 2 + z" 2 )
2 2

which is the kinetic energy of the particle

• Finally:

d ∂  ∂  ∂x ∂y ∂z 
 " T − T =  Fx + Fy + Fz 
dt  ∂q1  ∂q1  ∂q1 ∂q1 ∂q1 

• Similarly:

d ∂  ∂  ∂x ∂y ∂z 
 " T − T =  Fx + Fy + Fz 
dt  ∂q2  ∂q2  ∂q2 ∂q2 ∂q2 

• The general form of Lagrange’s equation is thus:

d ∂  ∂
 " T − T = Qqr
dt  ∂qr  ∂qr
 ∂x ∂y ∂z 
Qqr =  Fx + Fy + Fz 
 ∂qr ∂qr ∂qr 

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 20
16.61 Aerospace Dynamics Spring 2003

• Some observations:
o One Lagrange equation needed for each DOF
o Easily extendable for a system of particles
o T – Expression of system kinetic energy
o All inertial forces contained in the LHS
o Qq only contains external forces
r

• How to use this ….

1. Determine number of DOF and constraints
2. Identify generalized coordinates and equations of constraint
a. Iterate on 1 and 2 if needed
3. Write expression for T
a. v inertial velocity that can be written in terms of the
coordinates of any frame
b. Find required derivatives of T
4. Find generalized forces Qqr

a. If forces are known in inertial coordinates, transform

them to generalized coordinates
b. Apply generalized force equation for each force

 ∂x ∂y ∂z 
Qqr =  Fx + Fy + Fz 
 ∂qr ∂qr ∂qr 

5. Substitute into Lagrange’s equation

6. Solve analytically or numerically

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 21
16.61 Aerospace Dynamics Spring 2003

Example: Projectile Problem:

1, 2, 3. DOF = 3, no constraints

1 2 1
4. T= mv = m ( x" 2 + y" 2 + z" 2 )
2 2

∂T ∂T ∂T
5. = mx" , = my" , = mz"
∂x" ∂y" ∂z"

d  ∂T  d  ∂T  d  ∂T 
  = ""
mx ,   = ""
my ,   = mz""
dt  ∂x"  dt  ∂y"  dt  ∂z" 

∂T ∂T ∂T
= = =0
∂x ∂y ∂z

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 22
16.61 Aerospace Dynamics Spring 2003

6. Generalized forces:
 ∂x ∂y ∂z 
Qqr =  Fx + Fy + Fz 
 ∂qr ∂qr ∂qr 
F = − mg zˆ
∂z
Qqx = Qq y = 0, Qqz = Fz = −mg
∂z

7. EOMs: mx"" = 0, = my"" = 0, mz"" = −mg

8. Solve differential equations.

• Comments:
o Method is overkill for this problem
o Inspection shows agreement with Newton

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 23
16.61 Aerospace Dynamics Spring 2003

Example: Mass moving along a frictionless track.

Z
θ ρ

φ
X

• Track geometry defined such that:

ρ = az , and φ = −bz
DOF = 3 – 2 =1
• Constraint equations: ρ = az , and φ = −bz
• Generalized coordinate: z

1 2
• Find T = mv , what is v?
2

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 24
16.61 Aerospace Dynamics Spring 2003

• Define rotating coordinate frame such that mass remains in

xˆ − zˆ plane.

ẑ ρ
Z
ρ z
x̂

z
ŷ

Y
x̂
φ
X

r = ρx# +zz# = azx# + zz# and ω = φ" z# = −bzz#

v 2 = r" • r"
r" = azx
" # + zz
"# + ( −bzz# ) × azx# + zz# = (az")2 + (abzz
" )2 + z" 2
= azx
" # − abzzy
" # + zz
"# !
= (1 + a 2 + a 2 b 2 z 2 )z" 2

m ∂T
2 2 2 2 2
T = (1 + a + a b z )z" and = m(1 + a 2 + a 2b2 z 2 ) z"
2 ∂z"

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 25
16.61 Aerospace Dynamics Spring 2003

d  ∂T  2 2 2 2
z + 2m a 2b2 z z" 2
 "  = m(1 + a + a b z )""
dt  ∂z 
( )
∂T
= m(a 2b2 z ) z" 2
∂z

• External force is gravity

 ∂x ∂y ∂z 
Qqr =  Fx + Fy + Fz 
 ∂qr ∂qr ∂qr 

F = −mg zˆ

∂z
Qqx = Qq y = 0, Qqz = Fz = −mg
∂z

• Equation of Motion:

( a 2
+ a 2 2 2
b z + 1) ""
z + a 2 2
b z "
z 2
= −g

• Comments:
o Solution highly nonlinear
o “Trick” was finding inertial velocity
o Still need to use FARM approach

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 26
16.61 Aerospace Dynamics Spring 2003

Extending Lagrange’s Equation to Systems

with Multiple Particles

• Assume a system of particles and apply Newton’s laws:

Fx1 = mx""1 , Fy1 = my""1 , Fz1 = mz""1
$ $ $
Fx p = mx""p , Fy p = my""p , Fz p = mz""p

• As before, the F’s contain both external and constraint forces.

• Multiply both sides of each equation by the appropriate

virtual displacement and add all the equations together.
p p

∑ m ( ""x δ x + ""y δ y
i i i i + "" (
ziδ zi ) = ∑ Fxi δ xi + Fyi δ yi + Fzi δ zi )
i =1 i =1

• Recall that this is D’Alembert’s equation

• Assume the system has n DOF, n ≤ 3 p

• Select generalized coordinates, qi that enforce the constraints:

xi = f i ( q1 , q2 , …, qn , t )
yi = gi ( q1 , q2 , …, qn , t )
zi = hi ( q1 , q2 , …, qn , t )

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 27
16.61 Aerospace Dynamics Spring 2003

• Express virtual displacements in terms of generalized

coordinates:
∂xi ∂x ∂x
δ xi = δ q1 + i δ q2 + … + i δ qn
∂q1 ∂q2 ∂qn
∂yi ∂y ∂y
δ yi = δ q1 + i δ q2 + … + i δ qn
∂q1 ∂q2 ∂qn
∂zi ∂z ∂z
δ zi =
δ q1 + i δ q2 + … + i δ qn
∂q1 ∂q2 ∂qn
• Substitute the relations into D’Alembert’s equation
p
 ∂xi ∂y ∂z 
∑ m  ""
xi + ""
yi i + ""
zi i δ qr
i =1  ∂qr ∂qr ∂qr 
 ∂x p
∂y ∂z 
= ∑  Fxi i + Fyi i + Fzi i  δ qr
i =1  ∂qr ∂qr ∂qr 
• As before, have used fact that the generalized coordinates
automatically enforce the constraints.
o Sum over the entire system of particles decouples for each
of the generalized coordinates.
o This leaves us n such equations.

• Using the calculus relations (chain rule), one can show that
d ∂  ∂
 " T − T = Qqr
dt  ∂qr  ∂qr
• Once again, one Lagrange equation for each DOF.

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 28
16.61 Aerospace Dynamics Spring 2003

Lagrange’s Equation for Conservative Systems

• Conservative forces and conservative systems

o Forces are such that the work done by the forces in moving
the system from one state to another depends only on the
initial and final coordinates of the particles (path
independence).

• Potential Energy, V
o Work done by a conservative force in a transfer from a
general configuration A to a reference configuration B is
the potential energy of the system at A with respect to B.
o Note: V is defined as work from the general state to the
reference state.

• Examples of conservative forces:

o Springs (linear elastic)
o Elastic bodies
o Gravity force

• Non-conservative forces
o Friction
o Drag of a fluid
o Any force with time or velocity dependence

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 29
16.61 Aerospace Dynamics Spring 2003

• General Expression for V, the potential energy

A P

(
V = − ∫ ∑ Fxi dxi + Fyi dyi + Fzi dzi )
B i =1
• Note the “–“ sign since the path is from B to A. The sum is
over the P particles in the system.

• For path independence, integrand must be an exact

differential. Thus:
∂V ∂V ∂V
Fxi = − Fyi = − Fzi = − (C1)
∂xi ∂yi ∂zi
• Observe that:

∂Fx3 ∂  ∂V  ∂ 2V
= − =−
∂y4 ∂y4  ∂x3  ∂x3∂y4
∂Fy4 ∂  ∂V  ∂ 2V
= − =−
∂x3 ∂x3  ∂y4  ∂x3∂y4

• Thus, in general
∂Fxi ∂Fyr
= (C2)
∂yr ∂xi

• Equation (C1) represents a necessary condition for a force to

be conservative, Equation (C2) is a sufficient condition.

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 30
16.61 Aerospace Dynamics Spring 2003

• Recall expression for generalized forces:

 ∂xi
p
∂yi ∂zi 
Qqr = ∑  Fxi + Fyi + Fzi 
i =1  ∂qr ∂qr ∂qr 
o Separate forces into conservative and non-conservative

 ∂V ∂xi ∂V ∂yi ∂V ∂zi 

p
Qqr = −∑  + +  + Q N
qr
i =1  ∂xi ∂qr ∂yi ∂qr ∂yi ∂qr 
∂V
=− + Q N qr
∂qr
• Lagrange’s Equation:

d ∂  ∂
 " T − T = Qqr
dt  ∂qr  ∂qr
• Substitute in generalized force:

d ∂  ∂ ∂V
 T  − T = − + Q N
qr
dt  ∂q"r  ∂qr ∂qr
d ∂  ∂
⇒  T  − ( T − V ) = Q N
qr
dt  ∂q"r  ∂qr

• Since conservative forces are not functions of

∂
velocities: V =0
∂q"r

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 31
16.61 Aerospace Dynamics Spring 2003

• Thus, can define the Lagrangian L = T − V to obtain the

final form of Lagrange’s equation:
d  ∂L  ∂L
 − = Fqr
dt  ∂q"r  ∂q r

Example: Planar pendulum with an inline spring.

k
r
θ
m
x

• DOF = 3 – 1 = 2
• Constraint equation: z = 0
• Generalized coordinates: r, θ
• Coordinate mapping: x = r cosθ , y = r sin θ

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 32
16.61 Aerospace Dynamics Spring 2003

• Kinetic energy
1 2 1
mv = m ( x" 2 + y" 2 )
T=
2 2
• Derivatives of coordinates:
x" = r" cosθ − rθ" sin θ , y" = r" sin θ + rθ" cosθ
• Substitute into kinetic energy equation
1
T=
2
(
m r" 2 + r 2θ" 2 )
• Potential energy
1
k ( r − ro ) − mgr cosθ
2
V=
2
• Lagrangian
1 1
L = T −V =
2 2
( 2
)
m r" 2 + r 2θ" 2 − k ( r − ro ) + mgr cosθ
• Derivatives of L (note need to do this for each GC)
∂L d  ∂L  ∂L
= mr",  "  = mr"", = mrθ" 2 − k ( r − ro ) + mg cosθ
∂r" dt  ∂r  ∂r
∂L d  ∂L  ∂L
= mr 2θ",  " = mr 2 ""
θ + 2mrr"θ", = −mgr sin θ
∂θ" dt  ∂θ  ∂θ

• Substitute into Lagrange’s Equation:

mr"" − mrθ" 2 + k ( r − ro ) = mg cosθ
mr 2θ"" + 2mrr"θ" − mgr sin θ = 0

Massachusetts Institute of Technology © How, Deyst 2003 (Based on Notes by Blair 2002) 33