Sunteți pe pagina 1din 27

Chapter 4

Analytical Mechanics (Basic


Concepts)

Analytical mechanics is also called Lagrangian mechanics. The method is based on Calculus of Variation.

So far, we have done Newtonian mechanics where F = ma, where we spent most of the time obtaining a.
In analytical mechanics, the same equations of motion can be found using Hamilton’s Principle, D’Alembert’s
principle, and Euler-Lagrange’s equations. In this chapter, we will cover the following topics.

1. degree of freedom

2. generalized coordinates

3. constraints (a) holonomic and (b) nonholonomic

4. virtual displacement

5. virtual work

6. generalized force

7. Principle of Virtual Work

8. Hamilton’s Principle

9. D’Alembert’s Principle

10. Euler-Lagrange equations

4.1 Vocabulary
Degree of freedom: Number of degrees in a system is the minimum number of coordinates needed to describe
a system. For instance, a point mass has 3, a rigid bar has 6, two point masses have 6, and a 2D pendulum
has 1. The minimum number of eom required to solve for the motion of the system (position, velocity,...) is
same as the number of degrees of freedom. Figure ?? shows some examples.
Generalize coordinates: is a set of coordinates (do not need to be independent) that can completely
describe the position of a dynamic system. They are usually denoted as qi , and they may and may not be
physical. The space spanned by the general coordinates is the configuration space. It should be noted that
a set of generalized coordinates are not a unique.

Example 4.1 Generalized Coordinates.

1
2 CHAPTER 4. ANALYTICAL MECHANICS (BASIC CONCEPTS)

Figure 4.1: Examples of various systems and their degrees of freedom.


4.1. VOCABULARY 3

Figure 4.2: Generalized coordinates for various systems.

Consider systems in Figure 4.2. (a) For a particle in two dimensions, it has two degrees of freedom. Its
location can be expressed in terms of Cartesian coordinates, (x, y) , or in terms of polar coordinates, (r, θ)
so that

q1 = x, q2 = y or
q1 = r, q2 = θ.

Another candidate is q1 = x and q2 = r. However, this choice of coordinates renders more than one possible
location for the particle. Therefore, these are called ambiguous coordinates and should be avoided.
(b) For a two mass system (also 2DOF), the two displacements of the masses can be the generalized
coordinates. The displacements are measured from unstretched locations of the mass. The other candidate
is the sum and the relative displacements,

q1 = x1 , q2 = x2
q1 = x1 + x2 , q2 = x2 − x1

Let us consider the case where the distance between the origin and the point mass in Figure 4.2(a) is
fixed at L. The position of the point mass is constrained by a constraint equation given by x2 + y 2 = L2 . In
addition, the system is now an 1 DOF system. It is possible to describe the system location by 1 coordinate.
In this case, the most reasonable coordinate is θ.

As we have implied, constraints reduces the number of coordinate used. That is

DOF = N o. of coordinates - No. of constraints.

Example 4.2 2D Pendulum


This is a 1 DOF system. We can let q1 = θ or q1 = x and q2 = y with a constraint, x2 + y2 = L2 .

Constraints: There are two types of constraints. One is holonomic constraint, and the other is
nonholonomic constraint. When the constraint equation can be written in terms of generalized coordinates
and time only (no derivatives), it is called holonomic constraint. It is also called configuration constraint
because it contraints the geometric configuration. Holonomic constraint is usually denoted as f (qi , t) = 0. For
example, the constraint equation for the pendulum is holonomic because we can write f = x2 + y 2 − L2 =
0. When the constraint equation cannot be written in terms of displacements and time only, it is called
nonholonomic.
Pfaffian form: A constraint equation can be written in Pfaffian form so that
N

aj dqj + at dt = 0.
j=1
4 CHAPTER 4. ANALYTICAL MECHANICS (BASIC CONCEPTS)

Figure 4.3: 2D pendulum.

If there are M constraints, we can write


N

aij dqj + ait dt = 0, for i = 1, .., M.
j=1

The term, aij , is called the Jacobian constraint matrix. Any holonomic constraint can be written in
Pfaffian form by taking a derivative. For example, taking a derivative of f = x2 + y 2 − L2 = 0, we obtain
df = 2xdx + 2ydy = 0.
When a holonomic constraint is written in Pfaffian form, it is a perfect differential. That is, we can always
integrate to find f (q, t) .
Nonholonomic constraints are written in Pfaffian form only, and it is not a perfect differential. When a
constraint equation is given by
Adx + Bdy + Cdt = 0,
it is a perfect differential if one can find an integrating factor g (x, y) so that (TEST for PERFECT DIF-
FERENTIAL)
∂gA ∂gB
= . (4.1)
∂y ∂x
This makes sense because if it were a perfect differential,
∂f ∂f ∂f
= gA, = gB, = gC
∂x ∂y ∂t
∂2f ∂2f ∂2f ∂2f ∂2f ∂2f
= , = , =
∂x∂y ∂y∂x ∂x∂t ∂t∂x ∂t∂y ∂y∂t
There is no general method finding g, and sometimes it is hard to tell whether the equation is perfect
differential or not.

Example 4.3 Perfect Differential


Consider a constraint equation in Pfaffian form,
 2   
3x y + y 2 dx + 2x3 + 3xy dy = 0.
Is this holonomic or nonholonomic?
∂gA ∂gB
Solution: If we can find g that makes ∂y = ∂x , the constraint is holonomic. Let us assume that g is a
function of x only. Then,
 
∂g ∂A ∂B
B = − g
∂x ∂y ∂x
 3  ∂g  
2x + 3xy = −3x2 − 2y g
∂x  
∂g −3x2 − 2y
= g.
∂x (2x3 + 3xy)
4.1. VOCABULARY 5

There is no way that g can be a function of x only since the equation contains y. Let us now assume that g
is a function of y only. Then,
 
∂g ∂B ∂A
A = − g
∂y ∂x ∂y
 2  ∂g  
3x y + y 2 = 3x2 + 2y g
∂y
∂g 1
= g
∂y y
g = y.

The integrating factor is y. With this integrating factor, the constraint equation is rewritten as
 2 2   
3x y + y 3 dx + 2x3 y + 3xy 2 dy = 0,

which integrates to
f = x3 y2 + xy 3 + C.

Example 4.4 Ginsberg Example 6.2.


Two bars, pinned at B, move in the horizontal plane subject only to the restriction that the velocity of
end C must be directed toward end A. Determine the corresponding velocity constraint. Is this constraint
holonomic.
Solution: It is a 4 DOF system, and let us use xB , yB , θ1 , and θ2 (we could choose, xA , yA , xB , yB ). Let
us write the position vector and the velocity of point C in terms of these generalized coordinates,

rA/C = −l (cos θ1 + cos θ2 )ı̂ − l (sin θ1 + sin θ2 ) ̂


vC = vB + θ̇1 k̂ × rC/B
   
= ẋB − lθ̇1 sin θ1 ı̂ + ẏB + lθ̇ 1 cos θ1 ̂.

The constraint that the velocity vector points to A can be written mathematically as

rA/C × vC = 0
   
−l (cos θ1 + cos θ2 ) ẏB + lθ̇ 1 cos θ1 + l (sin θ1 + sin θ 2 ) ẋB − lθ̇1 sin θ 1 = 0
l (sin θ1 + sin θ 2 ) ẋB − l (cos θ1 + cos θ2 ) ẏB − l2 {1 + cos (θ2 − θ1 )} θ̇1 = 0

The last line in the previous equation is the constraint equation written in Pfaffian form. If we let q1 = xB ,
q2 = yB , q3 = θ1 , and q4 = θ 2 , we can write

a1 dq1 + a2 dq2 + a3 dq3 + a4 dq4 = 0,

where a4 = 0. This is a BIG hint because we cannot find an integrating factor g (other than zero) that makes
∂ga1 ∂ga4
= .
∂q4 ∂q1
Therefore, it is not a perfect differential, and the constraint is nonholonomic. In addition, I am not even
sure how to enforce this type of constraint. How should I even draw constraint force?

There is no general method finding the integrating factor. Sometimes, it may be difficult to say whether
the constraint is holonomic or not. In this class, we will usually deal with holonomic constraint.
The force required to enforce the constraint is called the constraint force. For a simple pendulum, the
tension is the constraint force. One thing that you should know about the constraint force is that if the
6 CHAPTER 4. ANALYTICAL MECHANICS (BASIC CONCEPTS)

Figure 4.4: Rotating shaft and point mass.

constraint is time independent (scleronomic) holonomic, the constraint force does NO WORK. This is easy
to prove. The constraint force F is perpendicular to the configuration space, and therefore, can be written
as
∇f
F = F ,
|∇f|
where f is the time independent holonomic constraint equation. The incremental work done is given by
N

dW = Fi dqi ,
i=1

where N is the number of coordinates used and


F ∂f
Fi = .
|∇f | ∂qi
Then,
N
F  ∂f
dW = dqi
|∇f | i=1 ∂qi
 
F ∂f
= df − dt .
|∇f | ∂t
For a holonomic constraint where f = 0, df = 0. If f is time independent (scleronomic, ∂f /∂t = 0. Therefore,
incremental work done by a time independen holonomic constraint force is zero. If f is a explicit function
of time (rheonomic constraint), the work is not equal to zero. The work is given by
 
F ∂f
dW = df − dt
|∇f| ∂t
F ∂f
= − dt = 0.
|∇f | ∂t
It should be noted that incremental work is done through incremental time. That is, if the
constraint force of a holonomic system does work, it takes time.

Example 4.5 Consider a shaft and a mass attached to the cord. The cord is wrapping around the shaft of
radius ro . Find the work done by the tension (constraint force).
Solution: The velocity and acceleration of the mass is given by

v = rθ̇êt
2
a = −rθ̇ êr
4.1. VOCABULARY 7

The tension can be found from force balance in the r direction,


2
T = mrθ̇ .

From watching figure skating a lot, we know that the skater spins faster as she brings her arms closer to
the body. Therefore r and θ̇ are related. The relationship can be found by writing conservation of momentum
statement in the vertical direction,

Hz = (r × mv) · k̂ = h (constant)


h = mr2 θ̇. (4.2)

Now, the work done to the system is



W = T dr

2
= mrθ̇ dr,

where θ̇ can be replaced by h/mr2 using conservation of momentum. Then,


ro 2
h
W = 3
dr.
L mr

Let us re-enforce the idea that the work done by constraint (of holonomic system) can only be done over
finite time by writing r as a function of time. From figure, r = L − ro θ. Differentiating, we have ṙ = −ro θ̇.
Let us substitute this into Equation 4.2 to obtain

h = −mr2 .
ro
Integrating, we can find
ro h
r3 = L3 − 3 t.
m
The work can be written as tf
ro h3
W =  5/3 dt.
ti m2 L3 − 3 rmoh
t

Virtual Displacement is denoted as δqi . It is a imaginary displacement (thought up in your head).


Since it was imagined, there is no time elapse or the time is frozen. There are a couple of properties that
you should be aware.
The first is that the virtual displacement must satisfy the constraint equation with the time frozen. For
example, consider a pendulum whose length is time dependent so that the constraint equation is given by
f = x2 + y 2 − L (t)2 . In Pfaffian form, we can write

∂L
xdx + ydy − L dt = 0,
∂t
which the actual incremental displacement dx must satisfy. However, since there is no time lapse, the virtual
displacements satisfy
xδx + yδy = 0.
Therefore, the virtual displacement in this case is NOT physical displacement. If the constraint equation is
scleronomic (time independent), the virtual displacement is physical. The second property is that the forces
are assumed to be constant.
Do Example 6.4 (introduce two methods (a) starting from r (b) starting from v)
8 CHAPTER 4. ANALYTICAL MECHANICS (BASIC CONCEPTS)

Figure 4.5: 2D Pendulum

It was said that constraint force (of holonomic system) does no work if the constraint is independent of
time. Since the time is frozen, and dt does not enter the constraint equation in the virtual world (where
virtual displacement and virtual work live), it should be noted that the constraint force of holonomic system
does no virtual work.
Let us write the virtual work done by the constraint force (besides the fact that it is zero). For now, let
us write R for the constraint force (synonymous with reaction force). The virtual work can be written as
 · δr,
δW = R
 In terms of generalized coordinates,
where we assumed there is only one constraint and one constraint force R.
we can write
N
∂r
δW = R· δqj ,
j=1
∂q j

where N is the number of coordinates used. For a particle, N is less than or equal to three. R · ∂r/∂qj is the
 and is denoted as as Cj so that we can write
generalized force associated with qj due to constraint force R
N

δW = Cj δqj .
j=1

Let us compare this equation with the constraint equation


N

aj δqj = 0. (4.3)
j=1

 is ’parallel’ to a. They are parallel in geometrical sense if q are displacements in orthogonal
This says that R
directions. Mathematically,
Cj = λaj for j = 1, ..., N, (4.4)
where the proportionality constant λ is called the Lagrange multiplier (yes, the same Lagrange multiplier from
Calculus). The Jacobian constraint matrix (a j ) tells us the direction of the constraint force or the contribution
of the constraint force to the generalized force. In addition, the virtual work done by  the constraint force
(which is zero) is proportional to the Pfaffian form of the constraint equation ( δW = λ N j=1 aj δqj ).
Let us consider an example with a 2D pendulum (Figure 4.5). The constraint equation is given by
xdx + ydy = 0,
where q1 = x, q2 = y, a1 = x and a2 = y.
The virtual work done by the constraint force is
N

δW = λ aj δqj (4.5)
j=1
= λxδx + λyδy. (4.6)
4.2. PRINCIPLE OF VIRTUAL WORK 9

Because q1 and q2 are physical displacements in the orthogonal directions, we also find that the constraint
force R can be written as
 = λxı̂ + λy̂.
R
In this case, the Lagrange multiplier is T /L.
For the case where there are M constraint equations, we can rewrite the constraint equation and constraint
force as
N

aij δqj = 0, for i = 1, ..., M
j=1
Rij = λaij , for j = 1, ..., N, and i = 1, ..., M,

where M is the number of constraints, and Rij refers to the constraint force in the j th direction that enforces
ith constraint.

4.2 Principle of Virtual Work


Let us consider a system of n particles. that are connected somehow. For a system in static equilibrium, the
sum of the applied forces must be equal to zero. For a system of n particles, the sum of the applied forces on
each particle must be equal to zero. Let us denote Fi′ as the resultant force on the ith particle. The resultant
force consists of applied force Fi on the ith particle, and the internal constraint force i on the ith particle.
R
Note that R i is sum of the all the internal constraint forces on the particle or Ri = n fij , where fij
j=1,j=i
is the internal force on the ith particle due to j th particle. The virtual work is given by
n

δW = Fi′ · δri , for n particles.
i=1
n 
 
= Fi + R
 i · δri .
i=1

For static equilibtrium, Fi + R


 i = 0, for i = 1, ..., n. Then, the virtual work is also zero for the static
equilibrium, δW = 0.
Let us write the virtual work in terms of generalized coordinates as
N

n
   ∂r
i
δW =  
Fi + Ri · δqj = 0,
j=1 i=1
∂qj

where N is the number of generalized coordinates. The coefficients of δqj must be zeros or

n
  ∂r
i
 
Fi + Ri · = 0 for j = 1, ..., N. (4.7)
i=1
∂qj

For simplicity sake, let us write


n
 ∂ri
Fi · = Qj
i=1
∂q j
n
 M

 i · ∂ri
R = λk akj ,
i=1
∂qj
k=1

where Qj is the generalized force associated with δqj due to the external forces Fi , i = 1, ..., n, and
λk akj is the contribution of the kth constraint force on the j th generalized coordinate. Note that if Fi are
10 CHAPTER 4. ANALYTICAL MECHANICS (BASIC CONCEPTS)

Figure 4.6: Two linked bars.

conservative, we have
N

δWconservative = −δV = Qi δq
i=1
∂V
= −Qi for conservative system.
∂qi
We know from the previous section that the virtual work done by the configuration constraint (holonomic
constraint) force is zero, or
 i · δri = 0 for i = 1, ..., n.
R
Then, Equation 4.7 can be written as
n
 ∂ri
Fi · = Qj = 0 for j = 1, ..., N. (4.8)
i=1
∂qj
If δqi are independent, that is, if N is same as the degrees of freedom, the coefficients of δqj in Equation
4.8 or the generalized forces (total N of them), Qj , must be zeros,
Qj = 0 for j = 1, ..., N.
However, if δqi are not independent, the constraint force in Equation 4.7 cannot be omitted.
In this case, we have
M

Qj − λk akj = 0 for j = 1, ..., N.
k=1
We will demonstrate this in Example 4.7.

Example 4.6 Baruh example 4.8


Find the equilibrium position of the two links as shown in Figure 4.6. The springs are unstretched when
both rods are horizontal. Both springs deflect vertically.
Solution: This is a 2 DOF system. We first have to decide which variables to use. Let us use x1 and x2
as generalized coordinates. The virtual work done by the gravity and the spring force is given by
 
x1 x1 + x2
δW = m1 gδ + m2 gδ − k1 x1 δx1 − k2 x2 δx2 ,
2 2
   
1 1 1
= m1 g + m2 g − k1 x1 δx1 + m2 g − k2 x2 δx2 .
2 2 2
Since δW = 0 for static equilibrium,
1 1
m1 g + m2 g − k1 x1 = 0
2 2
1
m2 g − k2 x2 = 0.
2
4.2. PRINCIPLE OF VIRTUAL WORK 11

θ m
P

Figure 4.7:

x1 and x2 can be obtained by solving for these two equation.


If we wanted to solve this problem in terms of θ1 and θ 2 , we write the virtual work in terms of θ 1 and
θ2 . The displacements, x1 and x2 are given by
x1 = L1 sin θ1
x2 = L1 sin θ1 + L2 sin θ2 ,
and the virtual displacements are given by
δx1 = L1 cos θ 1 δθ1
δx2 = L1 cos θ 1 δθ1 + L2 cos θ 2 δθ2 .
Substituting the displacements and the virtual displacements into the vitual work expression, and setting
the coefficients of of δθ1 and δθ2 to zeros will result in two algebraic equations in terms of θ1 and θ2 .

Example 4.7 Consider a simple pendulum shown. Find the equilibrium position.
Solution: This is an 1 DOF system. Let us use θ. The external forces are the gravity, mg, and the force,
P. The virtual work is given by
δW = mgδy + P δx
= mgδ (L cos θ) + P δ (L sin θ)
= (−mgL sin θ + P L cos θ) δθ.
Setting coefficient of θ to zero, we have
P
tan θ =
.
mg
Let us solve this problem using x and y, where x and y are related by the constraint equation,
x2 + y 2 = L2 . (4.9)
The virtual work is given by
δW = mgδy + P δx + λxδx + λyδy.
The virtual work by the constraint force is obtained geometrically (with λ = −T /L.)or from Equation 4.6
Then,
δW = (P + λx) δx + (mg + λy) δy = 0.
We know that the virtual work done by the constraint force T is zero or λxδx+λyδy = 0, but the contribution
of the virtual work due to the constraint force to each term are not zeros. Setting the coefficients of δx and
δy zeros, we have
P + λx = 0 (4.10)
mg + λy = 0. (4.11)
12 CHAPTER 4. ANALYTICAL MECHANICS (BASIC CONCEPTS)

Omitting the virtual work due to the constraint force will results in P = 0 and mg = 0, which is obviously
wrong. With the constraint equation (Equation 4.9), these equations (Equations 4.10 and 4.11) constitute
three simultaneous equations with three unknowns, x, y and λ. Note that if we used solved this problem in
terms of θ, we have one equation to solve, whereas if we solved this problem in terms of x and y, we have
three equations to solve simultaneously. In general, if N coordinates are used and they are related by M
constraints, the system has N − M degrees of freedom and we need to solve N + M algebraic equations to
find the equilibrium configuration. The unknowns are the N coordinates and M constraint forces.

4.3 D’Alembert’s Principle


D’Alembert’s principle extends the Principle of Virtual Work to the dynamic case. Again, consider n particle
systems. If the system is not rest, we can write Newton’s second law as

dpi
Fi′ = , i = 1, ..., n,
dt

where pi is the linear momentum for the ith particle ( pi = mai ), and F ′ is the resultant force that include

the applied force, Fi , and the internal constraint force, R i .Then, we can rewrite the dynamic equilibrium
equation as
 i − dpi = 0.
Fi + R
dt
Multiplying by the virtual displacement of the ith particle, and summing over all the particles, we have

n  
 i − d
Fi + R
pi
· δri = 0. (4.12)
i=1
dt


The term ni=1 R  i δri is the virtual work done by the constraint forces, and we already have established that
it equals zero. Then, we are left with

n  
 d
pi
Fi − · δri = 0, (4.13)
i=1
dt

which is the D’Alembert’s Principle.


D’Alembert’s Principle can be written in terms of generalized coordinates using

N
∂ri
δri = δqj
j=1
∂qj

we can rewrite Equation 4.12 as

N  n  
 i − d
Fi + R
pi
·
∂ri
δqj = 0.
j=1 i=1
dt ∂q j

We recognize that
n
 ∂ri
Fi · = Qj
i=1
∂q j

n
 M

 i · ∂ri
R = λk akj .
i=1
∂qj
k=1
4.4. D’ALEMBERT’S PRINCIPLE FOR A RIGID BODY 13

If the generalized coordinates are independent, the contribution of the constraint force, Cj , disappears and
the coefficients of δqj can be set to zeros or
n
 d
pi ∂ri
Qj − · = 0, for j = 1, ..., N.
i=1
dt ∂qj

Again, if the generalized coordinates are not independent, the work by the constraint forces must be added.
Then, the N differential equations are given by
M
 n
 d
pi ∂ri
Qj + λk akj − · = 0, for j = 1, ..., N.
i=1
dt ∂qj
k=1

With M constraint equations given by


N

λk akj δqj = 0, for k = 1, ..., M,
j=1

We have N +M differential algebraic equations (DAE) that need to be solved (in terms of qi (t) for i = 1, ..., N
and λk for k = 1, ..., M ). Note that this equation can be used for nonholonomic system as well if
the constraint force does not do virtual work.
A system described by N coordinates with M constraints relating the coordinates has N − M degrees
of freedom. The equations of motion for this system constitutes either (a) N − M second order differential
equations in terms of N − M independent coordinates, or (b) a set of N + M differential algebraic equations
with N second order differential equations and M algebraic constraint equations in terms of N coordinates
and M Lagrange’s multipliers. As we can see, it is much easier to solve for the N −M second order differential
equations than N + M differential algebraic equations. Therefore, we should be in the habit of using the
number of coordinates equal to the system degrees of freedom. The exception applies when we are interested
in finding the constraint forces explicitly.

4.4 D’Alembert’s Principle for a Rigid Body


Now, let us extend D’Alembert’s Principle to rigid bodies. Consider a rigid body rotating at ω  and whose
center of gravity (CG) is translating with position vector rG . We can write the displacement, velocity and
acceleration as

ri = rG + ρi ,


vi = vG + ω × ρi
ai = aG + ω × (ω × ρi ) + α × ρi

where ρi is the distance between the ith particle and the center of gravity, and therefore, is constant. Let us
assume that
ω
 = θ̇1 ê1 + θ̇2 ê2 + θ̇3 ê3 .
Then, the virtual displacement becomes

δri = δrG + δθ × ρi ,

where
δθ = δθ 1 ê1 + δθ 2 ê2 + δθ 3 ê3 .
D’Alembert’s priniciple in Equation 4.13 for a continuous system (with infinite particles) can be rewritten
as (Fi → dF, mi → dm, ρi → ρ)  
0= dF − adm · δri
14 CHAPTER 4. ANALYTICAL MECHANICS (BASIC CONCEPTS)
   
0= dF − (aG + ω × (ω × ρ) + α × ρi ) dm · δrG + δθ × ρ

Note that

dF · δrG = F · δrG ,

where F is the resultant force on the rigid body.


   
dF · δθ × ρ = δθ · ρ × dF


= δ θ · ρ × dF

= δθ · M
 G.


aG δrG dm = maG δrG ,

where m is the total mass of the system.



ω × (ω × ρ) · δrG dm = δrG · ω × (ω × ρdm) = 0
 
α × ρ · δrG dm = δrG · α × ρdm = 0

because of definition of center of gravity.


 
{ω × (ω × ρ)} · δθ × ρ dm = 0,

  
because ω × (ω × ρ) ⊥ δθ ×ρ. The last term, (α × ρ) · δθ × ρ dm, is a bit difficult. For now, let us assume
that the rotation is about the z axis so that we can write

δθ = δθ k̂
α = θ̈ k̂.

Then, we can write


 
(α × ρ) · δθ × ρ dm
   
= θ̈k̂ × ρ · δθ k̂ × ρ dm

= θ̈δθ ρ2 dm

= θ̈δθIG3 ,

where IG3 is the centroidal mass moment of inertia about the z axis. Then, we can write D’Alembert’s
principle for a rigid body as
   
F − maG · δrG + MG − IG3 θ̈ · δθ = 0.
4.4. D’ALEMBERT’S PRINCIPLE FOR A RIGID BODY 15

  
Let us go back to the last term, (α × ρ) · δθ × ρ dm. This can be simplified if we use the tensor notation
(also known as indicial notation, Einstein notation).
 
(α × ρ) · δθ × ρ dm

= ǫijk ǫipq αj δθ p ρk ρq dm

= (δ jp δ kq − δ jq δ kp ) αj δθp ρk ρq dm

2
= αj δθ j ρk dm − αj δθk ρk ρj dm,

When it is written out, we have


 
 2 2

α1 ρ2 + ρ3 dm − α2 ρ2 ρ1 dm − α3 ρ3 ρ1 dm δθ1
 
 2 2

+ α2 ρ1 + ρ3 dm − α3 ρ2 ρ1 dm − α1 ρ3 ρ2 dm δθ 2
 
 2 2

+ α3 ρ1 + ρ2 dm − α2 ρ3 ρ1 dm − α1 ρ3 ρ2 dm δθ 3 .

This can be rewritten as


G
dH
· δθ,
dt
where
G  
dH  2 2

= α1 ρ2 + ρ3 dm − α2 ρ2 ρ1 dm − α3 ρ3 ρ1 dm ê1
dt
 
 2 2

+ α2 ρ1 + ρ3 dm − α3 ρ2 ρ3 dm − α1 ρ1 ρ2 dm ê2
 
 2 2

+ α3 ρ1 + ρ2 dm − α2 ρ3 ρ2 dm − α1 ρ3 ρ1 dm ê3 .

In matrix form, we can write the angular momentum as

{HG } = [IG ] {ω} ,

where   2    
ρ2 + ρ23 dm  −
 2 ρ2 ρ12dm
 −  ρ3 ρ1 dm
[IG ] =  −  ρ1 ρ2 dm ρ1 + ρ3 dm  −
 2 ρ2 ρ32dm

.
− ρ3 ρ1 dm − ρ2 ρ3 dm ρ1 + ρ2 dm
Then, we can write D’Alembert’s principle for a rigid body as
   
F − maG · δrG + M G − d H  G · δθ = 0
dt
{HG } = [IG ] {ω} .

Example 4.8 Bead on a Hoop


Consider a bead that is free to slide along the hoop that is rotating at Ω as shown in Figure 4.8. Obtain
the equation of motion using D’Alembert’s principle.
Solution: D’Alembert’s principle says
 
F − ma · δr = 0.
16 CHAPTER 4. ANALYTICAL MECHANICS (BASIC CONCEPTS)

Figure 4.8: A bead on a rotating hoop.

Let us use θ as the generalized coordinate. The only external force on the bead is due to gravity so that
F = −mgb3 .
The position vector is given by
r = R sin θb̂2 − R cos θ b̂3 ,
where b̂i is rotating at Ωb̂3 . The velocity and accelerations are given by
v = −RΩ sin θ b̂1 + Rθ̇ cos θb̂2 + Rθ̇ sin θb̂3 (4.14)
  2    2 
a = −2Rθ̇Ω cos θ b̂1 + −R sin θ θ̇ + Ω2 + Rθ̈ cos θ b̂2 + Rθ̇ cos θ + Rθ̈ sin θ b̂3 .

From the velocity expression, we can obtain the virtual displacement expression,
δr = R cos θδθ b̂2 + R sin θδθb̂3 .
Assembling terms,
 
F − ma · δr = 0
  2    2 
−m −R sin θ θ̇ + Ω2 + Rθ̈ cos θ R cos θδθ − m Rθ̇ cos θ + Rθ̈ sin θ R sin θδθ − mgR sin θδθ = 0
 
−mR2 θ̈ − mgR sin θ + mR2 Ω2 cos θ sin θ δθ = 0.

The equation of motion is given by


mR2 θ̈ − mR2 Ω2 cos θ sin θ + mgR sin θ = 0.

Using D’Alembert’s principle, reaction forces (constraint forces such as normal forces) never enter the
formulation. It is much easier to use D’Alembert’s principle than Newton’s second law. This will be more
pronounced if the system is more complicated.

4.5 Hamilton’s Principle


We can derive Hamilton’s principle from D’Alembert’s principle for a n particle system as follows. Let us
start with Equation 4.13, which is conveniently written below.
n  
d
pi
Fi − · δri = 0 (4.15)
i=1
dt
4.5. HAMILTON’S PRINCIPLE 17

where we recognize the first term as the virtual work,


n

Fi · δri = δW.
i=1

The second term can be written as


d
pi
· δri
dt  
d dri
= mi · δri
dt dt
    2 
d dri 1 dri
= mi · δri − δ mi
dt dt 2 dt
 
d dri
= mi · δri − δTi .
dt dt
Then, Equation 4.15 can be written as
n
  
d dri
δW + δT = mi · δri .
i=1
dt dt

When above expression is integrated over time, we have


tf
tf n

dri 

(δW + δT ) dt = mi · δri  .
ti i=1
dt 
ti

The last equation is known as Hamilton’s principle of varying action.


A special case of Hamilton’s principle is obtained when the configurations at the initial and the final time
are specified. That is, if δri (ti ) = δri (tf ) = 0, then the Hamilton’s principle is reduced to
tf
(δW + δT ) dt = 0. (4.16)
ti

Above equation (Equation 4.16) is also called the Extended Hamilton’s principle.
Recall that work can be done by either conservative and nonconservative force. The work done by
conservative force is also negative potential energy. Then, we can write
tf
(δT − δV + δWn ) dt = 0, (4.17)
ti

where δWn is virtual work done by any force other than conservative forces. We often call T − V Lagrangian
or L. When δWn = 0, we can write tf
δLdt = 0,
ti
which is called the Hamilton’s principle or Principle of least action (first stated by Lagrange). When will δWn
be zero? This is the case when (of course) the system is conservative. In addition, this is when constraint
force does not do virtual work. The constraint force does not do virtual work if the system is holonomic.
Note that constraint forces of nonholonomic system can do virtual work.
t
Let us examine the meaning of Hamilton’s principle. The equation δ tif (T + W ) dt = 0 says that ‘Out of
many
 tf possible path that dynamic system can take, the true path is the one that renders the action integral,
ti (T + W ) dt, stationary.’ The stationary value turns out to be the minimum.
If we are using more coordinates than the number of degrees, the work must include that of the constraint
forces. In that case, we would have
tf  M

δT − δV + δWn + λk akj δqj dt = 0,
ti k=1
18 CHAPTER 4. ANALYTICAL MECHANICS (BASIC CONCEPTS)

where M constraint equations are given by


N

λk akj δqj = 0, for k = 1, ..., M.
j=1

4.6 Lagrange’s Equation


The Lagrange equation are also known as Euler-Lagrange equation. These can be derived easily from either
D’Alembert’s principle or Hamilton’s principle. Let us start from Equation 4.17,
tf
(δT − δV + δWnc ) dt = 0.
ti

The kinetic energy is a function of q and q̇ at most, and the potential energy is a function of q at most.
Then, we can write
N
 ∂T ∂T
δT (q, q̇) = δqi + δ q̇i
i=1
∂qi ∂ q̇i
N
 ∂V
δV (q) = δqi
i=1
∂qi
N

δWnc = Qnc i δqi ,
i=1

where Qnc i is the generalized force due to forces other than conservative forces. Then, we can write
N tf  
∂T ∂T ∂V
0= δqi + δ q̇i − δqi + Qnc i δqi dt.
i=1 ti
∂qi ∂ q̇i ∂qi

Integrating by parts, we have


N
 tf  
∂T d ∂T ∂V
0 = δqi − δqi − δqi + Qnc i δqi dt
i=1 ti
∂qi dt ∂ q̇i ∂qi

N
∂T tf

+ δqi  ,
i=1
∂ q̇i 
ti

where the terms evaluated at the boundaries are zeros (because δqi = 0 at boundaries). Then,
 N tf  
∂T d ∂T ∂V
0= − − + Qnc i δqi dt,
i=1 ti
∂qi dt ∂ q̇i ∂qi

which results in N equations (same number as the number of generalized coordinates used) given by
 
∂T d ∂T ∂V
− − + Qnc i = 0 for i = 1, ..., N.
∂qi dt ∂ q̇i ∂qi
This is same as
∂L d ∂L
− + Qnc i = 0 for i = 1, ..., N.
∂qi dt ∂ q̇i
If the system has fewer than N degrees of freedom, some of the generalized coordinates must be related
through constraint equations. In that case, the contribution of constraint force on generalized force must be
taken into account. Using the Lagrange multipliers, for a system with M constraint, we have then
 M
∂L d ∂L
− + Qnc i + λk aki = 0, for i = 1, ..., N.
∂qi dt ∂ q̇i
k=1
4.6. LAGRANGE’S EQUATION 19

where aki is the Jacobian constraint matrix.


If you recall the pendulum problem, the virtual work done by the tension is zero so that we can write

δWtension = T · δr = 0.

When written in terms of δx and δy, we have (Equation 4.6)

δWtension = λxδx + λyδy = 0

where λ = T/L. As a whole, δWtension = 0, but the contribution of the forces to the generalized forces in
the x and y directions are NOT zeros.
One should also note that the kinetic energy for a rigid body is given by
1 T
T = mv 2 + {ω} [IG ] {ω} .
2 G

Example 4.9 Consider the bead problem again. Find the equation of motion using Hamilton’s principle
and Lagrange’s equation.
Solution: The velocity is given by Equation 4.14 or

v = −RΩ sin θ b̂1 + Rθ̇ cos θb̂2 − Rθ̇ sin θb̂3 .

Then, the kinetic energy is given by


1  2

T = mR2 Ω2 sin2 θ + θ̇ .
2
The potential energy is given by
V = mgR (1 − cos θ) .
The Lagrangian is given by
1  2

L = mR2 Ω2 sin2 θ + θ̇ − mgR (1 − cos θ) .
2
Using Hamilton’s principle,
tf  
1 2 2 2 2
0 = δ mR Ω sin θ + θ̇ − mgR (1 − cos θ) dt
ti 2
tf  
= mR2 Ω2 sin θ cos θδθ + mR2 θ̇δ θ̇ − mgR sin θδθ dt
ti
tf  
= mR2 Ω2 sin θ cos θδθ − mR2 θ̈δθ − mgR sin θδθ dt
ti
tf

+ mR2 θ̇δθ 
ti
tf  
= mR2 Ω2 sin θ cos θδθ − mR2 θ̈δθ − mgR sin θδθ dt.
ti

Then,
mR2 Ω2 sin θ cos θ − mR2 θ̈ − mgR sin θ = 0.
Using Lagrange’s equation, we have
d ∂L ∂L
− = 0
dt ∂ q̇i ∂qi
mR2 θ̈ − mR2 Ω2 sin θ cos θ + mgR sin θ = 0.
20 CHAPTER 4. ANALYTICAL MECHANICS (BASIC CONCEPTS)

Figure 4.9: 2D pendulum.

Example 4.10 Lagrange’s equation with a constraint


Consider a 2D pendulum shown in Figure 4.9. Use x and y as generalized coordinates with constraint
x2 + y 2 = L2 . Find equations of motion in terms of x and y.
solution: The kinetic and potential energy is given by
1  2 
T = m ẋ + ẏ2
2
V = −mgy.

The virtual work done by F is given by


δWF = F δx,
so that Qncx = F. The constraint equation for the virtual displacements in Pfaffian form is

2xδx + 2yδy = 0

Lagrange’s equations are given by


∂L d ∂L
− + Qnc x + λax = 0
∂x dt ∂ ẋ
∂L d ∂L
− + Qnc y + λay = 0,
∂y dt ∂ ẏ
where
1  2 
L = m ẋ + ẏ 2 + mgy
2
Qnc x = F
Qnc y = 0
ax = 2x
ay = 2y.

The resulting equations of motion are given by

−mẍ + F + λ2x = 0
mg − mÿ + λ2y = 0

with the constraint equation


x2 + y 2 = L2 ,
4.7. NATURAL AND NONNATURAL SYSTEMS 21

where we have three equations (two differential and one algebraic) and three unknowns (x, y, λ), which are
functions of time. This is known as Differential Algebraic Equations (DAE). There are books dedicated on
the subject. The easiest way to solve it is by multiplying the first equation by y and second by x, and
subtracting the second from the first to eliminate λ to obtain
−myẍ + mxÿ + F y − mgx = 0.
Differentiate the constraint equation twice to obtain
xẍ + yÿ + x2 + y2 = 0.
These two equation can be solved numerically for x (t) and y (t) .

4.7 Natural and Nonnatural Systems


Let us consider an n particle sytem again. In general, the position vector to the ith particle can be expressed
in terms of generalized coordinates and time or
ri = ri (q1 , q2 , ..., qN , t) for i = 1, ..., n,
where we did not exclude the possibility that the position is explicit function of time. We have seen such
case where a shaft has a constant angular velocity Ω in Figure 4.8. The angle is then, θ = Ωt. This is,
in reality, not possible. The motion of the shaft is affected by the motion of the bead through Newton’s
second law (more specifically through the normal forces, forces between the hoop and the bead). In order to
keep the angular velocity constant, there must be a control feedback system that keeps adjusting the torque
on the motor. In any case, this does not keep us from omitting the feedback system and simplifying it to
the system with the shaft rotating at constant speed. We should note that natural and nonnatural system
refers to the system models not the physical system itself. For the same hoop problem, if we let the angular
velocity vary and denote as φ̇, the system becomes natural.
This is an example of nonnatural system. We will show mathematical definition for natural and nonnat-
ural system in the following. We can write the velocity of the ith particle as
dri
vi =
dt
N
∂ri ∂ri
= q̇j + .
j=1
∂qj ∂t

The kinetic energy is given by


   
n N N
1  ∂ri ∂ri    ∂ri ∂ri 
T = mi  q̇j + · q̇j +
2 i=1 j=1
∂q j ∂t j=1
∂q j ∂t
 
n  N  N  N
1 ∂ri ∂ri ∂ri ∂ri ∂ri ∂ri 
= mi  · q̇j q̇k + 2 · q̇j + · .
2 i=1 j=1
∂q j ∂q k ∂t j=1
∂q j ∂t ∂t
k=1

Let us write
n
 ∂ri ∂ri
αjk = mi ·
i=1
∂qj ∂qk
n
 ∂ri ∂ri
βj = ·
i=1
∂t ∂qj
n
 ∂ri ∂ri
τ = · .
i=1
∂t ∂t
22 CHAPTER 4. ANALYTICAL MECHANICS (BASIC CONCEPTS)

Then, we can write


N N N
1  
T = αjk q̇j q̇k + β j q̇j + τ .
2 j=1 j=1
k=1

We call
N N
1 
T2 = αjk q̇j q̇k
2 j=1
k=1
N

T1 = β j q̇j
j=1
T0 = τ. (4.18)

T
Note that αjk is symmetric so that [α] = [α] . In addition, [α] is positive definite and invertable.
In natural system, the displacements are NOT explicit functions of time and T1 and T0 are
zeros. T1 is usually related to Coriolis effects and T0 to centrifugal effects. The matrix, αjk , is called the
mass matrix or inertial matrix, vector β j is called gyroscopic vector.
The nonnatural system can be catagorized into two groups. The first is when both T1 and T0 are nonzeros,
and the second is when only T0 is nonzero. The second case is treated as NATURAL system with modified
potential energy U = V − T0 . U is also called the dynamic potential energy.
For nonnatural system, equilibrum is defined as the state where all generalized velocities and accelerations
are zero. This allows velocities and accelerations (viewed from inertial frame) to be nonzero. Let us consider
the bead and hoop problem in Figure 4.8 again. The velocity is given by (Equation 4.14)

v = −RΩ sin θ b̂1 + Rθ̇ cos θb̂2 + Rθ̇ sin θb̂3 ,

and the kinetic energy is given by


1  2

T = mR2 Ω2 sin2 θ + θ̇ .
2
This is the second type of nonnatural system because it is missing T1 . Then, this problem can be thought of
as a natural system with the modified potential energy,

U = V − T0
1
= −mgR cos θ − mR2 Ω2 sin2 θ.
2

The equilibrium is when the potential energy is minimum (for conservative system) or

dU
= 0

mgR sin θ − mR2 Ω2 sin θ cos θ = 0
 
g − RΩ2 cos θ sin θ = 0

The possible equilibrium positions are

sin θ = 0 → θ = 0, π
 
g − RΩ2 cos θ = 0 → θ = cos−1 g/RΩ2 .
 
Equilibria at θ = 0 and π are unstable and at θ = cos−1 g/RΩ2 is stable. In terms of principle of virtual
work, we would write
δW = δ (−V + T0 + Wnc ) = 0.
4.8. GENERALIZED MOMENTUM 23

4.8 Generalized Momentum


The generalized momentum is defined by
∂L
πk = .
∂ q̇k
We know that the terms that contains q̇k in the Lagrangian are T2 and T1 . Then, we can write
∂T2 ∂T1
πk = + .
∂ q̇k ∂ q̇k
From Equation 4.18, we find that
N

πk = αkj q̇j + β k
j=1
{π} = [α] {q̇} + {β} .

There is one-to-one relationship between {q̇} and {π} (because [α] is invertable). Therefore, we can replace
{q̇} with {π} so that
L = L (q1 , · · · , qn , π1 , · · · , πn , t) .

4.9 First Integrals


The equations of motion can be written in two forms. For n degree of freedom system, we can express the
equations of motion as n set of second order differential equations. Another method of expressing the general
solution is to obtain 2n independent functions of the form

fi (q1 , · · · , qn , q̇1 , · · · , q̇n , t) = αi , i = 1, · · · , 2n,

where αi are constants. fi are called the first integrals or integrals of motion. The highest order in
the first integrals is the same as the highest order that occurs in T − V. Obtainining all 2n first integrals is
difficult or impossible. However, we can find a few first integrals in some special cases. We will discuss three
integrals of motion

1. The statement of conservation of energy, T + V = E. This is an integral of motion if there is no work


done by nonconservative force (or the system is conservative).
2. Conservation of (either linear or angular) momentum, ∂L/∂ q̇k = π k . Consider the case when the
Lagrangian is not a function of qk (such coordinate is called the cyclic or ignorable coordinate).
The Euler-Lagrange’s equation is given by
d ∂L
= Qnc k .
dt ∂ q̇k

If the kth generalized force due to non-conservative forces is zero, we have


∂L
= constant πk .
∂ q̇k

where πk is called the generalized momentum associated with the kth generalized coordinate, qk . This
is one of the integral of motion.
N
3. Jacobi integral, k=1 π k q̇k − L = h. The Jacobi integral is an integral of motion if the system is
conservative and L is not an explicit funciton of time. The time derivative of the Lagrangian can be
written as
N N
dL  ∂L  ∂L ∂L
= q̇k + q̈k + .
dt ∂qk ∂ q̇k ∂t
k=1 k=1
24 CHAPTER 4. ANALYTICAL MECHANICS (BASIC CONCEPTS)

The Euler Lagrange equation for the conservatiṽe system is given by


∂L d ∂L
− = 0.
∂qk dt ∂ q̇k
Substitituting ∂L/∂qk with d(∂L/∂ q̇k )/dt in the first equation, we have

N   N
dL d ∂L ∂L ∂L
= q̇k + q̈k +
dt dt ∂ q̇k ∂ q̇k ∂t
k=1 k=1
N  
d ∂L ∂L
= q̇k + .
dt ∂ q̇k ∂t
k=1

Then,  
N

d ∂L ∂L
L− q̇k = .
dt ∂ q̇k ∂t
k=1
If L is not an explicit function of t, the right hand side becomes zero. Integrating, we obtain the Jacobi
integral, h.
N
∂L
L− q̇k = h.
∂ q̇k
k=1
In calculus of variation, this is called the Beltrami Identity. The Jacobi integral can be written in
a simpler form realizing
N
 ∂L
q̇k = 2T2 + T1 .
∂ q̇k
k=1
Then,
h = 2T2 + T1 − L = T2 + U.
For the natural system, the Jacobi integral is reduced to the statement of conservation of energy.

4.10 Ignorable Coordinates and Routh’s method


If a system has n degrees of freedom and l of those coordinates are ignorable, Routh’s method allows us to
reduce the n equations of motion to n− l equations by eliminating l ignorable coordinates from the equations
of motion. Let us replace the ignorable coordinate q̇k with generalized momentum, πk , so that

L = L (q1 , · · · , qn−l , q̇1 , · · · , q̇n−l , πn−l+1 , · · · , πn , t) .

The Euler-Lagrange equations are given by


∂L d ∂L
− = 0 for i = 1, ..., n − l (4.19)
∂qi dt ∂ q̇i
∂L
= 0 for k = n − l + 1, ..., n. (4.20)
∂πk
Let us define Routhian, R, by

R (q1 , · · · , qn−l , q̇1 , · · · , q̇n−l , πn−l+1 , · · · , πn , t)


n

= L (q1 , · · · , qn−l , q̇1 , · · · , q̇n−l , πn−l+1 , · · · , πn , t) − πk q̇k ,
k=n−l+1

where the ignorable coordinate q̇k is replaced by generalized momenta. Comparing with the Lagrangian, we
observe
∂L ∂R ∂L ∂R
= and = , for i = 1, ..., n − l.
∂qi ∂qi ∂ q̇i ∂ q̇i
4.10. IGNORABLE COORDINATES AND ROUTH’S METHOD 25

Then, Lagrange’s equation in Equation 4.19 can be written as


∂R d ∂R
− = 0 for i = 1, ..., n − l. (4.21)
∂qi dt ∂ q̇i
For the ignorable coordinates, we have
∂R ∂L ∂ q̇k
= − q̇k − πk for k = n − l + 1, ..., n.
∂πk ∂π k ∂πk
We recognize, πk = ∂L/∂ q̇k . Then,

∂R ∂L ∂L ∂ q̇k
= − q̇k −
∂π k ∂πk ∂ q̇k ∂πk
∂L ∂L
= − q̇k −
∂πk ∂π k
= −q̇k for k = n − l + 1, ..., n. (4.22)

Procedure for Using Routhian

1. First identify the ignorable coordinates (coordinates that only q̇k appears in L).

2. Find the generalized momenta πk for the ignorable coordinates, and solve for q̇k .

3. Form Routhian,
n

R=L− πk q̇k ,
k=n−l+1

and express ignorable coordinates, q̇k , in terms of generalized momenta, πk .

4. Write Euler-Lagrange’s equation, 4.21.

Example 4.11 Consider the bead and hoop problem again. This time, let us assume that hoop has
a variable angular velocity φ̇. The kinetic and potential energies are given by
1 2 1 2 1 2
T = Ihoop φ̇ + mR2 φ̇ sin2 θ + mR2 θ̇
2 2 2
V = mgR (1 − cos θ) .

φ is the ignorable coordinate, and the generalized momentum πφ is constant and is given by

∂L  
πφ = = Ihoop + mR2 sin2 θ φ̇. (4.23)
∂ φ̇
The Lagrangian is given by
1  2 1 2
L= Ihoop + mR2 sin2 θ φ̇ + mR2 θ̇ − mgR (1 − cos θ) .
2 2
The Equation of motion is given by
∂L d ∂L
− = 0
∂θ dt ∂ θ̇
2
mR2 θ̈ + mgR sin θ − mR2 φ̇ sin θ cos θ = 0

with
πφ
φ̇ =  .
Ihoop + mR2 sin2 θ
26 CHAPTER 4. ANALYTICAL MECHANICS (BASIC CONCEPTS)

Then, the equation of motion in terms of θ is given by


 2
πφ
mR2 θ̈ + mgR sin θ − mR2 sin θ cos θ   =0
Ihoop + mR2 sin2 θ

Or, the Routhian is given by


n

R = L− πk q̇k
k=n−l+1
1 2 1 2 1 2
= Ihoop φ̇ + mR2 φ̇ sin2 θ + mR2 θ̇ − mgR (1 − cos θ)
2 2 2
2
−πk φ̇ .

Replacing φ̇ in terms of πφ in Routhian, we have


  1 π2φ 1 2
R θ, θ̇, πφ = −   + mR2 θ̇ − mgR (1 − cos θ) .
2 Ihoop + mR2 sin2 θ 2

Equation 4.22 is given by


∂R πφ
= −  = −φ̇,
∂πk Ihoop + mR2 sin2 θ
which is identical to Equation 4.23. The Euler-Lagrange’s equation in terms of Routhian (Equation
4.21) is given by

∂R d ∂R
− = 0
∂θ dt ∂ θ̇
π 2φ
−mR2 θ̈ +  2
2 mR sin θ cos θ − mgR sin θ = 0.
2 2
Ihoop + mR sin θ
Chapter 5

Analytical Mechanics (Advanced


Topics)

5.1 Hamilton’s Canonical Equations


5.2 Gibbs-Appell Equations for Quasicoordinates

27

S-ar putea să vă placă și