0% found this document useful (0 votes)
71 views6 pages

18.03 Study Guide and Practice Hour Exam II, March, 2006 I. Study Guide

The document provides study guides and practice exams for solving second order linear differential equations. It summarizes key solution methods including: - The exponential response formula to find particular solutions when the forcing term is an exponential function. - The resonant response formula to handle cases of resonance. - The exponential shift law to eliminate exponential terms in the forcing function. Sample problems from the practice exams demonstrate applying these methods to find particular and general solutions for different types of forcing terms.

Uploaded by

emmyrovic
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
71 views6 pages

18.03 Study Guide and Practice Hour Exam II, March, 2006 I. Study Guide

The document provides study guides and practice exams for solving second order linear differential equations. It summarizes key solution methods including: - The exponential response formula to find particular solutions when the forcing term is an exponential function. - The resonant response formula to handle cases of resonance. - The exponential shift law to eliminate exponential terms in the forcing function. Sample problems from the practice exams demonstrate applying these methods to find particular and general solutions for different types of forcing terms.

Uploaded by

emmyrovic
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 6

18.

03 Study Guide and Practice Hour Exam II, March, 2006


I. Study guide.

Homogeneous equations. The second order linear ODE m¨


x + bẋ + kx = q(t) models
a spring/mass/dashpot system, where an external force q(t) is acting on the mass. We
have always assumed m, b, and k are constant in time. The characteristic polynomial
is p(s) = ms2 + bs + k, and the operator p(D) = mD2 + bD + kI is a second order (if
m �= 0) LTI (linear time invariant, i.e. linear constant coefficient) differential operator.
The equation is homogeneous if q(t) = 0. Solutions then are given by the following table,
in which we suppose m = 1 (by dividing through by it if need be) and r1 and r2 are the
roots of p(s).

Name Overdamped Critically Damped Underdamped

Roots r1 �= r2 real r1 = r2 r1 = r2 not real

Condition b2 > 4k b2 = 4k b2 < 4k

Gen Real Sol c1 er1 t + c2 er2 t (c1 t + c2 )er1 t e−bt/2 (c1 cos(ωd t) + c2 sin(ωd t))
= Ae−bt/2 cos(ωd t − φ)

The names are only really appropriate when k, b ≥ 0. In this case, Re r1,2 ≤ 0, and
Re r1,2 < 0 if k > 0 and b �> 0. In that case all these solutions die off as t → ∞. In the
underdamped case, ωd = k − (b/2)2 is the damped circular frequency. The two forms
of the general solution in the nonreal root case are related
� by the standard triangle: A, φ
are the polar coordinates of the point (c1 , c2 ) (so A = c21 + c22 and tan φ = c2 /c1 ).
If p(D) is a second order LTI operator and x1 and x2 is any pair of solutions to p(D)x = 0
such that neither is a multiple of the other, then the general solution to p(D)x = 0 is
given by c1 x1 + c2 x2 for c1 , c2 constants. Such a pair of solutions is called “basic.” For
any initial condition x(a), ẋ(a), one can solve for c1 and c2 . This process is very easy
if x1 , x2 is “normalized at t = a,” x1 (a) = 1, ẋ1 (a) = 0, x2 (a) = 0, ẋ2 (a) = 1. Then
x = x(a)x1 + ẋ(a)x2 . Example: cos(ωn t), (1/ωn ) sin(ωn t) is a normalized pair of solutions
to the harmonic oscillator ẍ + ωn2 x = 0.
The key to solving homogeneous constant coefficient linear equations is to look for ex­
ponential solutions. If c �= 0, cert is a solution to p(D)x = 0 exactly when r is a root of
p(s), p(r) = 0. Usually a degree n polynomial p(s) has n distinct complex roots. When
this happens, there are n distinct nonzero exponential solutions for p(D)x = 0, and the
general solution is a linear combination of them.
Inhomogeneous equations. The superposition principle states that the general solu­
tion to p(D)x = q(t) is given by x = xp + xh where xp is any solution to p(D)x = q(t) and
xh is the general solution to p(D)x = 0 (called by EP the complementary solution). Also,
if p(D)x1 = q1 (t) and p(D)x2 = q2 (t), then p(D)x = c1 q1 (t) + c2 q2 (t) if x(t) = c1 x1 + c2 x2 .
We found particular solutions xp for various different input signals q(t). These methods
work for any p(D).
(1) The Exponential response formula ERF: A solution to p(D)x = ert is given by
xp = ert /p(r), as long as p(r) =
� 0.
Application: To solve p(D)x = q(t) with q(t) = cos(ωt) or sin(ωt), write down the new
ODE p(D)z = eiωt ; solve it using ERF; and extract the real (or if the input signal is
sine, the imaginary part). Case: x¨ + ωn2 x = cos(ωt) has solution xp = cos(ωt)/(ωn2 − ω 2 )
(and same with sine), as long as ω = � ωn . Case: if b and k are positive then there is just
one sinusoidal solution to x¨ + bẋ + kx = F cos(ωt). It is given by A cos(ωt − φ) where
A and −φ are the polar coordinates of the “complex gain” W (iω) = F/p(iω); that is,
W (iω) = Ae−iφ .
More generally, if q(t) = eat cos(ωt) or eat sin(ωt), solve p(D)z = e(a+iω)t and continue.
(2) Undetermined coefficients: If q(t) is polynomial of degree at most k, and p(0) = � 0,
then there is exactly one solution of p(D)x = q(t) which is polynomial of degree at most
k. Case: if q(t) = c, a constant, then x = c/p(0) is a solution.
(3) “Resonance” means that the exponent r is a root of p(s), so the ERF fails. In
that case we have the Resonant Response Formula: If p(r) = 0 but p� (r) = � 0, then
p(D)x = ert has as solution xp = tert /p� (r). See the Supplementary Notes or Notes and
Exercises for a more general statement.
Application: x¨ + ωn2 x = cos(ωn t) has solution xp = (t/2ωn ) sin(ωn t). Case: x¨ + bẋ = 1.
Since ert = 1 if r = 0, we find xp = t/b as long as b �= 0. In any case, solutions grow
faster that the exponential growth/decay predicted by the real part of the root.
(4) The exponential shift law ESL: p(D)(ert u) = ert p(D + rI)u. If q(t) = ert q1 (t)
where q1 (t) is some other function, then x = ert u is a solution to p(D)x = q(t) provided
that u is a solution to p(D − rI)u = q1 (t). ESL eliminates exponentials. Application: if
b �= 0, a solution to x¨ + ω 2 x = cos(ωt) is xp = (1/2ω)t sin(ωt). The function q1 (t) might
be polynomial for example.
Putting all this together, we have actually proven the following theorem:
Theorem. If q(t) is a linear combination of products of polynomials and exponential
functions, then all solutions to p(D)x = q(t) are too.
Here we mean to include complex linear combinations of products of polynomials with
complex coefficients and complex exponential functions. For example sin(t) = (eit −
ie−it )/2i is included.
II. Practice Hour Exams.

Exponential Response Formula:


xp = Aert /p(r) solves p(D)x = Aert
� 0.
provided p(r) =

Resonant Response Formula:


xp = Atert /p� (r) solves p(D)x = Aert
provided p(r) = 0 and p� (r) =� 0.

Exponential Shift Law:


p(D)(ert u) = ert p(D + rI)u.

First Practice Exam (50 minutes)

1. A spring/mass/dashpot system is modeled by 2ẍ + 2bẋ + 4x = −10.


(a) What is the steady state (constant) solution?
(b) For what values of the damping constant b does the system “ring,” i.e. oscillate?
(c) When it does oscillate, what is the pseudoperiod (in terms of b)?

2. Find a particular solution to x¨ + 4ẋ + 4x = e−2t cos(2t).

3. Find the solution to x¨ + 4ẋ + 4x = 4 such that x(0) = 0 and ẋ(0) = 0.

4. Find a particular solution to x¨ + 4ẋ + 4x = 8t2 + 8.

5. (a) Find the amplitude of the sinusoidal solution to x¨ + 2ẋ + 2x = cos(ωt), as a


function of the input signal circular frequency ω.
(b) For what value of ω is the phase lag zero?
(c) For what value of ω is the phase lag 90◦ ?

Second Practice Exam (90 minutes?)

1. Find ω, A, and φ such that xp = A cos(ωt− φ) is a solution to ẍ+2ẋ+15x = 2 cos(3t).

2. (a) For what value of ω does resonance occur in ẍ + 2x = cos(ωt)?


(b) For what value of c does critical damping occur in ẍ + cẋ + 4x = 0?

3. (a) What is the general real solution of ẍ + 4ẋ + 5x = 0?


(b) Find a particular solution to x¨ + ẋ + 2x = 2t2 + 2t + 4.

4. (a) If e−2t + 2e−t is a solution to x¨ + cẋ + kx = 0, what are the constants c and k?
(b) Same question if te−t is a solution instead.
(c) What are the exponential solutions of ẍ+2ẋ+2x = 0? Find the general real solution.
What is the damping type of this equation?
5. Suppose x¨ + 4x = e−t t10 . For a suitable constant r, if we write x as ert u then u
satisfies a differential equation of the form ü + cu̇ + ku = t10 . What is r, and what is this
new ODE? (Don’t try to solve it!)

6. True or False: For appropriate c, k, and q, both e−t and sin(t) are solutions to the
single equation ẍ + cẋ + kx = q(t).

7. (a) The substitution x = ert u (for an appropriate value of r) lets you replace the
ODE x¨ + 2ẋ + 2x = te−t sin(t) with a different ODE (for u) having right hand side t.
What is the new ODE?
(b) Find a solution of ẋ + x = t sin t by replacing it by a complex equation and solving
using the substitution x = ert u for appropriate r.
(c) Find a polynomial solution of x¨ + x = t2 + 1. What is the general solution?

8. (a) Find a periodic solution to x¨ + 2ẋ + 2x = 1 + 2 cos(t).


(b) What is the amplitude of the sinusoidal solution to x¨ + 2ẋ + 2x = 2 cos(t)? What is
the phase lag φ?
(c) Find a solution to ẍ − 4x = e2t .
III. Solutions.
First Practice Exam
1. (a) x = −5/2.

(b) The characteristic polynomial has roots −b/2 ± (b/2)2 − 2. They are complex as

long as (b/2)2 < 2, i.e. |b| < 2 2.
� �
(c) ωd = 2 − (b/2)2 so Pd = 2π/ωd = 2π/ 2 − (b/2)2 .
2. The equation is the real part of p(D)z = e(−2+2i)t , where p(s) = s2 +4s+4. p(−2+2i) =
(−2 + 2i)2 + 4(−2 + 2i) + 4 = −4, so by the ERF zp = e(−2+2i)t /(−4) and so xp = Re zp =
−(1/4)e−2t cos(2t). Other methods work too: try xp = e−2t (a cos(2t) + b sin(2t)); or use
ESL to eliminate e−2t .
3. xp = 1, so xp (0) = 1 and ẋp (0) = 0. For x = xp +xh to satisfy the given initial condition
we need xh (0) = −1 and ẋh (0) = 0. p(s) = (s + 2)2 so the general homogeneous solution
is xh = (c1 t + c2 )e−2t . ẋh = (−2c1 t + (c1 − 2c2 ))e−2t , so xh (0) = c2 , ẋh (0) = c1 − 2c2 .
This gives c2 = −1 and c1 = −2, so x = 1 + (−2t − 1)e−2t .
4. x = at2 + bt + c, ẋ = 2at + b, x¨ = 2a, so we want a, b, c such that 8t2 + 8 =
4at2 +(4b+8a)t+(4c+4b+2a). This gives a = 2, then b = −4, then c = 5: xp = 2t2 −4t+5.
5. (a) xp = Re eiωt /p(iω)

has amplitude |1/p(iω)|. p(iω) = (iω)2 +2iω+2 = (2−ω 2 )+2iω
so A = |1/p(iω)| = 1/ (2 − ω 2 )2 + 4ω 2 .
(b) Zero phase lag occurs when p(iω) is real (and positive), which happens only when
ω = 0.
(c) Phase lag of 90◦ occurs when p(iω) is purely imaginary
√ with positive imaginary part.
2
This happens only when ω = 2 and ω > 0, i.e. ω = 2.

Second Practice Exam


1. Start by writing down a complex equation having this as its real part: p(D)z = 2e3it ,
with p(s) = s2 + 2s + 15. By the key formula this has solution zp = (2/p(3i))e3it .
p(3i) = −9 + 6i + 15 = 6 + 6i. The clever√ way to solve √ the problem is to switch to
polar coordinates right away:√ p(3i) = (6 2)eiπ/4 , so zp = ( 2/6)ei(3t−π/4)
√ . The original
equation has solution xp = ( 2/6) cos(3t − π/4). Thus ω = 3, A = 2/6, and φ = π/4.

2. (a) It occurs when ω equals the natural frequency of the system, which is 2.
(b) It occurs when the two roots of the characteristic polynomial are equal, which is
when c = 4 (so both roots are −2)
3. (a) The roots of the characteristic polynomial p(s) = s2 + 4s + 5 are r = −2 ± i, so
x = e−2t (a cos t + b sin t).
(b) Try x = at2 + bt + c, so ẍ + ẋ + 2x = (2a) + (2at + b) + 2(at2 + bt + c) = 2at2 +
(2a + 2b)t + (2a + b + 2c) Equating this with 2t2 + 2t + 4 gives, successively, a = 1, b = 0,
c = 1: so xp = t2 + 1. This is easy to check.
4. (a) The only way this sum can occur is if both e−2t and e−t are solutions, so the roots
of p(s) = s2 + cs + k are −1 and −2: p(s) = (s − (−1))(s − (−2)) = s2 + 3s + 2: so c = 3,
k = 2.
(b) The only way te−t can be a solution is if −1 occurs as a double root: so p(s) =
(s + 1)2 = s2 + 2s + 1: so c = 2, k = 1.
(c) The roots of p(s) = s2 +2s+2 are −1± i, so the exponential solutions are e(−1+i)t and
e(−1−i)t (and their constant multiples). The general real solution is e−t (a cos t + b sin t).
This is an underdamped equation.
5. Take r = −1 and use ESL: e−t t10 = p(D)(e−t u) = e−t p(D − I)u, so p(D − I)u = t10 .
p(s) = s2 +4, so p(s− 1) = (s− 1)2 +4 = s2 − 2s+5, and the equation is u−
¨ 2u̇+5u = t10 .
Or substitute and compute directly.
6. False. Reason: If these are system responses to the same signal, then their difference
must be a system response to the null signal, that is, a homogeneous solution. But such
solutions are linear combinations of two exponentials (or, in the case of a repeated root,
linear combinations of an expnential and t time the same exponential). sin(t) − e−t is
not of this form.
7. (a) The equation is the imaginary part of p(D)z = te(−1+i)t with p(s) = s2 + 2s + 2.
Take z = e(−1+i)t u and use ESL: p(D)z = p(D)(e(−1+i)t u) = e(−1+i)t p(D + (−1 + i)I)u.
Equating and canceling the exponential gives p(D + (−1 + i)I)u = t. We compute
p(D + (−1 + i)I) = (D + (−1 + i)I)2 + 2(D + (−1 + i)I) + 2I = D2 + 2iD, so the new
ODE is ü + 2iu̇ = t. This can also be done by direct substitution.
(b) This equation is the imaginary part of p(D)z = ż + z = teit . Take z = eit u apply
ESL: p(D)z = p(D)(eit u) = eit p(D + iI)u. Equating and canceling the exponential gives
p(D + iI)u = t. We compute p(D + iI) = (D + iI) + I = D + (1 + i)I, so the equation
is u̇ + (1 + i)u = t. This can be obtained also by direct substitution. Use undetermined
coefficients, with u = at + b: u̇ = a, so the equation reads a + (1 + i)(at + b) = t, so
a = 1/(1 + i) = (1 − i)/2 and b = −a/(1 + i) = i/2. Thus up = ((1 − i)/2)t + (i/2). Then
zp = (((1 − i)/2)t + (i/2))(cos t + i sin t), and xp = �zp = (t/2)(sin t − cos t) + (1/2) cos t.
(c) Try x = at2 + bt + c; ẋ = 2at + b, x¨ = 2a, so x¨ + x = at2 + bt + (c + 2a). For this to
be equal to t2 + 1 we must have a = 1, b = 0, and so c = 1 − 2a = −1: xp = t2 − 1. The
general solution is x = (t2 − 1) + a cos t + b sin t.
8. (a) Use superposition: the signal 1 has periodic solution given by the constant
1/2. For the other term, once again I recommend replacing the problem with a complex
one: p(D)z = 2eit , which has solution zp = 2eit /p(i) by the Exponential Response
Formula. p(s) = s2 + 2s + 2 so p(i) = −1 + 2i + 2 = 1 + 2i and zp = (2/(1 + 2i))eit =
((2 − 4i)/5)(cos t + i sin t) has real part xp = (2/5) cos t + (4/5) sin t. Thus the original
problem has a periodic solution (1/2) + (2/5) cos t + (4/5) sin t.
(b) We just saw that ẍ + 2ẋ + 2x = 2 sin(2t) has for √ a iφ
solution the imaginary part
it
of (2/p(i))e
√ , and that p(i) = 1 + 2i. Write √ p(i) = 5 e , so φ = arctan(2). Then
zp = (2/ 5)ei(t−φ) , so xp has amplitude 2/ 5 and phase lag φ = arctan(2). Alternatively,
use the usual triangle with sides 2/5 and 4/5 from the final form of the solution to (a).
(c) 2 is a characteristic root here so we can’t apply the Exponential Response Formula.
We are in a resonance situation, and multiplying by t comes to the rescue: Try x = Ate2t .
Then ẋ = A(2t+1)e2t , x¨ = A(2(2t+1)+2)e2t = A(4t+4)e2t and x−4x
¨ = 4Ae2t . (This can
2t 2t
also be done using ESL: p(D)(e At) = e p(D + 2I)(At), and since p(D) = (D + 2I)(D −
2I), p(D + 2I) = (D + 4I)D and p(D + 2I)At = (D + 4I)D(At) = (D + 4I)A = 4A.)
Setting this equal to e2t we find A = 1/4 and xp = (1/4)te2t .

You might also like