Errata for Book Operations Research Models and Methods

Paul A. Jensen and Jonathan F. Bard

September 15, 2010

Corrections marked with a (1) superscript were added to the text after the first printing.
Other corrections have not been made. Please send corrections or suggestions to one of
the authors with addresses at http://www.me.utexas.edu/~jensen/ORMM/ . Updated versions
of the errata are maintained on the web site.

Chapter 2
Page 43, car rental problem. Change first sentence in Solution section to: “The solution,
whose corresponding objective function value is z* = 2150, is shown…”

Table 2.15. The Name of Variable Number 8 should be: MTuW.

Page 45, Solution section. Change first sentence to: “The variable … provides the
solution, whose corresponding objective function value is z* = 866,050.”

Page 50, paragraph 3, line 3. Change to: “… one column per variable (the columns for
PA, PB, and PC are omitted).”

Ex. 15. The solution to the LP is not integer. There is no guarantee of integrality when the
series of 1’s in each column of the LP is broken by 0’s. An integer solution is obtained by
requiring the Solver to give integer answers. This makes the model into an integer
program.

On page 36 in 3rd paragraph change “hours” to “minutes”.

Chapter 3

Page 61 in the last line should be “m-dimensional null vector” instead of “k-dimensional
null vector”
Page 77 in the last of tables 3.6 (at the bottom of the page). Both instances of "–0.5" in
the x4 column should be positive.
Page 79, Table 3.7. The last basic variable should be x5 not x1.
Page 81, Table 3.11. The entry for x3 in row 2 should be 0 rather than 1.
Page 82, In the middle of the page the reference to Definition 4 should be to Definition 5.
Page 83, Table 3.12. The row for x5 (row 3) should be "0 1 0 0 0 1" rather
than "1 0 0 0 1 1"
Page 83, Table 3.14. The circle should be around the 0.25 in the x3 column and the arrow
should be pointing at the x3 column not the x4 column.
Page 83, Table 3.15. xI should be x3.

Page 87. The model at the bottom should be

1
 ˆ  i
Minimize w
iF

n
subject to a x
j 1
ij j  i  bi , i  1,..., m

xj  0, j = 1,…,n; i  0, i = 1,…,m
Pages 87 and 88. Replace the last sentence on page 87 starting with “Stop and indicate
…” through the first paragraph on page 88 with
“Accordingly, the original problem is infeasible so the computations should be halted.
The alternative situation indicates that a BFS has been found. At this point, all nonbasic
artificial variables can be deleted along with all original problem variables xj whose final
phase 1 reduced costs are positive. This follows because there is no feasible solution to
the original problem when any of these variables are positive. By implication then, only
original problem variables with phase1 reduced costs equal to zero need be considered in
phase 2.
If some artificial variables remain in the basis at a zero level, however, then one
of the following two situations arise.
(i) When one or more entries in the corresponding row are nonzero (not counting the “1”
in the artificial column), the artificial variable can be left in the problem until it is
replaced by some original problem variable during phase 2. The value of all basic
artificial variables will remain at zero in all phase 2 solutions.
(ii) If all the entries in the corresponding row are zero except for a single “1” in
the artificial column, then that row corresponds to a redundant equation in the original
model. All such rows and the associated artificial columns can be deleted from the
tableau before entering phase 2.”

Page 88. Replace the first sentence after the heading Phase 2 with
“Delete the nonbasic artificial variables and any original problem variables with positive
phase 1 reduced costs from the model and revert to the original objective function”

Page 88 in Phase 1 of the Example, write as

Phase 1: Minimize w = + 1 + 2

(or Maximize ŵ = – 1 – 2)

Phase 2: Maximize z = –7x1 + 3x2

Page 89, Table 3.22, the 0' row is not correct. The coefficients should be 7 and –3 rather
than the other way.

Page 92, Table 3.25. In the Basic column, it should be x4 and x5, not x1 and x2.

2
Page 94, Table 3.30. The last entry in row 0 should be 46 (not 55).

On page 100 (last line) and 102, the table reference should be to Table 3.35 not Table
3.24.

Page 104, Exercise 5. Replace parts (b), (c) and (d) with

(b) Using elementary row operations put the tableau in the simplex form by making x1
and x4 basic. What is the value of all variables in the new tableau.
(c) From the new tableau predict the effects of increasing x5 by 0.5, by 1 and by 2 on z,
x1 and x4. Note that feasibility should be maintained after x5 is increased.
(d) From the new tableau predict the effects of increasing x3 by 0.5, by 1 and by 2 on z,
x1 and x4. Note that feasibility should be maintained after x5 is increased.

Page 106, Exercise 20: should be xj  0, not 0  xj  1. 1

Chapter 4

Page 125, Table 4.6. Bases 2, 3, 4, and 5 have typos:

#2 should be x1, xs2, xs3 not xs1, xs2, xs3;
#3 should be x2, xs2, xs3 not xs2, xs2, xs3;
#4 should be x1, xs1, xs3 not xs1, xs1, xs3;
#5 should be xs1, x2, xs3 not xs1, xs2, xs3.

Exercise 11. The column in the tableau for x4 should be 0,0,1, rather than 0,1,0.1

Exercise 13(g). Change 4th to 3rd

Exercise 13(k). Say that x3 has an upper bound of 1.5 and x3= 1.5. The problem is …1

Chapter 5

Exercise 3. In constructing the network, create a combined super source and super sink
node for all external flows. Call it node 0. Arcs originating at node 0 and terminating at
the plant will then represent production, while arcs leaving the outlets and terminating at
node 0 will represent sales. Appropriate capacities (upper bounds) should be placed on
all such arcs.
Also, the first three column headings in the table should be as follows.

3
 Manufacturing data

Item Capacity
Period cost ($) (units)
1 $8 175
2 10 200
3 11 150

Page 168, line 4. Should be x2* = 2.7273.

Exercise 10, page 177. Replace with the following.

You have two options for providing yourself with the use of a new Honda:

A. You can purchase a contract for $100,000 that provides you with the car for 10
years. There are no other costs associated with this option. For purposes of
analysis, assume that the same contract will be available 10 years from now, and
forever after.

B. You can purchase the Honda for $20,000 and keep it for either 2 years, 4 years
or 6 years. The annual maintenance and salvage values are listed below. Resale
values for odd years are not shown because they are not included as options.
Again it is assumed that this or similar car will be available forever.

Year Operating cost Salvage value

of ownership during year at year end

1 $500 ––
2 500 $10,000
3 1000 ––
4 2000 6000
5 3000 ––
6 5000 2000

Maintenance and salvage values occur at the end of the year. Purchase costs are at
the beginning of the year. The selection depends on your minimum acceptable rate
of return, which is 25% per year. The objective is to minimize the present worth of
the cost of ownership for an infinite time period.

a. Provide a single network diagram of the problem that includes both options A
and B (see machine replacement example in the text to help with the construction
of the network).

4
 b. Solve the problem and indicate the optimal resale time.

Exercise 14, page 178. Change last sentence to “Provide a node-arc structure that can be
used to represent this revenue function for a minimum cost network flow programming
model.”

Exercise 15. Change the last sentence to “Provide a network diagram that represents the
facility.”

Exercise 16. Add the sentence “Provide a network diagram of the problem.”

Exercise 17. Change the last sentence to “Provide a network diagram of the problem and
solve.”

Exercise 18. At the end of parts (a) and (b) in the problem statement, add the sentence
“Provide a network diagram for this part of problem.”

Exercise 19. Add the sentence “Provide a network diagram of the problem.”

Exercise 20. Change the last sentence to “Provide a network diagram for the problem and
find the optimal solution.”

Exercise 21. After the last sentence but before the Hint, add the sentence “Provide a
network diagram for the problem.”

Chapter 6
Page 210, top of the page π4 should equal 28 not 27.

Chapter 7
Page 247. In demand constraint for Days-Off Scheduling problem, the index range should
be i = 1,…,7

Exercise 1, page 255. Rephrase parts (b) and (c) as follows.

(b) Rewrite the model in part (a) as a minimization problem with all “less than or equal
to” constraints.
(c) Rewrite the model in part (b) as a minimization problem with all positive objective
function coefficients while keeping the variables nonnegative.

Exercise 11(a), page 258. In definition of xij, change to “… covers the demand for the
next j months”

5
Exercise 11(b), page 259. Change the inventory variable from zj to Ij. Also, change the
instructions to “Give an MILP formulation for the problem using the new notation.”

Exercise 21(a), page 264. Should be “… shortest path tree problem …”

Chapter 8
Page 299. In Step 3 of the algorithm, Equation 14 should be Equation 13.

Chapter 9
Page 321. Change Definition 3 to:
n
Definition 3: A function f(x) is convex on the convex set S   if and only if
f(x1 + (1–)x2) ≤ f(x1) + (1–)f(x2)

for all x1, x2  S and for all 0 <  < 1. It is strictly convex if the inequality sign ≤ is
replaced with the sign <.

Page 322. Change any to every in Definition 5

Page 326. The constraints in the set S are not separated. The set should be
2 2
S = {(x1,x2) : (0.5x1 – 0.6)x2 ≤ 1, 2x + 3x ≥ 27; x1, x2 ≥ 0}
1 2

Page 327. Change solution: If the equality constraint is removed the solution becomes
x*= (2, 2 ) with f(x*) = 0.586.

Page 335, Figure 9.13. The coordinates of the center point should be (3, 2.115).

Page 354, Exercise 16. In the equation for Ci(xi), it should be “for i = 1,…,6”

Definition 5: A set S  n is convex if every point…

Page 340, middle of page. Replace the sentences starting with the one that begins with
“The arc from …” with:
“The arc from node 2 to node 3 allows flow from station B to station E. A nonlinear cost
function of the type given in Equation (6) is associated with each station arc. For
example, the flow through the arc from node 1 to node 2 is the flow rate entering station
B, fB, where fB  B. The cost measure is the average number of units at that station,
which can be written as LB = fB/(3 – fB).

Page 348: In the table in the array describing geometric programming, the word
polynomial should posynomial both with respect to f(x) and gi(x)

6
Page 349: posinomial should be spelled posynomial

Exercise 3(a), page 351: last term should be x22

Exercise 9(f): In second summation, change c(xj) to cj(xj)

Exercise 11, page 353; rephrase: “Use the definition of convexity and induction to prove
Lemma 1.” 1

Exercise 14, page 353; should read

“Following the suggestions in Section 9.3, prove that f(x) is convex if and only if

f(x1) ≥ f(x2) + Tf(x2)(x1 – x2) for all x1, x2  S

where S is a convex set.” 1

Exercise 16. Replace with the following.

You are in charge of providing labor for a manufacturing shop for the next 6 months.
Currently, there are 20 employees in the shop. Each worker costs $1000 per month
and can manufacture five units of product during the month. The cost of hiring and
training a new worker is $500. The cost of laying off a worker is $1000. At the end
of the 6-month period you want 20 persons working in the shop. All products must
be sold in the month they are produced.
Demand in each month is a random variable uniform distributed over the
ranges specified in the table below. If demand is lower than production capacity,
some workers will be idle. If demand is higher than capacity, sales are lost with a
penalty of $200 per item not sold. To compute the expected cost of lost sales,
consider month i with minimum demand at and maximum demand bt. Say xt is the
capacity for production in month t. Then the expected cost of lost sales is


200 b  a / 2  200 a  x , for 0  x  a
  t t  t t t t


Ct  xt    200 t  1,..., 6
 2  b  a   bt  xt  , for at  xt  bt
2

 t t


0, for bt  xt

7
 Set up and solve the NLP to determine how many workers to hire or lay off
at the beginning of each month so that the expected cost of lost sales plus the cost of
providing the workforce is minimized. Workers may also be laid off at the end of
month 6. Assume that sufficient capacity is provided to meet the minimum demand.

Minimum Maximum
Month, t demand, at demand, bt
1 50 150
2 100 200
3 75 175
4 50 150
5 200 250
6 100 200

Note that the above expression for Ct(xt) is a piecewise nonlinear function and
does not lend itself to an NLP model in continuous variables. Also, it would be
better to define capacity in terms of the number of employees. This can be done by
letting 5wt = xt, where wt is the size of the workforce in month t. Next you should
derive a continuous expression for the expected cost of lost sales in terms of wt for
the case where 5wt  bt, t = 1,…,6. Because the expression that you derive will not
be valid when 5wt > bt, it will be necessary to break wt into two variables, ut and vt,
where ut is the number in the workforce who are idle (i.e., not producing because
there is no demand) and vt is the number in the workforce who are contributing to
production.

Exercise 20, page 356; Add the following at the after the values of .
Note that from Little’s law discussed in Chapter 16, the average time in the system is
equal to the average number in system / entering flow rate, where the entering flow rate
here is equal to 0.5/min. Therefore minimizing the average time in the system when the
entering flow rate is constant is equivalent to minimizing the sum of the average number
at each station.

Chapter 10
Page 403; missing parenthesis in equation. Should be

f(x)  q(x) = . . .

8
Chapter 11

Page 423, Exercise 11: Replace parts (a) and (b) with:

(a) Whenever a tire fails it is sent to the shop, repaired, and returned to the truck. The
truck continues to operate as long as it has four good tires; otherwise, it must wait
until a repaired tire is returned.

(b) When both spares fail, they are sent to the shop, repaired one at a time, and
individually returned to the truck. The truck continues to operate as long as four
good tires are available; otherwise, it must wait until a repaired tire is returned.
(Hint: the state vector should have two components.)

Chapter 12

Page 429, Computer Repair example, paragraph 2, sentence 2; change to:

“Computers that fail during the day are picked up the next morning, repaired, and then
returned to the office the following morning.”

Page 432, definition of s2 at bottom should be:

s2 = status of the second machine (working or failed)

Page 433, line 2. Add the following sentence after the first sentence:
“For s2, we assign a value of 0 if the second machine has not failed and a value of 1 if it
has.”

Page 433, line 8, change from

"One failure leads to (1,0) and two failures lead to (2,0) ..."
to
"One failure leads to (1,0) and two failures lead to (1,1) ..."

Page 433. In Table 12.3, change the following state definitions:

s1 = (1,0): One machine has failed and will be in the first day of repair today.
s2 = (2, 0): One machine has failed and will be the second day of repair today.
s3 = (1,1): Both machines have failed, one will be in the first day of repair today and the
other is waiting.
s4 = (2,1): Both machines have failed, one will be in the second day of repair today and
the other is waiting.

Page 439. In matrix equation for q after 2 transitions, should be: q(2) = …

Page 457, Exercise 1. Update the text as follows:

9
“… Assume that the order is placed just after taking inventory at the end of the week.
Cars on order arrive just before inventory is taken so at that time there is always at least
one car on the lot.”

Page 459, Exercise 12. Change problem statement as follows.

“Consider a process in which a single worker must perform five stages of a

manufacturing process, as indicated in the figure below. For purposes of analysis, divide
time into one-hour segments. When the worker is idle raw material enters the system
during the next hour with probability pA. At stage i, the probability of completing the
corresponding tasks and moving on to stage i + 1 during the current hour is pC(i), i =
1,...,5. Assume that no more than one stage can be completed in an hour.”

Page 460. Exercise 16. Hint: For part (c), you must consider two states when there are k
printers is in the shop, k = 1, 2, 3. The first is associated with the repair work that is just
starting; call it repair k_start. The second is associated with the repair work about to
finish; call it repair k_finish.

Page 463. Exercise 26 is meaningless as written. It should be replaced with the following.
1

26. Heart patients at a local hospital can be found in one of two places: the coronary care
unit or in a regular room.
a. If we assume that the number of heart patients remains constant and that the 1-
day transition probabilities are as shown, what are the steady-state probabilities
for an individual patient?

One-day transition probabilities — heart patients

Hospital Not
CCU rehabilitation hospitalize
d
Coronary care unit 0.700 0.200 0.100
(CCU) 0.050 0.800 0.150
Hospital rehabilitation 0.015 0.005 0.980
Not hospitalized

b. Assume persons leaving the hospital from the CCU actually die. For each
fatality, a new heart patient enters a competing hospital. There is a 1-day
probability of 0.05 that a patient leaves the competing hospital and enters the
CCU. How would you change the 1-day transition matrix? Compute steady-state
probabilities.

Page 464, Exercise 29. In table, letter “n” should be italic only, not bold. 1

10
Chapter 13
Exercise 10(e). The expected cost vector should be: C = (1250, 1400, 900, 0)T. 1

Exercise 12(a). Add to the problem statement of part a:

“The stock price is currently $39.” 1

Exercise 13. Labels on parts d and c are switched. 1

Chapter 14

Page 501. Second P matrix is better written as:

1   2 2 0 0 0 0 
  2.5 1   4.5 2 0 0 0 

 0  2.5 1   4.5 2 0 0 
 
 0 0 2.5 1  4.5 2 0 
 0 0 0 2.5 1  4.5 2 
 
 0 0 0 0 2.5 1  2.5

Page 503. The proportion of customers who wait; should be: 1 –  0P   5P (note error is
subscript “5”).

Page 504: First heading should be: Adding ATMs

Page 506. Paragraph before rate matrix. Should be: 1 = 1 and 2 = 2.5.
Should be: matrix element r43 = 1 + 2 = 1 + 2.5 = 3.5.

Page 510. It might be better to replace Table 14.5 with

Table 14.5 Probability of Completing a Specific Number of Units

Completed units, k 10 9 8 7 6 5 4 3
Probability, p10k(14) 0.891 0.047 0.03 0.017 0.009 0.004 0.001 0.0005

Table 14.6 (Similar change should be made.

Second row of table should be pk(14).

11
Table 14.6 Updated Probability of Completing a Specific Number of Units

Completed units, k 10 9 8 7 6 5 4 3
Probability, p10k(14) 0.681 0.104 0.084 0.06 0.037 0.02 0.009 0.0042

Page 516, Table 14.9. In summation term, index “K” on  should be lower case. 1

Page 518. Replace last sentence in last paragraph of Section Transition Probabilities
with:
If event x occurs with rate x, then the rate assigned to the transition would be
rij = xp(i, j | x)

Page 521. Exercises Section. Should say “Use the Markov Analysis Excel add-in…”

Exercise 1: Rephrase and convert to five parts:

Consider Omar’s barbershop described in the previous chapter with one chair for cutting
hair and three for waiting (four chairs in all). Assume that the mean time between
customer arrivals is 15 minutes and the mean time for a haircut is 12 minutes. Also
assume that both times are exponentially distributed and customers who arrive and find
the shop full balk. In answering the following questions, provide the rate matrix and
economics matrix in addition to the information requested.
a. If each customer pays $10 for a haircut use the steady-state probabilities to compute
Omar’s earnings in an eight-hour day.
Now make the following changes and compute Omar’s average earnings. The changes
are not cumulative.
b. Add a fourth waiting chair and recomputed Omar’s earnings.
c. Change one of the waiting chairs to a barber’s chair and add another barber. Omar
earns $5 for each hair cut the second barber does. When both barbers are idle, an
arrival goes to Omar.
d. When two or more of the waiting chairs are full Omar works faster and reduces his
average cut time to 10 minutes.
e. Change the arrival process. When all the waiting chairs are empty, all the arrivals
enter. When one waiting chair is occupied, 1/3 of the arrivals balk. When two of the
waiting chairs are occupied, 2/3 of the arrivals balk. When all three chairs are
occupied, all the arrivals balk.

Exercise 13, part a: Restate as “Draw the rate network for the number of printers in the
queue.”

12
Exercise 13, part c: Change to: “Find the steady-state probabilities. What is the
production (completion) rate of this system with and without failures? Hint: A board is
produced when no fault is detected during inspection or when a repair is completed. Use
the Markov Analysis add-in to create a matrix under the Economics button. Place a 1 in
each cell that represents a production event.”

Exercise 13: See updated solutions. The original rate matrix was not correct.

Exercise 14, parts c, d, e: The data in the tables for “Mean Time” should be shifted to the
left to line up with the column headers. 1

Exercise 15, parts a and b: The data in the tables for “Mean Time” should be shifted to
the left to line up with the column headers. 1

Chapter 16
Page 563, in table at bottom for M/M/2, change  = /s to  = /2. Also, give second
formula for L:
L = Lq + Ls = Lq + / = 0.87271

The corrections for pages 563, 572 and 573 were suggested by R.G. Vickson, University
of Waterloo.

Page 563. In the expression for Pr{Tsys > t} the term in the inner parentheses becomes
indeterminate when s  1  s   0 . To address this situation, change the sentence before
the expression for Pr{Tsys > t} to:

Without going through the details, the result for s 1  s  0 is

 ( s  ) s ð0  1e t  s  s  )  
Pr{Tsys > t} = e–t 1     , t  0 and 0 <  < 1
 s !(1 )  s1s   

(add the following)

When s  1  s   0 , a singularity occurs in the denominator of the inner term in

parentheses. To analysis this case, it is more convenient to write the above expression in
terms of  and  by substituting  = /s everywhere. Now, taking the limit of the
fraction in the inner parentheses as   (s–1) gives
 ( /  ) s 0  
Pr{Tsys > t} = e–t 1   t  , t  0 for  s  1
 s !(1   / s )  

where

13
 1
 s 1   /  n   /  
s

 0    
 n0 n! s !(1   / s ) 

in both equations for Pr{Tsys > t}. Note that when taking the limit, it is necessary to use
L’Hospital’s rule because the fraction is 0/0 at s  1  s   0 .

(It may be a good idea to derive the above expression for the case where s  1  s   0 .)

Page 572-573. The section of Finite Input Source Systems has several errors. The last
sentence of the first paragraph of the section should read.
We assume arrivals balk when n = K and K ≤ N. The results of the section also hold when
the maximum number in the system is equal to the population.

The expression for qn should be:

( N  n) n
qn = for n = 0,…,K – 1,
N  L  ( N  K ) K

Page 573. In Figure 16.10, the expression for PB should be:

K
PB =  n
ns

The expression for the average arrival rate should be

    N  L   K  N  K  for K  N

Page 578. In Table 16.9 for Case 3, should be: Lq = 5.657

Exercise 7. Change last sentence in first paragraph to: “Assume that the system is in
steady state, and in parts a – e,…”

Exercise 7. Change part (b) to:

b. When a motorist is filling up, all other customers must wait on the street. How many
spaces for cars should be made available on the station property to assure that there is
sufficient room to wait there 85% of the time?

Exercise 15. Remove second occurrence of sentence “The company has two technicians
who can … to effect a repair.” Also, change “affect” to “effect” in first occurrence. 1

Exercise 17. Change the service rate to 8 customers per hour for a better problem. 1

14
Chapter 18

Exercise 4(d). Use t = 1. 1

Exercise 10(b) and (c). Should ask for 12 replications rather than 10. 1

Exercise 15. Add the sentence: Simulate the process of passing from system 1 to system 2
with a Bernoulli random variable. 1

Exercise 21. Should refer to Table 18.19 rather than Table 18.20. Note that Table 18.19
in Chapter 18 is in error as well as Appendix A1 of the simulation chapter. 1

Table 18.19 is in error. The correct table should be: 1

Table 18.19 Error as Function of Sample Size for Inventory Simulation

Measures Demand (D) Lead time (TL)
Estimated mean 11.65 2.8
Estimated standard deviation 2.594 0.980
Estimated size (n) for 1% error 3290 8127
Sample size (n) for 5% error 77 188
Sample size (n) for 10% error 14 33

Supplement

The discussion concerning confidence limits in the Simulation supplement distributed on

the original edition student disk is in error. It is corrected on the Teach ORMM CD and
on the web. The correct confidence limit discussion is below.

Confidence Intervals
Once x and the standard error of the mean,  X , are determined, the confidence interval
and maximum error for X are given by
x  x  z 2 x (A.6)
or
x  x  t 2ˆ x (A.7)

as the case may be, where ˆ X is the estimated standard error when (A.5) is used in place
of  X2 in (A.3) or (A.4). For A.7, the value of t 2 depends on the number of degrees of
freedom, df, where df = n – 1. If X is known, then the maximum error  for a given level
of confidence can be found from

15
   z 2 x or z 2 x / n
when (A.3) applies. It follows that

(A.8)

provides the required sample size which satisfies a given maximum error and confidence
level.

Error in Probability Supplement

Page 24
The c.d.f. of the Triangular Distribution

0, for x  0
 2
 x , for 0  x  c
c
F(x) =  .
 x (2  x )  c
, for c  x  1
 (1  c)

1, for x  1

16