SUPPLEMENT TO

CHAPTER 7
Learning Curves
SUPPLEMENT OUTLINE

The Concept of Learning Curves, 2

Determining Unit Times, 5
Determining the Learning
Percentage, 7
Applications of Learning
Curves, 9
Cautions and Criticisms, 10
Summary, 10
Solved Problems, 11

Learning Objectives

Discussion and Review

Questions, 11
Internet Exercise, 12
Problems, 12
Mini-case: Product Recall, 19
Mini-case: Learning Curve in
Surgery, 20
Case: Renovating the Lions Gate
Bridge, 20

After completing this supplement,

you should be able to:
Explain the concept of
learning curves.
LO2 Make time estimates based on
a learning curve.
LO3 Determine the learning
percentage.
LO4 List and briefly describe some
of the main applications of
learning curves. Also,
determine the minimum
number of repetitions to
achieve a given standard.
LO5 Outline some of the cautions
and criticisms of learning
curves.
LO1

PART THREE System Design

Learning usually occurs when humans are involved; this is a basic consideration in the
design of work systems. It is important to be able to predict how learning will affect task
times and costs. This supplement addresses those issues.

THE CONCEPT OF LEARNING CURVES

LO1

Human performance of activities typically shows improvement when the activities are
done on a repetitive basis: The time required to perform a task decreases with increasing
repetitions. Learning curves display this phenomenon. The degree of improvement and
the number of tasks needed to realize the major portion of the improvement is a function
of the task being done. If the task is short and somewhat routine, only a modest amount
of improvement is likely to occur, and it generally occurs during the first few repetitions.
If the task is fairly complex and has a longer duration, improvements will occur over a
larger number of repetitions. Therefore, learning curves have little relevance for planning
or scheduling of routine short activities, but they do have relevance for complex long repetitive activities.
Figure 7S-1 illustrates the basic relationship between units produced and time per unit:
time per unit decreases as the number of units produced increases. In fact, the most common learning curve model, which is studied here, assumes that the rate of decrease in unit
time remains constant as the number of units produced doubles (this is illustrated on the
next page).
The general relationship is alternatively referred to as an improvement curve. When
unit cost is used instead of unit time, the relationship is usually referred to as experience curve or progress function. Experts agree that the learning effect is the result of
other factors in addition to actual worker learning (which results in increased dexterity,
reduced rework, etc). Some of the improvement can be traced to preproduction factors,
such as selection of tooling and equipment, product design and test, personnel training,
and, in general, the amount of effort expended prior to the start of the work. Other contributing factors may involve changes after production has begun, such as changes in
methods, layout, support services, tooling, design, and lot size increases. In addition,
management input can be an important factor through improvements in planning,
scheduling, motivation, and control.
Major changes that are made once production is under way, such as product redesign
or new equipment, can cause a temporary increase in time per unit until workers adjust to
the change. If a number of major changes are made during production, the learning curve
would be more realistically described by a series of scallops instead of a smooth curve, as
illustrated in Figure 7S-2. Nonetheless, it is convenient to work with a smooth curve,
which can be interpreted as the average effect.
From an organizational standpoint, what makes the learning effect more than an interesting curiosity is its predictability, which becomes readily apparent if the relationship is

The learning effect: time per

unit decreases as the number
of units produced increases

Time per unit

Figure 7S-1

Number of units

SUPPLEMENT TO CHAPTER 7 Learning Curves

Time per unit

Figure 7S-2

Improvements may create a

scallop effect in the learning
curve
Average

No. of units
Improvements

plotted on a log-log scale (see Figure 7S-3). The straight line that results reflects a constant
learning percentage, which is the basis of learning curve estimates: Every doubling of
units produced results in a constant percentage decrease in the time per unit. Typical decreases range from 10 percent to 20 percent. By convention, learning curves are referred to
in terms of the complements of their decrease rates. For example, an 80-percent learning
curve denotes a 20-percent decrease in unit time with each doubling of units produced and
a 90-percent curve denotes a 10-percent decrease rate. The 80 or 90 percent learning percentage in the above examples is referred to as slope of the learning curve. Note that a
100-percent curve would imply no decrease in unit time at all (i.e., no learning).
For some examples of learning curves in industry see Figure 7S-4, and for slopes of
typical industrial activities see Figure 7S-5.1 Note the log-log scales and the fact that the
unit time curves are approximately linear.
100
80

Figure 7S-3

60
50
40

On a log-log graph, learning

curves are straight lines

90%

30

Time per unit

20
80%
10
8
6
5
4
3
2

4 5 6

8 10

20

30 40

60 80100

Number of units produced

Both figures are from E. B. Cochran, Planning Production Costs: Using the Improvement Curve, San Francisco:
Chandler Publishing co, 1968.

PART THREE System Design

Figure 7S-4

IMPROVEMENT
CURVE
EXPERIENCE

W.W.II
DESTROYER (18)

2,000

1,000

1,000

(18)

W.W.II
TRANSPORT
SHIP (17)

(17)

500

200

500

200

W.W.II
BOMBER (16)
MAJOR

100

AIRCRAFT COMPONENT

100

D (13)

(16)

50

50

AIRCRAFT
ASSEMBLY (14)

D (13)

20
HOURS PER UNIT (000S)

Some examples of learning

curves in industry

(14)

B (10)

10

10

MAJOR

20

B (10)

AIRCRAFT COMPONENT
ELECTRONIC

(14)

GROUND CHECKOUT (7)

D (13)

A (4)
AIRBORNE COUNTER
MEASURES (9)

(9)

(5)
(7)

1.0

1.0
(5)

ELECTRO-MECHANICAL DEVICE (5)

B (10)

SMALL TRUCK BODY (3)

(9)

0.5

0.5

(7)
(4)

B (2)

(5)

(3)

0.2

(4)

MISSILE
GUIDANCE

0.2

B (2)

0.1

0.1
A (1)
(3)

.05

.05

(3)

.02

.02

A (1)

.01

10

20

50

UNITS PRODUCED

100

200

500

1,000

.01

SUPPLEMENT TO CHAPTER 7 Learning Curves

TYPICAL BASIC SLOPES

HOURS / UNIT

2000

Figure 7S-5

JOB MACHINE
SHOP
SHEET M
ETAL STA
MPING
WIR

Slopes of typical industrial

activities

E PREPAR
JOB M ATION
ACHIN
ING - LA
ELECTR
RGE PA
ONIC
RTS
GEN S SHIP AS
MAJ
ERAL
S
OR
SUBA EMBLY
AIRC
SSEM
RAF
T
B
LY
A
SSEM
ELECTRONICS
BLY

1500

SHEET METAL

1000

95
92
90
88
87
85
83
80

750

500
100

150

200

300

500

750

1,000

UNITS PRODUCED

An activity is known to have an 80-percent learning curve. It has taken a worker 10 hours
to produce the first unit. Determine expected completion times for these units: the 2nd,
4th, 8th, and 16th (note successive doubling of units).
Each time the cumulative output doubles, the time per unit for that quantity should be
approximately equal to the previous time multiplied by the learning percentage (80 percent
in this case). Thus:
Unit

Unit Time (hours)

=10

0.8(10)

=8

0.8(8)

=6.4

0.8(6.4)

=5.12

16

0.8(5.12)

=4.096

Example S-1 illustrates an important point and also raises an interesting question. The
point is that the time reduction per unit becomes smaller and smaller as the number of
units produced increases. For example, the second unit required two hours less time than
the first, but the decrease from the 8th to the 16th unit was only slightly more than one
hour. The question raised is: How are times calculated for units such as three, five, six,
seven, and other units that dont fall into the doubling pattern?

Determining Unit Times

LO2

There are two ways to obtain the unit times. One is to use a formula; the other is to use a
table of values.
First, consider the formula approach. The formula is based on the existence of a linear
relationship between the time per unit and the number of units produced when these two
variables are expressed in logarithms.
The unit time (i.e., the number of direct labour hours required) for the nth unit can be
calculated using the formula
Tn=T1nb

(7S-1)

Example S-1

Solution

PART THREE System Design

where
Tn=Time for the nth unit
T1=Time for the first unit
learning percentage
ln

100
; ln stands for the natural logarithm
b=
ln(2)
To use the formula, you need to know the time for the first unit T1 and the learning
percentage. For example, for an 80-percent learning curve with T1=10 hours, the time
for the third unit would be calculated as
T3=10(31n.8/1n 2)=10(3.223/.693)=10(3.322)=7.02
The second approach is to use a learning factor from Table 7S-1. The table shows
two things for some selected learning percentages. One is the unit time factor for the
number of units in the sequence (unit number). This enables us to easily determine how
long any unit in the sequence should take to produce. The other is the total (cumulative)
time factor. The calculation for both times is a relatively simple operation: Multiply the
learning factor by the time required for the first unit.
To find the time for unit n (e.g., n=10 the 10th unit), use the formula
Tn=T1unit time factor

(7S-2)

Thus, for an 85-percent learning curve with T1=4 hours, the time for the 10th unit
would be 4.583=2.33 hours. To find the time for all units up to a specified unit n (e.g.,
n=10 the first 10 units), use the formula
Tn=T1total time factor

(7S-3)

Thus, for an 85-percent curve with T1=4 hours, the total time for all first 10 units
(including the time for unit 1) would be 47.116=28.464 hours.
Example S-2

Solution

An airplane manufacturer is negotiating a contract for the production of 20 small jets. The
initial jet required 400 days of direct labour. The learning percentage is 80 percent. Estimate the expected number of days of direct labour for:
a. The 20th jet.
b. All 20 jets.
c. The average time for 20 jets.
Using Table 7S-1 with n=20 and an 80-percent learning percentage, you find these
factors: Unit time=.381. Total time=10.485.
a. Expected time for 20th jet: 400(.381)=152.4 days.
b. Expected total time for all 20 jets: 400(10.485)=4,194 days.
c. Average time for 20 jets: 4,19420=209.7 days.
Use of Table 7S-1 requires a time for the first unit. If for some reason the completion
time of the first unit is not available, or if the manager believes that the completion time
for some later unit is more accurate, the table can also be used to obtain an estimate of the
time for the first unit.

Example S-3

Solution

The manager in Example S-2 believes that some unusual problems were encountered in
producing the first and second jets and would like to revise the time for the first unit
based on the completion time of 276 days for the third jet.
The learning factor for n=3 and an 80-percent curve is .702 (Table 7S-1). Divide the
actual time for unit 3 by the learning factor to obtain the revised estimate for the time of
first unit: 276 days.702=393.2 days.

SUPPLEMENT TO CHAPTER 7 Learning Curves

70%

75%

80%
Total
Time

85%

Unit
Number

Unit
Time

Total
Time

Unit
Time

Total
Time

Unit
Time

Unit
Time

Total
Time

1.000

1.000

1.000

1.000

1.000

1.000

1.000

1.000

1.000

1.000

.700

1.700

.750

1.750

.800

1.800

.568

2.268

.634

2.384

.702

2.502

.850

1.850

.900

1.900

.773

2.623

.846

2.746

.490

2.758

.562

2.946

.640

3.142

.437

3.195

.513

3.459

.596

3.738

.723

3.345

.810

3.556

.686

4.031

.783

4.339

.398

3.593

.475

3.934

.562

4.299

.657

4.688

.762

5.101

Unit
Time

Total
Time

90%

.367

3.960

.446

4.380

.534

4.834

.634

5.322

.744

5.845

.343

4.303

.422

4.802

.512

5.346

.614

5.936

.729

6.574

.323

4.626

.402

5.204

.493

5.839

.597

6.533

.716

7.290

10

.306

4.932

.385

5.589

.477

6.315

.583

7.116

.705

7.994

11

.291

5.223

.370

5.958

.462

6.777

.570

7.686

.695

8.689

12

.278

5.501

.357

6.315

.449

7.227

.558

8.244

.685

9.374

13

.267

5.769

.345

6.660

.438

7.665

.548

8.792

.677 10.052

14

.257

6.026

.334

6.994

.428

8.092

.539

9.331

.670 10.721

15

.248

6.274

.325

7.319

.418

8.511

.530

9.861

.663 11.384

16

.240

6.514

.316

7.635

.410

8.920

.522 10.383

.656 12.040

17

.233

6.747

.309

7.944

.402

9.322

.515 10.898

.650 12.690

18

.226

6.973

.301

8.245

.394

9.716

.508 11.405

.644 13.334

19

.220

7.192

.295

8.540

.388 10.104

.501 11.907

.639 13.974

20

.214

7.407

.288

8.828

.381 10.485

.495 12.402

.634 14.608

21

.209

7.615

.283

9.111

.375 10.860

.490 12.892

.630 15.237

22

.204

7.819

.277

9.388

.370 11.230

.484 13.376

.625 15.862

23

.199

8.018

.272

9.660

.364 11.594

.479 13.856

.621 16.483

24

.195

8.213

.267

9.928

.359 11.954

.475 14.331

.617 17.100

25

.191

8.404

.263 10.191

.355 12.309

.470 14.801

.613 17.713

26

.187

8.591

.259 10.449

.350 12.659

.466 15.267

.609 18.323

27

.183

8.774

.255 10.704

.346 13.005

.462 15.728

.606 18.929

28

.180

8.954

.251 10.955

.342 13.347

.458 16.186

.603 19.531

29

.177

9.131

.247 11.202

.338 13.685

.454 16.640

.599 20.131

30

.174

9.305

.244 11.446

.335 14.020

.450 17.091

.596 20.727

DETERMINING THE LEARNING PERCENTAGE

LO3

If the learning percentage of the activity cannot be estimated based on similar previous
activities or the industry learning slope, given a few observations of unit times one can
estimate the learning percentage of the activity by fitting the power function of 7S-1 to
the chart of the data. For simplicity, we will change notation to more familiar symbols:
y=Tn, a=T1, and x=n. The equation for the power function is y=axb. The fitted power
function will provide the value for a=time of the first unit, and b=ln (learning
percentage/100) / ln 2. Solving the b equation gives:
Learning percentage=1002b

(7S-4)

The power function can easily be obtained from Excel by charting the data (enter the
data, then click Insert, and finally click Line) and using Trendline under the Layout
menu.

Table 7S-1

Learning factors

PART THREE System Design

Example S-4

The cost per patient for heart transplant in Temple University Hospital in Philadelphia2
for the first 17 patients is displayed below.
Heart Transplant Patient

Cost (in $1,000)

133

97

250

120

115

125

98

94

57

10

201

11

52

12

86

13

93

14

89

15

70

16

64

17

75

Chart the line plot of the data, fit the power function to the chart, and estimate the learning percentage.
Solution

Note that in this case, the data given are unit costs instead of unit times, and cost is
directly related to the times of surgeons and nurses. After charting the data in Excel as a
line plot, the Trendline command under Layout menu was used to fit the power function to the data.
300

y = 168.19x0.2732

250
200
150
100
50
0

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

Substituting b=0.2732 into 1002b, the learning percentage=82.75 percent.

Adapted from D. B. Smith and J. L. Larsson, The Impact of Learning on Cost: The Case of Heart Transplantation, Hospital & Health Services Administration 34(1), Spring 1989, pp. 8597.

SUPPLEMENT TO CHAPTER 7 Learning Curves

Figure 7S-6

Worker learning curves can

help guide personnel decisions
Time/cycle

A (underqualified)
B (average)

Standard time

C (overqualified)

One week

Training time

APPLICATIONS OF LEARNING CURVES

LO4

Learning curves have useful application in a number of management activities, including:

1. Labour planning and scheduling
2. Negotiated selling/purchasing
3. Assessing labour training needs and performance
Knowledge of output projections in learning situations can help managers make better
decisions about how many workers they will need. Of course, managers recognize that
improvement will occur; what the learning curve contributes is a method for quantifying
expected future improvements.
Negotiated selling/purchasing often involves contracting for specialized items that
may have a high degree of complexity. Examples include aircraft, ships, and specialpurpose equipment. The direct labour cost per unit of such items can be expected to
decrease as more units are produced. Hence, negotiators negotiate cost/price on that
basis. For contracts that are terminated before delivery of all units, suppliers can use
learning curve data to argue for an increase in the unit price for the smaller number of
units. Conversely, the customer can use that information to negotiate a lower price per
unit on follow-on orders on the basis of projected additional learning gains.

Determining the minimum number of repetitions to achieve

a given standard
Learning curves can be used to determine the length of training for new workers doing
complex-long-cycle jobs. The progress of each worker can be evaluated by measuring each
workers performance, graphing the results, and comparing the learning to an expected rate
of learning. The comparison will reveal if a worker is underqualified, average, or
overqualified for a given type of work (see Figure 7S-6). Also, learning curves can be used
to determine the minimum number of repetitions to achieve a given standard.
Use learning curves to predict the number of units that a trainee needs to produce to
achieve a unit time of 6 minutes if the trainee took 10 minutes to produce the first unit
and learning curve of 90 percent is expected. Use both
a. Formula 7S-2 and Table 7S-1 in reverse.
b. Logarithm version of Formula 7S-1:
ln(Tn)=ln(T1) + b ln(n)

(7S-5)

Rewriting (where e=2.718282):

n=e[ln(T ) - ln(T )]/b
n

(7S-6)

Example S-5

10

Solution

PART THREE System Design

a. Formula 7S-2:
Tn=T1unit time factor
Setting Tn equal to the specified time of 6 minutes, T1 to 10 minutes, and solving for
the unit time factor:
6min=10minunit time factor unit time factor=6min10min=.600.
From Table 7S-1, under 90 percent in the Unit Time column, we find .599 at 29 units
is the closest to .600. Hence, approximately 29 units will be required to achieve the
specified time.
b. Using the Formula 7S-6:
(1)Calculate b, b=ln(learning percentage/100)ln(2)=ln(.90)ln(2)=
0.10540.6931=0.152.
(2) n=e [ln(Tn) ln(T1)]/b=e [ln(6) ln(10)]/.152=e [1.7922.303)]/.152=e3.36=28.8
Round to 29. Hence, 29 units will be needed to achieve a time of 6 minutes.

CAUTIONS AND CRITICISMS

LO5

Managers/staff using learning curves should be aware of their limitations and pitfalls.
This section briefly outlines some of the major cautions and criticisms of learning
curves.
1. Learning percentage may differ from organization to organization and by type of work.
Therefore, it is best to base learning percentage on empirical studies rather than
assumed percentage.
2. Projections based on learning curves should be regarded as approximations of actual
times and treated accordingly.
3. If time estimates are based on the time for the first unit, considerable care should be
taken to ensure that this time is valid. The first unit time (or even several units after
that) may not be accurate due to time compression, design changes, equipment problems, etc. It may be desirable to revise the first unit time as later times become available. Since it is often necessary to estimate the time for the first unit prior to production,
this caution is very important.
4. Learning curves do not apply to mass production (which have short cycle times)
because the decrease in time per unit is imperceptible.
5. Users of learning curves sometimes fail to include carryover effects; previous experience with similar activities can reduce unit times. In this case, instead of Formula
7S-1, the following model could be used: Tn=T1(n+np)b where np is the number of
units produced previously.
Summary

Unit time tends to decrease at a constant rate as number of units produced doubles. This effect
is called learning curve and is stronger for complex new products. Learning percentage (or
slope) is defined as (100percentage reduction in unit time as units double). For example, 90
percent learning percentage means that production time decreases by 10 percent if number of
units produced is doubled. Unit time can be calculated using Formula 7S-1 or Table 7S-1.
Learning percentages for various industries/activities are available, but it is better to determine
it empirically by fitting a power function to the plot of unit time drawn against the number of
units produced.

SUPPLEMENT TO CHAPTER 7 Learning Curves

11

Solved Problems
Problem1

An assembly operation has a 90-percent learning curve. The line has just begun work on a new
item. The initial unit required 28 hours. Estimate the time that will be needed to complete:
a. The first five units.
b. Units 20 through 25.

Solution

Use the total time factor in the 90-percent column of Table 7S-1.
a. Total time factor: 4.339.
Estimated time for the first five units: T1Total time factor=28(4.339)=121.492 hours.
b. The total time for units 20 through 25 can be determined by subtraction, using total time
factor for 90% learning of Table 7S-1.
Hours
Total time for 25 units:

28(17.713)=495.964

Total time for 19 units:

28(13.974)=391.272

Total time for 20 through 25

104.692 hours

Problem2

A manager wants to determine an appropriate learning percentage for a new type of

work his company will undertake. He has obtained unit times for the initial six repetitions
(see below).
i. Estimate the value of b in Formula 7S-1 using the power function of Excel.
ii. What learning rate is appropriate?
Unit

Completion Time (hours)

15.9

12.0

10.1

9.1

8.4

7.5

Solution
(i)

i. b=-.4094

Unit
1
2
3
4
5
6

Completion
Time (hrs)
15.9
12.0
10.1
9.1
8.4
7.5

y = 15.928x .4094

20
15
10
5
0

ii. The learning percentage=100(2.4094)=75.3%

1. What is a learning curve and why is it important? (LO1)

2. Under what circumstances is most learning possible? (LO1 & 5)
3. What would a learning of 80 percent imply? (LO1)
4. List the factors that create the learning effect. (LO1)
5. Give an example of learning effect from your own education. Note: It must be a new complex
task that had to be repeated many times, and you got better and faster in it, with an initial
sharp drop in the unit time. (LO1)
6. Why should the learning percentage be higher (i.e., less learning) for a machine shop or an
automated process than a manual assembly line? (LO1)
7. Compare using Formula 7S-1 and Table 7S-1. (LO2)

Discussion and
Review Questions

12

PART THREE System Design

8. Explain how, instead of using the power function, one can use linear regression to estimate the
learning percentage. (LO3 & 4)
9. In what management activity are learning curves most useful? (LO4)
10. Users of learning curves sometimes fail to include carryover effect. What is meant by this?
(LO5)
11. If the learning phenomenon applies to all human activities, why isnt the effect noticeable in
short-cycle mass production? (LO5)

Internet Exercise

Problems

Verify one of the values in Table 7S-1 using the learning curve calculator in http://cost.jsc.nasa.gov/
learn.html (scroll down). Choose Crawfords method. Note: Crawfords method measures learning
by the unit time whereas Wrights method measures learning by the average time. (LO2)

1. An aircraft services company has an order to refurbish the electronics of 18 planes. The work
has a learning percentage of 80 percent. The first plane has required 300 hours to refurbish.
Estimate the time needed to complete: (LO2)
a. The fifth plane.
b. The first five planes.
c. All 18 planes.
2. Estimate the time it will take to complete the 4th unit of a multi-unit job involving a large
assembly if the initial unit required 80 hours and the learning percentage is: (LO2)
a. 72 percent.
b. 87 percent.
c. 95 percent.
3. A contractor intends to bid on installing 30 in-ground swimming pools. Because this will be a
new line of work for the contractor, he believes there will be a learning effect. After reviewing
time records from a similar type of activity, the contractor is convinced that an 85-percent
learning curve is appropriate. He estimates that the first pool will take his crew eight days to
install. How many days should the contractor schedule for: (LO2)
a. The first 10 pools?
b. The second 10 pools?
c. The final 10 pools?
4. A job is known to have a learning percentage of 82 percent. If the first unit took 20 hours,
estimate the times for the third and fourth units. (LO2)
5. A manager wants to determine an appropriate learning percentage for a certain activity. Times of
the first six units are given below: (LO3)
Units

Time (minutes)

1..........

46

2..........

39

3..........

35

4..........

33

5..........

32

6..........

30

a. Determine the learning percentage using Excel.

b. Using your answer from part a, estimate the average time if a total of 30 units are
planned.
6. Students in an operations management class have been assigned four similar homework problems. One student took 50 minutes to complete the first problem. Assume that a 70-percent

SUPPLEMENT TO CHAPTER 7 Learning Curves

learning curve is appropriate. How much time can this student plan to spend solving all the
remaining problems? (LO2)
7. A subcontractor is responsible for outfitting six ships with new electronics. Four of the six
ships have been completed in a total of 600 hours. If the task has 75-percent learning curve,
how long should it take to finish the last two units? (LO2)
8. The 5th unit of a 25-unit job took 14.5 hours to complete. If a 90-percent learning curve is
appropriate: (LO2)
a. How long should it take to complete the last unit?
b. How long should it take to complete the 10th unit?
c. Estimate the average time for the 25 units.
9. A lot of 20 units is to be produced. Labour cost is $8.50 per hour. Setup cost is $50 and material
cost is $20 per unit. The learning percentage is expected to be 90 percent. Overhead is charged
at the rate of 50 percent of total labour, material, and setup cost. Determine the average unit
cost for the lot, given that the first unit took 5 hours to complete. (LO2)
10. A firm has a training program for a certain operation. The progress of trainees is carefully
monitored. An established standard requires a trainee to be able to complete the sixth repetition of the operation in six hours or less. Those who are unable to do this are assigned to other
jobs. Currently, three trainees have each completed two repetitions. Trainee A had times of 9
hours for the first and 8 hours for the second repetition; trainee B had times of 10 hours and
8 hours for the first and second repetitions; and trainee C had times of 12 and 9 hours. Which
trainee(s) do you think will make the standard? Explain your reasoning. (LO2)
11. The first unit of a job took 40 hours to complete. The work has a learning percentage of
88percent. Determine time estimates for units 2, 3, 4, and 5. (LO2)
12. Estimate the remaining time that will be needed to complete a five-unit job. The initial unit required
12 hours, and the work has a learning percentage of 77 percent. (LO2)
13. A job is supposed to have a learning percentage of 82 percent. Times for the first four units were
30.5, 28.4, 27.2, and 27.0 minutes. Does a learning percentage of 82 percent seem reasonable?
Fit a power function to the line plot of the data in Excel. (LO3)
14. The 5th unit of a 10-unit job took five hours to complete. The 6th unit has been worked on
for two hours, but is not yet finished. Estimate the additional amount of time needed to finish
the 10-unit job if the work has a 75-percent learning percentage. (LO2)
15. Estimate the number of repetitions each of the workers listed below will require to reach a
time of 7 hours per unit. Times of the first two repetitions (in hours) are given below. (LO4)
Trainee

T1

T2

Art

11

9.9

Sherry

10.5

8.4

Dave

12

10.2

16. Estimate the number of repetitions that a new worker will require to achieve the standard if
the standard is 18 minutes per unit. She took 30 minutes to do the initial unit and 25minutes
to do the next unit. (LO4)
17. Estimate the number of repetitions each of the workers listed below will require to achieve a
standard time of 25 minutes per unit. Times of the first two repetitions (in minutes) are given
below. (LO4)
Trainee

T1

T2

Tracy

36

31

Darren

40

36

Lynn

37

30

13

14

PART THREE System Design

18. A research analyst performs database searches for clients. According to her, a new search
requires approximately 55 minutes. Repeated requests on the same or similar topic take less
and less time, as shown below: (LO4)
Request no.

Time (min.)

55.0

41.0

35.2

31.0

28.7

26.1

24.8

23.5

How many more searches will it take until the search time gets down to 19 minutes?
19. The following data are the hours it took to assemble 20 identical sections of an aircraft fuselage.3
Use Excel to chart the line plot of the data, fit a power function to the chart, and determine
the learning percentage. (LO3)
Units Produced

Actual Unit Time

2,122

1,512

1,283

848

755

798

697

825

759

10

798

11

788

12

771

13

774

14

770

15

778

16

786

17

777

18

785

19

781

20

764

20. The following data are the cumulative number of Model T cars produced by Ford and their unit
manufacturing cost for the period 1910 to 1926.4 Chart the line plot of the data, fit a power
function to the chart, and determine the learning percentage, assuming that the cost of material
remained the same over these years. (LO3)
Fords Model T
Year

Cumulative Number of
Cars Produced (1,000s)

Manufacturing Cost
per Car ($)

1910

50

2,578

1911

120

2,243

1912

290

1,765

1913

493

1,431

3
C. J. Waterworth, Relearning the Learning Curve: A Review of the Derivation and Applications of LearningCurve Theory, Project Management Journal 31(1), March 2000, pp. 2431.
4
L. E. Yelle, Adding Life Cycles to Learning Curves, Long Range Planning 16(6), 1983, pp. 8287.

SUPPLEMENT TO CHAPTER 7 Learning Curves

(continued)
Year

Cumulative Number of
Cars Produced
(1,000s)

Manufacturing Cost
per Car ($)

1914

801

1,458

1915

1,302

1,602

1916

2,037

1,258

1917

2,701

1,273

1918

3,199

1,059

1919

4,140

941

1920

4,603

862

1921

5,574

831

1922

6,881

769

1923

8,900

804

1924

10,829

709

1925

12,749

661

1926

14,312

641

21. In a study of forgetting and relearning, 31 subjects individually built an Erector Set, dismantled
it, and rebuilt it again several times, then went away and returned after a few months, and
repeated the above process.5 For one subject, the following 7 building times (in minutes) were
observed (in order): 33, 25, 22, 20, 19, 16, and 15. After a 101-day break, the subject had the
following 8 building times (in order): 21, 17, 20, 15, 17, 15, 15, and 13. Fit a power function
to each set of data, and calculate the learning percentage and relearning percentage for the
subject. Explain your results. (LO3)
22. The following data are the man-hours it took to build the concrete frame of each floor of
four buildings in Porto, Portugal.6 Chart the line plot of each set of data, fit a power function to each chart, and calculate the learning percentage for each building. Interpret your
results. (LO3)
Floor Level
Building

SGL 3A

11

SGL 3B

12

11

11

11

SGL 2A

11

10

SGL 2B

12

11

10

10

10

11

12

13

*23. A study was performed to determine the efficacy of laser to extract leads of pacemakers that
attach themselves to heart and veins of patients after a few years.7 Patients were divided in 3
groups of 25 each. Average surgery times fell from 19 minutes for the first 25, to 11 minutes
for the next 25, and to 8 minutes for the final 25, without any major complications. Using
midpoints of each group and the average times, chart a scatter plot of the three points, fit a
power function to the chart, and calculate the learning percentage. Is there learning? (LO3)
5

C. D. Bailey, Forgetting and the Learning Curve: A Laboratory Study, Management Science, 35(3), March
1989, pp. 34352.
6
J. P. Couto and J. C. Teixeira, Using Linear Model for Learning Curve Effect on Highrise Floor Construction,
Construction Management and Economics, 23(4), 2005, pp. 355364.
7
N. Ghosh et al, Laser Lead Extraction: Is There a Learning Curve?, Pacing and Clinical Electrophysiology,
March 2005, pp. 180184.

15

PART THREE System Design

9,000
8,000
Assembly time (hours)

16

7,000
6,000
5,000
4,000
3,000
2,000
1,000
0

30

60

90

120

150

180

210

240

270

Cumulative production (Collared Coil units)

*24. The manufacture of collared coils (see photo below) for CERN Particle Accelerator in Geneva,
Switzerland, was performed by three suppliers.8 The man-hours used by the third supplier for
each unit are displayed in the chart above by a circle. The dashed line is the learning curve fitted.
Check 2 points on the curve to verify that the learning percentage is 74 percent. (LO2 & 3)

25. A study was done to show that using a new equipment (Nucletrons FIRST system) will
reduce the learning time for implanting radioactive seeds for treating prostate cancer.9 For
an explanation of the procedure, see e.g., http://www.prostate-cancer.com/brachytherapy/
cancer-treatments/treatment-brachy-ldr.html. Two institutions were used for the study: Tom
Baker Cancer Center (T.B.) and Center Hospitalier Universitaire de Quebec (CHUQ). T.B.
had no previous experience in this procedure but CHUQ had treated 740 patients before.
The results of the study were similar in terms of outcome, but CHUQ had faster procedure
times (in minutes; see below). Chart the line plot of each set of data, fit a power function to
each chart, and calculate the learning percentage for each institution. Explain the differences
between the two results. (LO3)
Case Number

Time T.B.

Time CHUQ

260

130

340

120

250

80

100

100

8
P. Fessia et al, Application of the Learning Curve Analysis to the LHC Main Dipole Production: First
Assessment, IEEE Transactions on Applied Superconductivity, 16(2), June 2006, pp. 242247.
9
L. Beaulieu et al, Bypassing the Learning Curve in Permanent Seed Implants using State-of-the-art Techno
logy, International Journal of Radiation Oncology * Biology * Physics 67(1), 2007, pp. 7177.

SUPPLEMENT TO CHAPTER 7 Learning Curves

(Continued)
Case Number

Time T.B.

Time CHUQ

220

70

180

90

220

80

200

90

160

100

10

170

130

11

200

90

12

210

70

13

200

75

14

130

90

15

230

80

16

190

85

17

140

70

18

110

80

19

130

75

20

140

80

26. A study was performed to determine the minimum number of tracheal intubation necessary for
an intern to achieve expert status.10 The objective was an intubation time of two minutes or
less without any complications. Four residents with no prior experience intubated 15 patients
each under supervision. A learning curve was generated of the mean time (in seconds) for
intubation of patients 1 to 15 for all residents combined (see below). Chart the line plot of the
data, fit a power function to the chart, and determine how many repetitions are necessary for
a resident to attain the expert status. (LO3 & 4)
Repetition Number

Average
Procedure Time

240

265

199

143

117

117

152

107

139

10

95

11

98

12

70

13

79

14

82

15

78

27*. An airplane component manufacturer has been approached by its customer (the airplane
manufacturer) to make a specific component. Based on similar components produced before,
the cost estimator has determined that the breakeven (standard) unit will be the 24th unit
10
C. Johnson and J. T. Roberts, Clinical Competence in the Performance of Fiberoptic Laryngoscopy and
Endotracheal Intubation: A Study of Resident Instruction, Journal of Clinical Anesthesia 1(5), 1989,
pp.344349.

17

18

PART THREE System Design

made and the standard time for the 24th unit will be 161 hours. Also, she expects a learning
percentage of 80 percent. The order is for 30 units. (LO2)
a. If a quotation is to be given to the customer for the total expected labour hours, what should
it be?
b. The quotation was accepted by the customer and the component manufacturer has started
making the component. Although the first few units took longer than expected, per-unit
labour hours are now approaching the estimates. However, the customer has asked for a
design change to be implemented after the 15th unit. The design change is estimated to
result in 30 hours of new work for the standard (24th) unit, and is expected to have the
same learning percentage (80 percent). The new work replaces the same amount of original
work (30 hours) for the standard (24th) unit. What should the revised total labour hours (for
all 30 units) be? Hint: The new and the original work behave differently at the mid-point
of production of this batch because the original work has already been learned to a large
degree whereas the new work is yet to be learned.
28. A study was performed on the new tube hydroforming line of a GM component plant to determine the future cost prospects of this new technology.11 One of the characteristics measured
was the cycle time to make one tube. The monthly averaged cycle time (in minutes) over a
period of 53 months of production (approximately 42,000 units per month) are displayed below.
Chart the line plot of the data, fit a power function to the chart, and calculate the learning
percentage. Why is the learning effect small in this case? (LO3)
Month

11

Cycle Time

Month

Cycle Time

1.09

28

0.83

1.1

29

0.79

1.36

30

0.7

1.16

31

0.75

1.13

32

0.79

0.9

33

0.82

34

0.81

0.93

35

0.85

0.79

36

1.03

10

0.74

37

0.79

11

0.91

38

0.68

12

0.8

39

0.71

13

0.77

40

0.74

14

0.71

41

0.69

15

0.75

42

0.78

16

0.71

43

0.77

17

0.81

44

0.81

18

0.91

45

0.79

19

0.8

46

0.77

20

0.89

47

0.78

21

0.78

48

0.91

22

0.91

49

0.82

23

0.8

50

0.83

24

0.81

51

0.88

25

1.12

52

0.84

26

1.03

53

0.65

27

0.73

M-C Nadeau et al, A Dynamic Process-based Cost Modeling Approach to Understand Learning Effects in
Manufacturing, International Journal of Production Economics 128, 2010, pp. 223234.

SUPPLEMENT TO CHAPTER 7 Learning Curves

19

*29. An airplane component manufacturer has been asked to quote for 30 units of a new component.
The following data are available: two departments are involvedfabrication with 90percent
learning slope and assembly with 85 percent learning slope; the breakeven (standard) unit for
both departments is the 20th unit; and the direct labour of the 20th unit is estimated to be 100
hours for fabrication and 75 hours for assembly. What should the quoted total labour hours
be? (LO2)
*30. A machine shop processes a component in batches (lots) through its operations. Suppose three
lots of the component have already been processed with the following lot sizes and lot average
hours, and the shop is about to process the fourth lot that consists of 40 units12 (LO3)
Lot No.

Lot Size

Lot Avg. Hours

6,800

1
2

4,500

15

3,500

40

Determine the appropriate lot plot points (the x-axis values of each lot), chart the
scatter plot of the data, fit a power function to the chart, and estimate the total hours for
the fourth lot.
*31. There has recently been a controversy over the expected cost of 65 F35 Joint Strike Fighter
jets that Government of Canada wants to purchase. F35s design phase has taken several years
longer than expected and costs are billions of dollars over budget. Also, Canada and the manufacturer, Lockheed Martin, have not yet signed a contract, so the price has not been determined
yet. In addition, accurate cost data for the first few prototypes are not available publicly, while
the prototypes are constantly being changed to improve them. This makes estimating the costs
of jets very difficult. The first lot of 32 production-grade jets are expected to be delivered to
the U.S. armed forces in 2012 at the cost of $207.6 million each (excluding the research and
development costs; see e.g., http://comptroller.defense.gov/defbudget/fy2012/FY2012_Weapons.pdf p15). Israel has negotiated to buy 20 jets to be delivered at the end of 2015 at the cost
of $137.5 million each (see e.g., http://www.aviationweek.com/aw/generic/story_channel.
jsp?channel=defense&id=news/awst/2010/08/23/AW_08_23_2010_p32-249396.xml). According to the production schedule, 330 jets will have been made by the end of 2015 (see e.g., the last
page of http://www.parl.gc.ca/PBO-DPB/documents/F-35_Cost_Estimate_EN.pdf).(LO3)
a. Using these two data points (midpoint of 16 for the U.S. purchase and midpoint of 340 for
the Israeli purchase, chart the scatter plot of the two data points, fit a power function to
the chart, and calculate the learning rate. Does it seem reasonable? Briefly explain.
b. Canadas jets are expected to be delivered between 2016 and 2020 in lots of approximately
13. The middle jet is expected to be the 848th jet made. Estimate the cost of the 848th jet
using the power function of part a, and then estimate the total cost of Canadas purchase.
Mini-Case

Product Recall

n automobile manufacturer is conducting a product recall

after it was discovered that a possible defect in the braking mechanism could cause loss of braking in certain cars. The
recall covers a span of three model years. The company sent out
letters to car owners promising to repair the defect at no cost
at any dealership.

12

The companys policy is to pay the dealer a fixed amount

for each repair. The repair is somewhat complicated, and the
company expects learning to be a factor. In order to set a reasonable rate for repairs, company engineers conducted a number
of repairs themselves. It was then decided that a rate of $88 per
repair would be appropriate, based on a flat hourly rate of $22
per hour and 90-percent learning.

http://www.acq.osd.mil/dpap/cpf/docs/contract_pricing_finance_guide/vol2_ch7.pdf.

PART THREE System Design

20

Shortly after dealers began making repairs, the company

received word that several dealers were encountering resistance
from workers who felt the flat rate was much too low and who
were threatening to refuse to work on those jobs. One of the
dealers collected data on job times and sent that information
to the company: three mechanics each completed two repairs.
Average time for the first unit was 9.6 hours and for the second
unit was 7.2 hours. The dealer has suggested a rate of $110
per repair.

You have been asked to investigate the situation and to

prepare a report.
Questions
1. Prepare a list of questions that you will need to have
answered in order to analyze this situation.
2. Comment on the information provided in the case.
3. What preliminary thoughts do you have on solutions/partial
solutions to the points you have raised?

Mini-Case

Learning Curve in Surgery

ew surgery techniques require a long learning curve on

the part of a surgeon. Random complications may arise
due to patients conditions. Therefore, it is important to know
the number of operations required to stabilize operating times
and complication rates.
Dr. Voitk of Salvation Army Scarborough Grace Hospital
reported the results of 100 consecutive operations for laparoscopic hernia repair on 98 patients. Approximately two-thirds of
surgeries were uni-lateral (right/left) and the remaining one-third
were bilateral (involving contra-lateral defects; many unsuspected
before surgery). The average surgery time (from skin incision
to skin closure) was 46 minutes for unilateral and 62 minutes
for bilateral. Surgery times for the unilateral procedure began
to level off after 50 operations. Dr. Voitk reported the average
surgery times (in minutes) for each quartile of the 100 operations,
classified by type of operation, as shown on the right.
At the end of the study the times had levelled off at 58 minutes (operating time), including 37 minutes (surgical time), for
unilateral type, which approached the historical times for open
repair. Learning also reduced the complication rates, which fell

in an approximately exponential manner, beginning to level off

at 50 operations and becoming stable after 75.
1st
Quartile

2nd
Quartile

3rd
Quartile

4th
Quartile

Unilateral

59

45

38

37

Bilateral

69

67

58

52

Questions
1. Determine the learning percentage for the unilateral laparoscopic hernia surgery. (Hint: Assume that the 8th, 24th,
40th, and 58th observations had unilateral surgeries.)
2. Determine the learning percentage for the bilateral laparoscopic hernia surgery. (Hint: Assume that the 4th, 12th,
20th, and 28th observations had the bilateral surgeries).
Compare your answers.
Source: Adapted from Dr. Andrus Voitk, The Learning Curve in
Laparoscopic Inguinal Hernia Repair for the Community General
Surgeon, Canadian Journal of Surgery 41(6) December 1998,
pp. 446450.

Case

Renovating the Lions Gate Bridge

Dr. William C. Wedley

en sections completed, 44 more to go. The repairs are

controversial, behind schedule, over budget, and subject
to penalties. American Bridge/Surespan staff is contemplating
how long it will take to install the remaining sections of Lions
Gate Bridge.
Lions Gate Bridge, the scenic entranceway to Vancouvers
Burrard Inlet, began service in 1938. At that time, its two lanes
of traffic appeared more than ample. By the 1950s however,

c ommuters to and from Vancouvers north shore were causing

rush-hour line-ups. In 1952 two lanes were converted to three
narrow lanes. But continued expansion on the north shore, growth
of the Whistler ski community, and increased ferry service to
Vancouver Island caused the bridge to reach its capacity.
Over the years, Lions Gate Bridge had undergone regular
maintenance. Although its suspension cables and steel superstructure have never been changed, the bridge has undergone
periodic painting and resurfacing. On the viaduct section in the
north, the cantilevering of sidewalks allowed increased lane
widths and separation of pedestrians from traffic. This accentuated the need for change on the rest of the bridge.

SUPPLEMENT TO CHAPTER 7 Learning Curves

By the 1990s, the annual cost of maintenance made the existing situation uneconomical. The bridge either had to be repaired
to good as new or an alternate crossing had to be built. The
debate became extreme. Some proposed doubling capacity with
a twin bridge or a new four-lane bridge. Other plans included
various locations for a tunnel or combinations of bridges and
tunnels. At the same time, the Parks Board and environmentalists
opposed cutting of trees or any encroachment on Stanley Park.
Consensus seemed impossible.
Complicating the issue was money. Highway construction is
a provincial responsibility, but deficit financing already overwhelmed the government. User tolls were a possibility, but north
shore communities claimed that tolls would be discrimination
when not applied to other bridges. Private ownership, while a
possibility, suffered from criticisms of tolls and monopoly. In
the end, money ruled. In May 1999, the provincial government
chose to replace only the deck of the existing bridge.
The outcry was instantaneous. While the bridge would be
made safer, the same capacity problems would remain. Public
meetings were held, citizens signed petitions, and stop work
injunctions were sought through the court system. When tempers

Q1

Exhibit 1Schematic Diagram of Lions Gate Bridge

Exhibit 2Schematic Diagram of the Jacking Traveller

13

Peter Lambur, The Vancouver Sun, October 5, 1999.

21

cooled, a detailed news article looked back and pronounced

Lions Gate Renewal Defies Reason.13
A team headed by American Bridge (AB) of Pittsburgh was
selected to rehabilitate the bridge. In order to carry out the work,
they assembled a group of other professionals for construction
design.
The total budget for the upgrade was slightly over $100
million. The most complex and challenging component of the
project is the requirement to replace the decking while the bridge
is still in use. American Bridges plan is to replace the bridge in
sectionscutting out and lowering an old section and raising
and inserting a wider prefabricated section during weekend or
overnight shutdowns. In the end, the bridge would be wider,
sleeker, and modernized.
In total, 54 sections must be removed and replaced. Removing and replacing deck panels requires tricky technology. Since
the bridge is suspended, removing a section releases the structural integrity that holds it together. In addition, without some
large clamping device to keep the bridge apart, it is possible
that the pieces would not fit.

22

PART THREE System Design

American Bridge designed and developed the jacking

traveller. This device travels along the bridge from section
to section as the work is done. Its essential function is twofold:
(1) to clamp the bridge in place while a section is removed and
(2) to position the jacks for lowering and raising sections. Once
a new section is securely in place, the jacking traveller moved
along to the next section, where the process is repeated.
Initially, AB had proposed to use the jacking traveller as a
car ramp as well, so that the bridge did not have to be closed
during panel replacement. However, later it was discovered that
the bridge may not be able to safely hold all the weight during
construction. This change in plan resulted in a six-month delay
in the start of construction, which was initially supposed to start
in January 2000 and finish in one year.13
The first panels were installed on the weekend of September
810, 2000. The plan was ambitiousto replace two panels in
one weekend closure.
The bridge was closed at 11 p.m. on Friday and work began.
At about 6 a.m., the first of the 54 panels was cut out and
lowered to the ground using the jacking traveller. Shortly later,
at daybreak, the new wider panel was lifted approximately 40
metres up to the bridge deck. Five hours later, after lateral jockeying, bolting, and welding, the panel was in place. This was
the first time that an entire piece of a suspension bridge of this
magnitude has been replaced at one time.

Q1

Exhibit 3Raising Panel 1, Daybreak, Sept 9, 2000 Photo from the

former project Website, courtesy of Stuart McCall/North Light

About noon Sunday, after 24 hours used for moving the jacking traveller, installation started on the second panel. Difficulty
was encountered removing rusted suspension pins and dealing
with a frozen expansion joint. In total, the second panel took
26 hours, and the bridge reopened at 2 p.m. Monday, well after
the scheduled time for completion.

13

The traffic jams on Monday morning were horrendous. Vehicles were backed up for kilometres, roads became grid-locked,
people were late for work, and economic activity was hindered.
From the first two panels, the project staff learned some
lessons. As the work at this stage was still above land, the old
section had been lowered 40 metres to dollies where it was
moved out of the way. Then a new panel was moved into place
and raised to the bridge deck. This lowering and raising process
took 2.5 hours. Toward the centre of the bridge and over water,
the vertical lifts will become longer.
The second lesson was that panels adjacent to towers (like
the first panel) are more difficult to replace. The detached old
section must be first swung sideways and then lowered without
striking the tower. Then, the replacement panel has to be raised
away from the tower and swung sideways once it gets to the top.
Since space is tight and limited, replacement of panels next to
towers takes longer.
They also discovered that moving and setting up the jacking
traveller between lifts takes a considerable time. In total, bridge
closure for the first two panels was 63 hours and 48 minutes. Of
this time, approximately 24 hours was used for repositioning the
jacking traveller and setting up for the second section. Hence,
actual removal and insertion time, exclusive of setup, was about
40 hours for the first two sections. Approximately 14 hours of
that time was for Panel 1. The remaining time, 26 hours, was
for Panel2, which required extra time to free a frozen expansion
joint.
Except for special circumstances at the towers and the south
abutment, bridge closures for future deck replacements were
planned for one deck section at a time. That way, the setup
procedure for the jacking traveller between lifts could be done
during nighttime closures or even when traffic was flowing. (The
jacking traveller did not obstruct the traffic, which just passed
under it.) The total disruption and closure time could thereby
be kept to a minimum.
To plan and monitor how long closures would take, management estimated replacement times. The time estimates and the
actual times for the first ten panels are shown in Table 1.
The actual times for the first three panels were well above
the teams estimates. The times for Panels 4 to 9 were much
better, being within two hours of the estimate. As well, the times
to complete Panels 4 to 8 improved as the work crews gained
experience. Panel 10, however, took much longer than expected.
Being adjacent to the north side of the North Tower, it was more
difficult to insert. As well, expansion problems complicated its
installation.
By looking at the replacement times for sections that were
adjacent to towers, the project staff could get an idea of the extra

A. Cho, Lions Gate Bridge Rehabilitation Proves a Lion-sized Challenge, ENR 248(3), January 28, 2002,
pp. 2225.

SUPPLEMENT TO CHAPTER 7 Learning Curves

23

Table 1
Bridge Closures
Date
FromTo

Time
Closed

Time
Opened

Time Work
Completed

Comments

Estimate
(hours)

Sept 811

22:12

14:00

63 hrs 48 mins

Panels #1 & #2 (three nights)

15 each

Sept 1617

20:05

13:30

17 hrs 25 mins

Panel #3

12

Sept 2324

22:06

12:26**

13 hrs 50 mins**

Panel #4 (** adjusted for 30 minute delay caused by a

person trying to jump from the bridge)

13

Sept 30Oct 1

22:05

10:42

12 hrs 37 mins

Panel #5

13

Oct 1415

22:03

8:27

10 hrs 24 mins

Panel #6

10.5

Oct 2122

22:07

8:07

10 hrs

Panel #7

10

Oct 2829

22:07

6:07**

9 hrs**

Panel #8 (** adjusted for the end of daylight savings time)

10

Nov 23

9:21

5:58

10 hrs 37 mins

Panel #9

9.75

Dec 23

22:30

17:30

19 hrs

Panel #10 (North of North Tower)

10

time required for working close to towers. The two panels adjacent
to towers (1 and 10) required 14 and 19 hours, respectively. Panel
10, however, was installed on a warm sunny day when expansion of metal caused the expected tolerances to disappear; this
compounded the space problem and lengthened the installation
process by four hours. Accordingly, the installation time for
panels next to towers is 14 or 15 hoursabout 5 hours longer
than sections that are clear of the towers.
With the experience of the first ten sections, the project
managers contemplated how much time would be required for
the remaining 44 sections. They knew that Panel 11 would be
difficult, because it is adjacent to a tower. To enable manoeuvrability around towers, Panels 10 and 11 are smaller in size
(10-metre sections).
Panels 12 to 34 are full-sized sections (20 metres) that cover
the main span between the north and south towers. Except for the
first two of those sections, all are delivered by sea. Two barges
with accompanying tugboats jockey into place below the opening, one to receive the old section and the other to deliver the
new panel. The required vertical lift from the barges depends
upon the tides at the time of the lift. The typical lift from the
water level is expected to be 55 metres, as opposed to 40 metres
over land. The increased distance and the jockeying of barges
are expected to increase the removal and lifting time by one hour
(i.e., 3.5 hours over water vs. 2.5 hours over land).
After Panel 34, the size changes to 10 metres but with no
change in the lifting, bolting, and welding times. Panels 35 and
36, next to the south tower, are expected to take longer. Panels
37 and 38 are spliced together beforehand, delivered on a narrow barge at high tide, and lifted and installed as one piece.
This splicing to form a double panel avoids one bridge closure,
but the horizontal jockeying of a larger section adds about two
hours to the replacement.

Panels 39 to 53 are installed differently. Since the land below

is steep cliff, delivery and installation is done from above. First,
the 10-metre section cut from the bridge is raised above the deck
level and turned 90 degrees. A truck with extended girders backs
onto the bridge from the south end so that its girders extend
over the gap. The old section is then lowered onto the girders,
moved onto the truck, and driven away. In the meantime, the
new section is backed over the entire bridge from the north end
on a similar truck. The new section is raised from the trucks
girders, rotated 90 degrees, and lowered into the opening after
the truck is removed. Although this process does not require as
much time jacking sections up and down, it requires the tricky
90-degree turn. In this regard, Panels 39 to 53 are new learning
experiences.
The final panel, number 54, is unique. Unlike Panels 39 to
53, it is longer (12 metres) and requires a different installation
procedure. Being longer, it cannot be backed over the bridge or
swung 90 degrees before installation. As a result, both removal
of the old section and delivery of the new panel are done from
the south abutment.
As a consequence of engineering delays and adverse weather
conditions in December and January, the replacement schedule
was cancelled for those months. This provided a time for reflection for the project managers. Given their experience to date, they
had to estimate how long each remaining section would take for
removal and replacement. With those estimates, they would be able
to determine a closure schedule that would minimize disruption to
the community, but ensure safety for workers and the public.
To help in this process, staff needed to prepare an estimate
sheet for panel installation times (see Table 2). With the actual
and estimated times for the first 10 panels and their knowledge
of challenges ahead, they agreed to map out a plan of action.
Their meeting is scheduled for next Monday. After that meeting,

PART THREE System Design

24

they will present their plan to their client, the B.C. Transportation Financing Authority. The client is insisting on an accurate and
reliable closure schedule.
Table 2
Estimate Sheet for Panel Installation Times
Panel #

Panel Size

Estimate (hours)

Actual Time (hours)

14.0

Land, At Tower

Lift Conditions

20m

15.0

26.0

Over Land

20m

15.0

17.4

Over Land

20m

12.0

13.8

Over Land

20m

13.0

12.6

Over Land

20m

13.0

10.4

Over Land

20m

10.5

10.0

Over Land

20m

10.0

9.0

Over Land

20m

10.0

10.6

Over Land

20m

10

19.0

Land, At Tower

10m

11

Land, At Tower

10m

12

Land and Water

20m

13

Over Water

20m

14

Over Water

20m

15

Over Water

20m

15

Over Water

20m

Questions
1. Given the data on the replacement of the first 10 panels,
is there any evidence that the work team is learning how
to make installations faster?
2. What techniques can be used to estimate the length of time
needed to perform each of the remaining installations?

9.75
10.0

3. How long will it take to perform each of the remaining

installations?
4. What allowances will have to be made for factors such
as (1) installations next to towers, (2) longer vertical lifts
over water, and (3) different installation procedures for
Panels 37 to 54?