Solutions

57:022 Principles of Design II

Homework Solutions Fall 1996
Dennis L Bricker
Dept. of Industrial Engineering
University of Iowa

•••••••••••••••••••••••••••• Homework # 1 ••••••••••••••••••••••••••••••••

1. A telephone exchange contains 6 lines. A line can be busy or available
for calls and all lines act independently. Each line is busy 75% of the
noon period (so that the probability that a line will be busy at any given
time during the noon period is 75%).
a. What is the probability of there being at least three free lines at any
given time during this period?
Sol’n: The number N6 of free lines at any given time will have the
binomial distibution with parameter n=6 and p=0.25, i.e., each of the
six lines corresponds to a “trial”, with the line being free corresponding
to “success”. Therefore
6 6

P {N 6 ≥ 3} = ∑   (0.25) (1 − 0.25) = 0.1694
i 6− i

i =3  i 

b. What is the expected number of free lines at any time during this
period?
Sol’n: The mean (expected value) of a random variable having a
binomial distribution is np, which in this instance is 6(0.25) =1.5

c. You need to make three calls to this exchange, and each time you
receive a busy signal you try again. What is the probability that you
require exactly six tries in order to complete your three calls?
Sol’n: Consider each call to the exchange to be a Bernouilli “trial”, and
Tk the number of the call on which you receive a free line for the kth
time. Then Tk has a Pascal Distribution (or, for the special case of k=1,
a geometric distribution). The required probability is therefore
 6 − 1
P {T 3 = 6} =   (0.25)3(1 − 0.25)6− 3
 3− 1
= 0.1318 + 0.0326 + 0.0044 + 0.0002
= 0.0659

2. The foreman of a casting section in a certain factory finds that on the

average, 1 in every 8 castings made is defective.
N10 = # defects in 10 castings.
N10 has the Binomial Distribution with parameters n=10 and
p=1/8.

57:022 HW#1 2/1/97 page 1

Solutions

a. If the section makes 10 castings a day, what is the probability that

none of these will be defective?
Sol’n: Applying the formula for the binomial distribution, we obtain
 10  1 0 1
10−0

P {N 10 = 0} =   1− = 0.2631
 0   8  8

b. If the section makes 10 castings a day, what is the probability that

two of these will be defective?
Sol’n: According to the same formula used in (a),
 10  1 2  1
10− 2

P {N 10 = 2} =   1− = 0.2416
 2   8  8

c. What is the probability that 3 or more defective castings are made in

one day?
10
Sol’n: P {N 10 ≥ 3} = ∑ P {N 10 = i} = 1− P {N 10 < 3}
i =3

= 1 - [0.2631+0.3758+0.2416]
= 0.1195
3. A light bulb in an apartment entrance fails randomly, with an expected
lifetime of 20 days, and is replaced immediately by the custodian. Assume
that this bulb's lifetime has an exponential distribution.
a. What is the probability that a bulb lasts longer than its expected
lifetime?
Sol’n: T1, the lifetime of the first bulb (i.e., the time of the first
failure), has exponential distribution with parameters λ =1 failure/20
days, i.e., the failure rate is 0.05 failures/day.
( ) 20) = 0.3679
P {T 1 > 20} = e 20 (
− 1

b. If the current bulb was inserted 10 days ago, what is the probability
that its lifetime (since it was inserted) will exceed the expected
lifetime of 20 days?
Sol’n: Because of the Memoryless Property of the exponential
distribution, the time between the 10th day and the failure again has
the same exponential distribution as T1. Therefore, since the
probability that the time from day 10 until the failure will exceed ten
days is
( ) (10)
P {T 1 > 10} = e 20
− 1
= 0.6065 ,
the probability that the total lifetime will exceed 20 days, given that it
is at least 10 days, is 60.65%.

c. If you were to test 10 of these bulbs, what is the probability that

more than half will exceed the expected lifetime?

57:022 HW#1 2/1/97 page 2

Solutions

Sol’n: Consider the testing of each bulb to be a Bernouilli trial, with

"success" corresponding to the bulb's exceeding the expected lifetime
of 20 days. Then the number of "successes" in the 10 trials will have
the binomial Distribution with parameters n=10 and p=0.3679 (from
the answer in (a)).
P {N 10 > 5} = P {N 10 = 6} + ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ + P {N 10 = 10}
= 0.083126 + 0.02765 + 0.00603 + 0.00078 + 0.00004
= 0.11763

d. If the custodian has 2 spare bulbs, what is the probability that these
(including the one currently in use) will be sufficient for the next 60
days?
Sol’n: Consider the Poisson process in which an event is defined to be
a failure of a bulb. Then the number of events during a 60-day period
will have the Poisson Distribution with rate λ =1 event per 20 days,
i.e., 1/20 = 0.05 events/day. The two spare bulbs will be sufficient for
the next 60 days, provided that the number of bulb failures is not
greater than 3. According to the formula for the Poisson distribution
with λ = 0.05 events/day and t=60 days is
P {N 60 ≤ 3} = 0.6472
•••••••••••••••••••••••••••• Homework # 2 ••••••••••••••••••••••••••••••••
up a hitchhiker is p=4%, i.e, an average of one in twenty-five drivers will stop;
different drivers, of course, make their decisions whether to stop or not
independently of each other.

1. Each car may be considered as a "trial" in a Bernouilli process, with "success"

defined as the car's stopping to pick up the hitchhiker.

2. Given that a hitchhiker has counted 15 cars passing him without stopping, what
is the probability that he will be picked up by the 25th car or before?

Soln: Let Z1 = the number of the Bernouilli trial in which “success”

first occurs, where p=0.04. Z1 has the geometric distribution, and
because the Bernouilli process is “memoryless”, we want the
probability that success occurs on or before the 10th trial, i.e.,
10 10
P {Z 1 ≤ 10} = ∑ P {Z 1 = n} = ∑ (1− p )n −1p
n =1 n =1

( )
= 0.04 0.96 + 0.961 + 0.962 +...... +0.969 =0.33517
0

57:022 HW#1 2/1/97 page 3

Solutions

Suppose that the arrivals of the cars form a Poisson process, at the average rate of 15
per minute. Define "success" for the hitchhiker to occur at time t provided that both
an arrival occurs at t and that car stops to pick him up. Let Y1 be the time (in
seconds) of the first "success", i.e., the time that he finally gets a ride, when he
begins his wait at time t =0.

3. What is the arrival rate of "successes"? _____0.6______

Soln: Arrival rate of "successes" λ = (λ ′) ⋅( p) = (15)⋅(0.04) = 0.6
(since the arrivals of the cars form a Poisson process at the average
rate of 15/min. and each of these cars have probability p =0.04 that
it will pick up a hitchhiker.)

4. What is the name of the probability distribution of Y1? Exponential

5. What is the value of E(Y1) ?___1.6667__

1 1
Soln: The mean of Exponential Distribution is =
λ 0.6

6.What's the value of Var(Y1)?___2.7778 ___

1 1
Soln: The variance of Exponential Distribution is 2 =
λ (0.6)2

7. What is the probability that he must wait less than 2 minutes for a ride (P{Y1<
2}? ___0.6988__
Soln: The CDF (cumulative distribution function) for Y1, which has
the exponential distribution, is
F(t) = P {Y 1 ≤ t } = 1 − e − λt where λ = 0.6.
Another equivalent approach:

57:022 HW#1 2/1/97 page 4

Solutions

{ } { }
If P T n < t is true, then P N t ≥ n is true, i.e., P {T n < t } ⇒ P {N t ≥ n}
P {T 1 < 2} = P {N 2 ≥ 1} = 1− P {N 2 < 1} = 1 − P {N 2 = 0}
−(0.6)(2) {(0.6)(2) }
0

=1- e = 0.6988
0!

8. What is the probability that he must wait more than 2 minutes for a ride(P{Y1>
2}? Soln: ___0.3012___
P {Y 1 > 2} = 1− P {Y 1 ≤ 2} = 1− F (2) = e ( ) = 0.3012
− 0.6 2

Equivalently,
Using the property " P {T n > t } ⇔ P {N t < n} "
−(0.6)(2) {(0.6)(2)}
0

P {T 1 > 2} = P {N 2 < 1} = P {N 2 = 0} = e =0.3012

0!

9. What is the probability that he must wait exactly 2 minutes for a ride? (P{Y1=
2}? Soln: ____0_____.
The probability that the event occurs in a time interval of length zero
is zero!

Note that your answers in 7-9 must have a sum equal to 1!

Suppose that after 2 minutes (during which 28 cars have passed by) the hitchhiker is
still there waiting for a ride.

10. What is the conditional expected value of Y1 (expected total waiting time,
given that he has already waited 2 minutes). ____3.6667___
Soln: Because of the memoryless property of the Exponential
distribution, the expected additional time until a car stops is identical
to the original expected time until the car stops. That is, he now
 1 
expects to wait a total of 2+ E (T ) = 2 +   = 3.6667
1  0.6 

10. What is your ID number? ___485-21-1250__ (Han-Suk’s ID #)

Note: Your solutions to #11, 12, 14, and 15 below will vary
because of the different random number seed used in the
simulation!

11. Using the first four digits of your ID number as the "seed" for the "Midsquare"
technique, generate a sequence of 10 pseudo-random numbers R1, R 2, ... R10
uniformly distributed in the interval [0,1].

Soln: Starting with the “seed” X0=4852, we get the “pseudo-random”

sequence

57:022 HW#1 2/1/97 page 5

Solutions

i Xi Xi 2 Ri
1 4852 23541904 0.5419
2 5419 29365561 0.3655
3 3655 13359025 0.3590
4 3590 12888100 0.8881
5 8881 78872161 0.8721
6 8721 76055841 0.0558
7 0558 00311364 0.3113
8 3113 09690769 0.6907
9 6907 47706649 0.7066
10 7066 49928356 0.9283

12. Using the "Inverse Transformation" technique and the 10 numbers generated in
(10.), generate the interarrival times τ1, τ2, ... τ10 for the first ten cars with
arrival rate λ = 15/minute, and then the arrival times T1, T2, ... T10 of those ten
cars.
Soln:
i Ri τi Ti
1 0.5419 0.05204452 0.05204452
2 0.3655 0.03032787 0.08237238
3 0.3590 0.02964839 0.11202077
4 0.8881 0.14600998 0.25803075
5 0.8721 0.13710044 0.39513119
6 0.0558 0.00382782 0.39895901
7 0.3113 0.02486330 0.42382231
8 0.6907 0.07822957 0.50205188
9 0.7066 0.08174789 0.58379978
10 0.9283 0.17568430 0.75948408

ln  R  ln 1− R 
( Since τ = − i =− i , where τ is the interarrival time,
i
i λ λ
i.e., τ1 =T1, τ2 = T2 - T1, τ3 = T3 - T2, etc., so that the arrival times are
T1=τ1, T2=τ1 + τ2, T3=τ1+ τ2 + τ3, etc.)

13. What is the expected number of arrivals during the first 20 seconds?
  = (15 / min)( 1 min) = 5
Soln: E N
1/3  3

14. What is the actual # of arrivals during the first 20 seconds of your simulation?
Soln: 4 arrivals in the first 20 seconds. ( Since the 5th arrival
occurs at 0.3952 minutes which is later than 0.3333 minutes(= 20
seconds). )

15. What is the probability that you would observe exactly this number of arrivals
in this Poisson process? 0.1755 .
Soln:

57:022 HW#1 2/1/97 page 6

Solutions

{(15)(1 / 3)}
4
54
{ }
P N 1/3 = 4 = e −(15)(1/ 3)

4!
= e −5
4!
= 0.1755

•••••••••••••••••••••••••••• Homework # 3 ••••••••••••••••••••••••••••••••

1. Consider again the proposed drive-up bank teller window in the Hypercard Stack
"Intro. to Simulation", but where the arrival rate of customers has increased to
20/hour (one every 3 minutes). The SLAM models for two models (one with a
single teller window and the second with two teller windows), together with
output, appear below.

••••••••••••••• Case A: single teller •••••••••••••••

GEN,BRICKER,BANKTELLERS,9/12/1996,,,,,,,72;
LIM,2,1,50;
INIT,0,480;
NETWORK;
CREATE,EXPON(3.0),,1;
QUE(1),0,4,BALK(OVFLO);
ACT(1)/1,EXPON(2.0);
COLCT,INTVL(1),CUSTOMER_TIME,20/.5/.5;
TERM;
OVFLO COLCT,ALL,OVERFLOW;
TERM;
END;
FIN;

S L A M I I S U M M A R Y R E P O R T

SIMULATION PROJECT BANKTELLERS BY BRICKER

DATE 9/12/1996 RUN NUMBER 1 OF 1

CURRENT TIME .4800E+03

STATISTICAL ARRAYS CLEARED AT TIME .0000E+00

During the 480 simulated time units, there were a total

of 158 cars served by the bank teller. The average time
in the system for these cars was 5.86 time units with a
s t a n d a r d d e v i a t i o n o f 5.20 time units and times ranged
from 0.0345 to 23.5 time units. The distribution for
time in the system is depicted by the histogram generated
by SLAM II. There was a total of 1 observation of
overflow.

A histogram of the values collected at a COLCT node can

be obtained. This is accomplished by specifying on input
t h e n u m b e r o f i n t e r i o r c e l l s , N C E L ; the upper limit of
t h e f i r s t cell, HLOW; and a cell width, HWID, for the
histogram. The number of cells specified, NCEL, each of
which will have a width of HWID. T w o additional cells
will be added that contain the interval (- ∞,HLOW] and
(HLOW+NCEL*HWID, ∞).

**STATISTICS FOR VARIABLES BASED ON OBSERVATION**

57:022 HW#1 2/1/97 page 7

Solutions

MEAN STANDARD COEFF. OF MINIMUM MAXIMUM NO.OF

VALUE DEVIATION VARIATION VALUE VALUE OBS

CUSTOMER_TIME .586E+01 .520E+01 .887E+00 .345E-01 .235E+02 158

OVERFLOW .182E+03 .000E+00 .000E+00 .182E+03 .182E+03 1

The second category of statistics for this example is

the “file statistics”. The statistics for file 1
correspond to the cars waiting for service at bank
t e l l e r w i n d o w w a s 1 . 2 5 c a r s , w i t h a standard deviation
o f 1 . 4 7 3 c a r s , a m a x i m u m o f 4 c a r s w a i t e d , and at the
end of the simulation there were no cars in the queue.

**FILE STATISTICS**

FILE AVERAGE STANDARD MAXIMUM CURRENT AVERAGE

NUMBER LABEL/TYPE LENGTH DEVIATION LENGTH LENGTH WAIT TIME

1 QUEUE 1.250 1.473 4 0 3.797

2 .000 .000 0 0 .000
3 CALENDAR 1.679 .467 3 1 1.729

The last category of statistics for this example is

“statistics on service activities”. The first row of service
activity statistics corresponds to the server at bank teller
window who was busy 67.9 percent of the time. Since the
capacity of the server is one, the values 12.84 and 87.10
refer to the maximum length of the server idle period and the
server busy period, respectively.

**SERVICE ACTIVITY STATISTICS**

ACT ACT LABEL OR SER AVERAGE STD CUR AVERAGE MAX IDL MAX BSY ENT
NUM START NODE CAP UTIL DEV UTIL BLOCK TME/SER TME/SER CNT

1 QUEUE 1 .679 .47 0 .00 12.84 87.10 158

**HISTOGRAM NUMBER 1**

CUSTOMER_TIME

OBS RELA UPPER

FREQ FREQ CELL LIM 0 20 40 60 80 100
+ + + + + + + + + + +
9 .057 .500E+00 +*** ˆ +
11 .070 .100E+01 +*** C | +
13 .082 .150E+01 +**** C | +
14 .089 .200E+01 +**** C | +
4 .025 .250E+01 +* C | +
8 .051 .300E+01 +*** C | +
9 .057 .350E+01 +*** C | +
2 .013 .400E+01 +* C | +
7 .044 .450E+01 +** C | +
9 .057 .500E+01 +*** C +
4 .025 .550E+01 +* |C +
6 .038 .600E+01 +** | C +
4 .025 .650E+01 +* | C +
5 .032 .700E+01 +** | C +
5 .032 .750E+01 +** | C +
5 .032 .800E+01 +** | C +
5 .032 .850E+01 +** | C +

57:022 HW#1 2/1/97 page 8

Solutions

6 .038 .900E+01 +** | C +

2 .013 .950E+01 +* | C +
4 .025 .100E+02 +* | C +
3 .019 .105E+02 +* | C +
23 .146 INF +******* ˘ C
--- + + + + + + + + + + +
158 0 20 40 60 80 100

**STATISTICS FOR VARIABLES BASED ON OBSERVATION**

MEAN STANDARD COEFF. OF MINIMUM MAXIMUM NO.OF

VALUE DEVIATION VARIATION VALUE VALUE OBS

CUSTOMER_TIME .586E+01 .520E+01 .887E+00 .345E-01 .235E+02 158

••••••••••••••• Case B: two tellers •••••••••••••••

GEN,BRICKER,BANKTELLERS,9/12/1996,,,,,,,72;
LIM,2,1,50;
INIT,0,480;
NETWORK;
CREATE,EXPON(3.0),,1;
QUE(1),0,4,BALK(OVFLO);
ACT(2)/1,EXPON(2.0);
COLCT,INTVL(1),CUSTOMER_TIME,20/.5/.5;
TERM;
OVFLO COLCT,ALL,OVERFLOW;
TERM;
END;
FIN;

S L A M I I S U M M A R Y R E P O R T

SIMULATION PROJECT BANKTELLERS BY BRICKER

DATE 9/12/1996 RUN NUMBER 1 OF 1

CURRENT TIME .4800E+03

STATISTICAL ARRAYS CLEARED AT TIME .0000E+00

During the 480 simulated time units, there were a total

of 179 cars served by the bank teller. The average time
in the system for these cars was 2.37 time units with a
standard deviation of 2.60 time units and times ranged
from 0.0117 to 23.6 time units. The distribution for
time in the system is depicted by the histogram
generated by SLAM II.

**STATISTICS FOR VARIABLES BASED ON OBSERVATION**

MEAN STANDARD COEFF. OF MINIMUM MAXIMUM NO.OF

VALUE DEVIATION VARIATION VALUE VALUE OBS
CUSTOMER_TIME .237E+01 .260E+01 .110E+01 .117E-01 .236E+02 179
OVERFLOW NO VALUES RECORDED

The second category of statistics for this example is

the “file statistics”. The statistics for file 1
correspond to the cars waiting for service at bank
teller window was 0.099 cars, with a standard deviation

57:022 HW#1 2/1/97 page 9

Solutions

of 0.355 cars, a maximum of 4 cars waited, and at the

end of the simulation there were no cars in the queue.

**FILE STATISTICS**

FILE AVERAGE STANDARD MAXIMUM CURRENT AVERAGE

NUMBER LABEL/TYPE LENGTH DEVIATION LENGTH LENGTH WAIT TIME
1 QUEUE .099 .355 4 0 0.266
2 .000 .000 0 0 .000
3 CALENDAR 1.785 .794 4 1 2.164

The last category of statistics for this example is

“statistics on service activities”. The first row of
service activity statistics corresponds to the service
activity at the bank teller window: the average
utilization was 0.785 which is the average number of busy
servers. Since there are two servers, the utilization of
each individual server was 0.785/2 = 39.25%, which means
that each server was busy an average of 39.25% of the
day. Note that since this activity has a capacity of 2,
the maximum idle and busy values refer to the number of
servers, instead of the service times.

**SERVICE ACTIVITY STATISTICS**

ACT ACT LABEL OR SER AVERAGE STD CUR AVERAGE MAX IDL MAX BSY ENT
NUM START NODE CAP UTIL DEV UTIL BLOCK TME/SER TME/SER CNT
1 QUEUE 2 .785 .79 0 .00 2.00 2.00 179

**HISTOGRAM NUMBER 1**

CUSTOMER_TIME

OBS RELA UPPER

FREQ FREQ CELL LIM 0 20 40 60 80 100
+ + + + + + + + + + +
27 .151 .500E+00 +******** ˆ +
30 .168 .100E+01 +******** C | +
28 .156 .150E+01 +******** C | +
16 .089 .200E+01 +**** C | +
17 .095 .250E+01 +***** C | +
11 .061 .300E+01 +*** C | +
11 .061 .350E+01 +*** C | +
9 .050 .400E+01 +*** C | +
9 .050 .450E+01 +*** C| +
4 .022 .500E+01 +* C +
3 .017 .550E+01 +* |C +
2 .011 .600E+01 +* | C +
2 .011 .650E+01 +* | C +
2 .011 .700E+01 +* | C +
2 .011 .750E+01 +* | C +
1 .006 .800E+01 + | C+
0 .000 .850E+01 + | C+
1 .006 .900E+01 + | C+
1 .006 .950E+01 + | C+
0 .000 .100E+02 + | C+
1 .006 .105E+02 + | C+
2 .011 INF +* ˘ C
--- + + + + + + + + + + +
179 0 20 40 60 80 100

57:022 HW#1 2/1/97 page 10

Solutions

**STATISTICS FOR VARIABLES BASED ON OBSERVATION**

MEAN STANDARD COEFF. OF MINIMUM MAXIMUM NO.OF

VALUE DEVIATION VARIATION VALUE VALUE OBS
CUSTOMER_TIME .237E+01 .260E+01 .110E+01 .117E-01 .236E+02 179

••••••••••••••••••••••••••••••

a. How many cars were unable to enter the queue each day, because the queue was
filled to capacity?
Case A: 1 Case B: none
b. What fraction of the time was each teller busy each day?
Case A: 67.9 % Case B: 39.25 %
c. Estimate the mean (average) time in the system for the customers.
Case A: 5.86 min . Case B: 2.37 min.
d. What fraction of the customers spend more than 5 minutes (total of both waiting
and being served) at the bank?
Case A: 72 / 158 = 45.6% Case B: 17 / 179 = 9.5%

•••••••••••••••••••••••••••• Homework # 4 ••••••••••••••••••••••••••••••••

1. The numbers of arrivals during 100 hours of what is believed to be a Poisson
process were recorded. The observed numbers ranged from zero to nine, with
frequencies O0 through O9:

The average number of arrivals was 3.8/hour. We wish to test the "goodness-of-fit"
of the assumption that the arrivals correspond to a Poisson process with arrival rate
3.8/hour. The first step is to compute the probability of each observed value, 0
through 9:

57:022 HW#1 2/1/97 page 11

Solutions

0.1943588

a. What is the value missing above? (That is, the probability that the number of
arrivals is exactly 4.)
(3.8)4
e −3.8 = 0.1943588
4!

b. Now, we can compute the expected number of observations of each of the

values 0 through 9, which we denote by E0 through E 9.
• What is the expected number of times in which we would observe four
arrivals per hour?
Soln: 19.44
• Did we observe more or fewer than the expected number?
Soln: The observed number (=24) is more than the expected number (=19.44)

c. Complete the table below:

0.1943588 19.43588 20.83119 1.07179

d. Ignoring the suggestion that cells should be aggregated so that they contain at
least five observations, what is the observed value of
2
D= Σi
Ei -Oi
Ei
?

57:022 HW#1 2/1/97 page 12

Solutions

Soln: 7.8241705
e. Keeping in mind that the assumed arrival rate λ =3.8/hour was estimated from
the data, what is the number of "degrees of freedom"?
Soln: 8 (=10-1-1) degrees of freedom
f. Using a value of α = 5%, what is the value of the quantity
2
χ
5% such that D

χ2
exceeds 5% with probability 5% (if the assumption is correct that the
arrivals form a Poisson process with arrival rate 3.8/hour)?
Soln: 15.507
2
g. Is the observed value greater than or less than 5% ? χ
χ 2
Soln: D (=7.82) < 5% (=15.507)

Should we accept or reject the assumption that the arrival process is Poisson
with rate 3.8/hour?
Soln: Accept the assumption that the arrival process is Poisson with rate
3.8/hour.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2. Consider again the SLAM II example of the garment factory with 50 sewing
machine operators and machines, 54 machines (including 4 backup machines) and
3 repairmen. The original system was modeled by the following SLAM statement.
Note that the time units are hours, and that the system is simulated for 2000 hours,
i.e., 50 weeks of 40 hours each.

GEN,BRICKER,GARMENTFACTORY,9/18/1996,,,,,,,72;
LIM,2,,55;
INIT,0,2000;
NETWORK;
Q1 QUE(1),4;
ACTIVITY(50)/1,EXPON(157);
Q2 QUE(2);
ACTIVITY(3)/2,UNFRM(4,10),,Q1;
END;
FIN;
The output of this SLAM model is
S L A M I I S U M M A R Y R E P O R T

SIMULATION PROJECT GARMENTFACTORY BY BRICKER

DATE 9/18/1996 RUN NUMBER 1 OF 1
CURRENT TIME .2000E+04
STATISTICAL ARRAYS CLEARED AT TIME .0000E+00

**FILE STATISTICS**

FILE AVERAGE STANDARD MAXIMUM CURRENT AVERAGE

NUMBER LABEL/TYPE LENGTH DEVIATION LENGTH LENGTH WAIT TIME
1 Q1 QUEUE 1.420 1.311 4 0 4.396
2 Q2 QUEUE .802 1.578 11 4 2.473
3 CALENDAR 51.778 1.390 54 50 48.778

**SERVICE ACTIVITY STATISTICS**

57:022 HW#1 2/1/97 page 13

Solutions

ACT ACT LABEL OR SER AVERAGE STD CUR AVERAGE MAX IDL MAX BSY ENT
NUM START NODE CAP UTIL DEV UTIL BLOCK TME/SER TME/SER CNT
1 Q1 QUEUE 50 49.542 1.29 47 .00 10.00 50.00 649
2 Q2 QUEUE 3 2.236 .98 3 .00 3.00 3.00 642

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
The modified example, in which the repair of the failed machines is subcontracted if
there are already two machines in the repair queue, is modeled by the SLAM
statements:

GEN,BRICKER,GARMENTFACTORY,9/18/1996,,,,,,,72;
LIM,2,,55;
INIT,0,2000;
NETWORK;
Q1 QUE(1),4;
ACTIVITY(50)/1,EXPON(157);
Q2 QUE(2),,2,BALK(SUB);
ACTIVITY(3)/2,UNFRM(4,10),,Q1;
SUB COLCT,BETWEEN,SUBCONTRACT;
ACTIVITY,UNFRM(20,30),,Q1;
END;
FIN;
The output of this SLAM model is

S L A M I I S U M M A R Y R E P O R T

SIMULATION PROJECT GARMENTFACTORY BY BRICKER

DATE 9/18/1996 RUN NUMBER 1 OF 1

CURRENT TIME .2000E+04

STATISTICAL ARRAYS CLEARED AT TIME .0000E+00

**STATISTICS FOR VARIABLES BASED ON OBSERVATION**

MEAN STANDARD COEFF. OF MINIMUM MAXIMUM NO.OF

VALUE DEVIATION VARIATION VALUE VALUE OBS

SUBCONTRACT .527E+02 .729E+02 .138E+01 .881E-01 .262E+03 35

**FILE STATISTICS**

FILE AVERAGE STANDARD MAXIMUM CURRENT AVERAGE

NUMBER LABEL/TYPE LENGTH DEVIATION LENGTH LENGTH WAIT TIME

1 Q1 QUEUE 1.488 1.270 4 1 4.694

2 Q2 QUEUE .277 .577 2 0 .926
3 CALENDAR 52.235 1.060 54 53 51.086

**SERVICE ACTIVITY STATISTICS**

ACT ACT LABEL OR SER AVERAGE STD CUR AVERAGE MAX IDL MAX BSY ENT
NUM START NODE CAP UTIL DEV UTIL BLOCK TME/SER TME/SER CNT

1 Q1 QUEUE 50 49.706 .83 50 .00 7.00 50.00 633

2 Q2 QUEUE 3 2.082 .98 3 .00 3.00 3.00 594

57:022 HW#1 2/1/97 page 14

Solutions

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Note: When “between” statistics are collected, the number of observations is the
number of intervals between arrival of entities, i.e., one less than the number of
arrivals!

Suppose that
• the hourly wage of a sewing machine operator (including benefits, etc.) is $15.
• the hourly wage of a machine repairman (including benefits, etc.) is $25.
• each machine operator (when busy) generates $27.50 per hour of revenue for
the company,
• the cost of subcontracting the repair of a machine is $250.

a. What is the annual payroll for machine operators?

Soln: $ 30,000 (=2000 × 15 ) per operator; total is 50($30,000) = $1.5 million
b. What is the annual payroll for repairmen?
Soln: $ 50,000 (=2000 × 25 ) per repairman; total is 3($50,000) = $150,000
c. What is the average direct cost of repairing a machine in-house?
Soln: $ 77.88 / machine =$25(2000/642)
d. How many repairs were subcontracted during the year?
Soln: 36 (Since there were 35 observations of the interval between such repairs,
the number of the subcontracted repairs is 36.)
e. How frequently were repairs subcontracted?
Soln: once every 52.7 hours (from SLAM II output).
f. What is the average length of time that machines wait in the repair queue?
Soln: Without subcontracting: 2.473 hours
Soln: With subcontracting: 0.926 hours
g. What is the maximum number of idle machine operators during the year?
Soln: Without subcontracting: 10 operators
Soln: With subcontracting: 7 operators
h. What is the utilization of each repairman, i.e. the % of time busy?
Soln: Without subcontracting: 74.5%
Soln: With subcontracting: 69.4 %
i.. How many more man-hours of machine operator time per year are utilized when
the repairs are subcontracted?
Soln: 328 hours more ( = 49.706-49.542) × 2000 )
j. How much more revenue per year is earned by the company when repairs are
subcontracted?
Soln: With subcontracting : (49.706 × 27.5 × 2000)-(250 × 36) = 2,724,830
Without subcontracting, revenue is (49.542 × 27.5 × 2000) = 2,724,810
Difference is $20
k. Is the decision to subcontract the repairs cost-effective? That is, will the increase
in the company ‘s revenues compensate for the cost of subcontracting?
Soln: Yes, but the increase is so negligible that the company should decide based
upon other non-quantifiable factors.

•••••••••••••••••••••••••••• Homework # 5 ••••••••••••••••••••••••••••••••

Regression Analysis. Tests on the fuel consumption of a vehicle traveling at
different speeds yeilded the following results:

Speed s (mph) 20 30 40 50 60 70 80 90
Consumption C (mile/gal.) 11.4 17.9 22.1 25.5 26.1 27.6 29.2 29.8

(Note: the above data is complete fictitious!)

57:022 HW#1 2/1/97 page 15

Solutions

It is suggested that the relationship between the two variables is of the form C = a + b/s.

a. Run the "Cricket Graph" software package (on the Macintosh), and enter the
above observed values of s and C (columns 1 and 2).

b. Plot the "scatter plot" of C versus s by choosing "scatter" on the Graph menu, and
specifying s on the horizontal axis and C on the vertical axis. Does the plot appear
to be linear? ___NO____

Data from "Homework#5"

30

20 C
C

10
0 20 40 60 80 100

c. Choose "transform" from the "data" menu to create a new variable 1/s which is the
reciprocal of s. (Put this new variable into Column 3.)

57:022 HW#1 2/1/97 page 16

Solutions

d. Plot the "scatter plot" of 1/s (horizontal axis) versus C (vertical axis). Does the plot
appear to be linear? ___Yes ____

Data from "Homework#5"

30

20 C
C

10
0.01 0.02 0.03 0.04 0.05 0.06

1/s

e. After plotting C versus 1/s, select "Simple" from the "Curve Fit" menu, in order to
fit a simple linear relationship between C and 1/s, i.e., to determine a and b such
that C ≈ a + b(1/s). What is the value of a? ___34.604 ___ of b? ___-476.94 ___

57:022 HW#1 2/1/97 page 17

Solutions

34.604 476.94

57:022 HW#1 2/1/97 page 18

Solutions

f. Another suggestion is that the relationship is of the form C = axb . Perform the
appropriate linear regression to fit a curve of this type. What is the value of a?
_2.14067(=e 0.76112 ) _ of b? __0.60572 _

Data from "Homework#5"

3.4

3.2

3.0
ln C

ln C
2.8

2.6

2.4
2 3 4 5

ln s
y = 0.76112 + 0.60572x R^2 = 0.923

◊◊◊◊◊◊◊◊◊◊◊◊◊◊◊◊◊◊

Next, start the shareware program “Curve Fit 0.7e ” (author: Kevin Raner) which
can be found on the “Public Software” fileserver of ICAEN. This program can
perform nonlinear regression directly, without first linearizing the function.

g. Enter the values of C and s as before:

2.80000E+01

2.40000E+01

2.00000E+01

1.60000E+01

1.20000E+01

3.00000E+01 6.00000E+01 9.00000E+01

57:022 HW#1 2/1/97 page 19

Solutions

Next, enter the function “f(x) = a + b/x” and give initial values of 1 to the
parameters a & b . Then choose “Custom Fn” from the menu in order to find the
values of a & b which will minimize the sum of the squared errors, using a nonlinear
optimizing algorithm, e.g. a Quasi-Newton algorithm. (You may have encountered
this type of algorithm, e.g. Fletcher-Powell, in 57:021 “Principles of Design I”.) Be
sure to indicate that both a & b are to be selected:

57:022 HW#1 2/1/97 page 20

Solutions

h. What is the optimal fitted curve? C = (34.6 ) + (-477 ) / s. What is the SSE (sum of
the squared errors)? ___2.48990 ___

2.80000E+01

2.40000E+01

a=3.46E+01
b=-4.77E+02
2.00000E+01

1.60000E+01
Sum of Squares
=2.48990

1.20000E+01

3.00000E+01 6.00000E+01 9.00000E+01

57:022 HW#1 2/1/97 page 21

Solutions

i. Next, enter the power function

and find the optimal curve of this type: C = 3.04 s0.519 . What is the SSE for this
fitted curve? 19.80037

2.80000E+01

2.40000E+01
a=3.04E+00
b=5.19E-01

2.00000E+01

1.60000E+01 Sum of Squares

=19.80037

1.20000E+01

3.00000E+01 6.00000E+01 9.00000E+01

j. You now have four candidate curves. Write them below, together with their SSE’s.
(You may need to compute the values of SSE for the curves fit by Cricket Graph
manually.)
C = asb SSE C = a + b/s SSE
--------------------------------------------------------------------------------------

Cricket Graph: C = 2.14067s0.60572 25.7847687 C = 34.604 - 476.94 / s 2.48989893

Curve Fit C = 3.04s0.519 19.80037 C = 34.6 - 477 / s 2.48990

---------------------------------------------------------------------------------------
Which one has the smallest SSE?

Soln : The question was perhaps vague, in that “SSE” did not specify in which
equation the error was to be computed, i.e., ln Ci = ln a + b×ln s i or Ci = asbi . My
intention was the latter, i.e., the error in the original curve rather than the
linearized curve.

57:022 HW#1 2/1/97 page 22

Solutions

There is neglible difference between the SSE in the fitted curves of the form C = a +
b/s found by Cricket Graph and Curve Fit 0.7e. (This is because the objective
functions for the optimization are identical.)

However, this SSE is considerably smaller than that of the two fitted curves of the
form C = asb . Note that for the form C = asb , the SSE in the curve found by Curve Fit
0.7e is less (19.8 compared to 25.78) than that in the curve found by Cricket Graph.
(In these two cases, the objective functions were different, namely
8
∑
2
Minimize ln Ci - ln a + b×ln s i in the case of Cricket Graph,
i=1
8 2
Minimize ∑ Ci - as bi in the case of Curve Fit 0.7e.
i=1

•••••••••••••••••••••••••••• Homework # 6 ••••••••••••••••••••••••••••••••

Reasoning that the failure time of a mechanical device is the minimum of the failure
times of its individual elements (all nonnegative random variables), we will
therefore assume that the failure time of the device has approximately a Weibull
distribution. Suppose that 500 units of this device are operated simultaneously for
280 days, at which time 100 have failed, with the failure times:

To estimate the expected lifetime (and its standard deviation) by recording the failure
times of all 500 units would require perhaps several years, an excessive amount of
time. Hence, we wish to estimate the Weibull parameters u & k from only the data
above, and use these to estimate µ and σ.

a. Group the observations as follows:

Cumulative
Interval # failures # failures % surviving
0- 50 2 2 99.6
50-100 4 6 98.8
100-150 15 21 95.8
150-200 26 47 90.6
200-250 32 79 84.2
250-280 21 100 80.0

57:022 HW#1 2/1/97 page 23

Solutions

b. Enter the observed values of t and R (failure time & reliability, i.e., the fraction
surviving) into the Cricket Graph program.
c. Using "transform" on the menu:
- compute ln t and place it into column 3 of the data matrix.
- compute R-1 and place it into column 4 of the data matrix.
- compute ln R-1 and place it into column 5 of the data matrix.
- compute ln (ln R-1) and place it into column 6 of the data matrix.

t # of failures NS FS % Surviving
(= NF ) ( = % FS)
5.00E+01 2.00E+00 4.98E+02 9.96E-01 9.96E+01
1.00E+02 6.00E+00 4.94E+02 9.88E-01 9.88E+01
1.50E+02 2.10E+01 4.79E+02 9.58E-01 9.58E+01
2.00E+02 4.70E+01 4.53E+02 9.06E-01 9.06E+01
2.50E+02 7.90E+01 4.21E+02 8.42E-01 8.42E+01
2.80E+02 1.00E+02 4.00E+02 8.00E-01 8.00E+01

t R In t 1/R In (1/R) In ( In(1/R) )

5.00E+01 9.96E-01 3.91E+00 1.00E+00 4.01E-03 -5.52E+00
1.00E+02 9.88E-01 4.61E+00 1.01E+00 1.21E-02 -4.42E+00
1.50E+02 9.58E-01 5.01E+00 1.04E+00 4.29E-02 -3.15E+00
2.00E+02 9.06E-01 5.30E+00 1.10E+00 9.87E-02 -2.32E+00
2.50E+02 8.42E-01 5.52E+00 1.19E+00 1.72E-01 -1.76E+00
2.80E+02 8.00E-01 5.63E+00 1.25E+00 2.23E-01 -1.50E+00

d. Plot the scatter graph of the data, with ln t (column 3) on the horizontal axis,
and ln (ln R-1) (column 6) on the vertical axis. Do the points appear to lie on a
straight line?
Yes, the points appear to lie on a straight line.

57:022 HW#1 2/1/97 page 24

Solutions

Data from "data.bc"

-1

-2
ln ( ln (1/R) )
-3

ln ( ln (1/R) )
-4

-5

-6
3 4 5 6

ln t

e. Fit a line to the points, using the Cricket Graph program. What is the equation
of the line? What is its slope and y-intercept?
The Equation is " y = -15.225 + 2.4244 x "
with slope 2.4244 and y-intercept -15.225

Data from "data.bc"

-1

-2
ln ( ln (1/R) )

-3

ln ( ln (1/R) )
-4

-5

-6
3 4 5 6

ln t
y = - 15.225 + 2.4244x R^2 = 0.984

57:022 HW#1 2/1/97 page 25

Solutions

f. Based upon the fitted line, what are the parameters u & k of the Weibull
distribution? u = _______________ , k= _______________
k = slope = 2.4244 ≈ 2.4

y-intercept = -k In u = -15.225 ⇒ In u = 6.2799 ⇒ u = e6.2799 = 533.7353

g. According to these values of u & k, what is

... the expected lifetime of the device? 473.1563 days
... the standard deviation of the device’s lifetime? 210.0341 days
... the probability that the device will fail during the first 400 hours is 0.3937


µ Y = u ⋅Γ  1 + 
1
Soln: k = (533.7353)(0.8865) = 473.1563 = expected lifetime
 2
Γ  1+ 
σY k −1
= σY
Γ 21 + 
µY 1 0.9407
= −1
 k ⇒ 473.1563 (0.8865)2 = 0.4439 ⇒ σ Y = 210.0341
Note: The above evaluations of the Gamma function were done using the tables
which were e-mailed to the class (& put onto the web site).
 400  2.4
− 
F(400) = P{t ≤ 400} = 1 − e  533.7353 
= 0.3937 ∴ 39.37 %

or using the "ProbLib" APL workspace on the macintosh, we can find that
the expected lifetime of the device is 473.1468,
the standard deviation of the device's lifetime is 209.9998, and
the probability that the device will fail during the first 400 hours is 0.3937

h. Using the parameters u & k, compute the expected number of failures (of the
100 units tested) in each of the intervals:

Observed Expected
Interval # failures # failures
0- 50 ___2___ __1.7__
50-100 ___4___ __7.2__
100-150 __15___ _14.3__
150-200 __26____ _22.0__
200-250 __32____ _29.55_
250-280 __21____ _21.0__
Note: The Weibull cumulative distribution function (CDF) may be evaluated
using the “ProbLib” APL workspace on the Macintosh, and used to calculate the
expected # of failures in the table above.

F(t) = 1− e ( ) 
− tuk

Soln: pi = F(ti ) − F(ti−1 ) ∀ i ≥ 1 , where   . Thus,

∴ p1 = F(t1 ) − F(t0 ) = F(50) - F(0) = 1 − e (

−( 50 u )
k

) (
− 1−e −( 0 u )
k

) = 0.0034
∴ E1 = 500p1 .= 1.7
p2 = F(t2 ) − F(t1 ) = F(100) - F(50) = 1 − e (
−(100 u )
) (
− 1 − e −( 50 u )
k k

) = 0.0144
∴ E1 = 500p1 .= 7.2
and so on ...

57:022 HW#1 2/1/97 page 26

Solutions

Following this procedure, we may obtain the following table:

Oi - Ei 2
Interval Pi Oi Ei Ei
0 - 50 0.0034 2 1.7 0.05294118
50-100 0.0144 4 7.2 1.42222222
100-150 0.0286 15 14.3 0.03426573
150-200 0.0440 26 22.0 0.72727273
200-250 0.0591 32 29.550.20313029
250-280 0.0420 21 21.0 0

D = 2.43983215

i. Perform a Chi-Square Goodness of Fit test to determine whether this distribution

provides an acceptable fit to the observed data, with a = 5%.
Soln: Since (from above) " D ( 2.4398) < χ 25% (= 7.815) ", we can conclude that
this distribution provides an acceptable fit to the observed data with α =5%.

j. According to these estimates of u & k, at what time should

... 5% of the units have failed? ___156.77__ hours
... 50% of the units? ___458.85__ hours
... 90% of the units? ___752.87__ hours

Soln: Exact estimates require solving the following equation for t:

k
tu
F(t) = 1 - e- where F(t) is 5%, 50%, and 90%, respectively. The inverse of
1
F(t) was derived in the notes: F-1 (p) = u - ln 1 -p k , where k = 2.4244, u=
533.7353
Thus we obtain
1
F-1 (0.05) = 533.7353 - ln 0.95 2.4244 = 156.77 hours
1
F-1 (0.5) = 533.7353 - ln 0.5 2.4244 = 458.85 hours
1
F-1 (0.9) = 533.7353 - ln 0.1 2.4244 = 752.87 hours

k. The manufacturer wishes to state a warranty period such that 99% of the
units produced are expected to survive the warranty period. According to these
estimates of u & k, what should be the length of the warranty period ?
-1 1
Soln: Again, we solve F (p) = u - ln 1 -p k where u & k are as above, and
p = 1%. This gives us a warranty period of 80 days (more exactly, 80.0352 days)

l. According to the value of k, is the failure rate increasing or decreasing with

time? ________________
Soln: k is greater than 1, which indicates that the failure rate is increasing
with time.

◊◊◊◊◊◊◊◊◊◊◊◊◊◊◊◊◊◊◊◊◊◊◊◊◊◊◊

Use the Curve Fit 0.7e software to try fitting a Weibull distribution in its original
(nonlinear) form. If x = upper limit of the intervals and y = fraction failed, then you
should enter the function f(x) = 1 - 2.718281828^(-(x/a)^b)
where a=u and b=k. (Note that Curve Fit allows only variables a, b, c, d, & e.)

57:022 HW#1 2/1/97 page 27

Solutions

m. Using as starting values for a & b the values which you found in (f), try fitting
the curve to the data. What are the values of a & b?
a = _______________ +/- __________________
b = _______________ +/- __________________
Was Curve Fit able to find values with small error estimates (e.g. with error less
than about 5 or 10% of the parameter)? ________
If yes, proceed to question (n); otherwise, stop.
Soln: According to the Curve Fit output, the parameters are
a = 533.73486 +/- 23.74058
b= 2.35061 +/- 0.13497
Yes, the Curve fit can find values with small error estimates.

2.00000E-01

1.60000E-01

1.20000E-01

8.00000E-02

4.00000E-02

1.00000E+02 2.00000E+02
Corr. Coeff. (R2) = 0.99607
Sum of Squares = 0.00013

n. What are the values of the parameters for this new fitted distribution?
u = 534 , k = 2.35

o. According to these values of u & k, what is

... the expected lifetime of the device? 473.2137
... the standard deviation of the device’s lifetime? 214.0274
... the probability that the device will fail during the first 400 hours? 0.3978
Soln: See solution of (g) above.

p. Using these new parameters u & k, compute the expected number of failures (of
the 100 units tested) in each of the intervals:

Observed Expected
Interval # failures # failures
0- 50 ________ _______
50-100 ________ _______
100-150 ________ _______

57:022 HW#1 2/1/97 page 28

Solutions

150-200 ________ _______

200-250 ________ _______
250-280 ________ _______


F(t) = 1− e ( ) 
− tuk

Soln: pi = F(ti ) − F(ti−1 ) ∀ i ≥ 1 , where   . Thus,

∴ p1 = F(t1 ) − F(t0 ) = F(50) - F(0) = 1 − e(−( 50 u )

k

) (
− 1 − e −( 0 u )
k

)
= 0.0038
∴ E1 = 500p1 .= 1.90
= F(t ) − F(t ) = F(100) - F(50) = (1 − e ) − (1 − e ) = 0.0155
−(100 u ) −( 50 u )
k k

p2 2 1
∴ E1 = 500p1 = 7.75
and so on ...

Following this procedure, we may obtain the following table:

Oi - Ei 2
Interval Pi Oi Ei Ei
0 - 50 0.0038 2 1.90 0.00526316
50-100 0.0155 4 7.75 1.81451613
100-150 0.0300 15 15.00 0
150-200 0.0454 26 22.70 0.47973568
200-250 0.0600 32 30.00 0.13333333
250-280 0.0423 21 21.15 0.00106383

D = 2.43391213

q. According to these new estimates of u & k, at what time should

... 5% of the units have failed? 156.84 days
... 50% of the units? 459.09 days
... 90% of the units? 753.26 days
Soln: See solution of (j) above.

◊◊◊◊◊◊◊◊◊◊◊◊◊◊◊◊◊◊◊◊◊◊◊◊◊◊◊

r. You needn’t perform a “goodness-of-fit” test to confirm it, but state your
opinion about which Weibull distribution seems to be a better fit.
Soln: I f we were to perform the “goodness-of-fit” test, then since " D ( =2.4339)

< χ 5% (= 7.815) ", we would conclude that this distribution also provides an
2

acceptable fit to the observed data with α =5%. The results are so similar that
there is probably no significant difference in the qualities of the fits. If each
program found the exact optimum of its objective function, then clearly the
result of Curve Fit would be a better fit. However, because Curve Fit is doing
computation with nonlinear functions while Cricket Graph is using linear
functions, the latter will perform the computations more accurately.

•••••••••••••••••••••••••••• Homework # 7 ••••••••••••••••••••••••••••••••

1. A system contains 4 types of devices, with the system reliability represented
schematically by

57:022 HW#1 2/1/97 page 29

Solutions

It has been estimated that the lifetime probability distributions of the devices are as
follows:
A: Weibull, with mean 1000 days and standard deviation 1200 days
B: Exponential, with mean 4000 days
C: Normal, with mean 3000 days and standard deviation 750 days
D: Exponential, with mean 1000 days

a.) Compute the reliabity of a unit of each device for a designed system lifetime of 1000
days:
Sol’n:
Device Reliability R(1000)
 1000  0.8376
− 
A 33.91% RA(1000) = 1 − FA (1000) = e  910.6269 

1000
−
B 77.88% RB(1000) = 1 − FB (1000) = e 4000

C 99.62% From CDF table of Normal Distribution

1000
−
D 36.79% RD (1000) = 1− FD (1000) = e 1000

b.) Using the reliabilities in (a), compute the system reliability:

Sol’n:
Subsystem Reliability R(1000)
0.5632 % Since 1 − (1− R A ) = 1 − (1− 0.3391)
2 2
AA
AA+B+C 0.4370 % R AA × R B × R C
B+DDD 0.5821 % {
R B × R DDD = 0.7788 × 1 − ( 1− 0.3679)
3
}
Total system: 0.7647 % 1 − (1− R AA × R B × R C )(1− RB × RDDD )
c.) Draw a SLAM network which can simulate the lifetime of this system. (You need not
perform the simulation.)

57:022 HW#1 2/1/97 page 30

Solutions

Notes:
• The CREATE node indicates that only 1 entity is initially created.
• “A”, “B”, “C”, and “D” above should be replaced by the lifetime distribution of the
corresponding devices, i.e., WEIBL( , ), EXPON(4000), RNORM(3000,750), and EXPON(1000),
respectively.
• The parameters of the Weibull distribution as defined by SLAM are named ALPHA and
BETA, and are not (µ,σ), the mean & standard deviation. ALPHA is identical to the
parameter k in the class notes, and BETA is uk . To determine u and k from the mean and
standard deviation, one may roughly estimate k given the coefficient of variation σ/µ
=1.2 from the table below (taken from the class notes):

The value of k is between 0.8 and 0.9; interpolating would give k=0.8 + .1(0.06/0.15) =
0.84. Next, the parameter u may be found from
µ = u Γ 1+ 1 ⇒ u = µ = 1000 ≈ 1000 = 909.09
k Γ 1+ 1 Γ 1+ 1 1.1
k 0.84
where the Gamma function is evaluated by interpolating in the table below (also found
in the notes):

57:022 HW#1 2/1/97 page 31

Solutions

Thus, “A” in the diagram above should be WEIBL(0.84, 909).

• The numbers in the ACCUMULATE nodes are: FR = number of arriving entities required
to cause the first entity to be “released” from the node; SR = number of arriving
entities required for subsequent entities to be “released” from the node.
• In the ACCUMULATE nodes N1, N3, and N5, the value of SR is irrelevant, since FR is
assigned a value equal to the total number of arrivals at the node, and no subsequent
entities will ever be released from the node. (The default value will be 1.)
• In the ACCUMULATE nodes N2 and N4, additional entities may arrive after the first entity
is released from the node, and therefore a value of SR (larger than the maximum
number of entities which might arrive) must be specified. In the diagram above, 9 was
specified, but any larger and some smaller values are possible.
• The COLCT (COLLECT) node will collect statistics on the time of the FIRST arrival; in this
case, the “1” on the TERMINATE node cause the simulation to then end, and so no
further arrivals can occur. For this reason, it would be equivalent to specify INTVL(1)
instead of FIRST if you also indicate on the CREATE statement that the creation time (0 in
this case) is to be recorded in attribute #1 (i.e., the parameter MA = 1).

2. A system has 3 types of components (A, B, & C) which are subject to failure. As shown
below, the system requires at least one of components A and B, and at least two of
component C. One of component A is in “stand-by”. That is, when the first component A
has failed, the second is to be switched on, and until then does not begin to "age" or fail.
Assume that the sensor/switch has 95% reliability. In the case of components B & C, on
the other hand, all units of these components are in operation simultaneously, so that
each unit is subject to failure from the beginning of the system’s operation. Note that the
system requires at least 2 of component C for proper functioning.

The lifetime distributions of the three component types are:

Component A: Erlang, being the sum of five random variables, each
having exponential distribution with mean 100 days.

57:022 HW#1 2/1/97 page 32

Solutions

Component B: Exponential, with expected lifetime 200 days.

Component C: Exponential, with expected lifetime 300 days.

a. Draw a SLAM network which can simulate the lifetime of this system.
Sol’n:

Notes:
• Again, the values “A”, “B”, and “C” on the diagram above should be replaced by
the lifetime distributions, namely ERLNG(100,5), EXPON(200), and EXPON(300),
respectively.
• The parameters of the Erlang distribution in SLAM are not the mean and
standard deviation (which would be in this case 500 and
2
5×100 = 100 5 = 223.61), but the mean of each of the component exponential
distributions and the number of such distributions.
• The “1” in the GOON node (labelled N1) indicates that the entity which arrives
there (at the time of the failure of the primary unit of device A) must leave
node N1 via a single activity only.
• The two activities leaving the GOON node N1 represent (1) the lifetime of the
standby unit of device A, which receives the entity with 99% probability, and
(2) the failure of the switch, requiring time = 0, which receives the entity with
probability 1%.
• See the note in problem 1 about the values of the parameters FR and SR of the
ACCUMULATE nodes.
• See the note in problem 1 about the FIRST statistics collected by the COLCT node.

b. Enter the network into the computer, and simulate the system 1000 times,
collecting statistics on the time of system failure. Request that a histogram be
printed. Specify about 15-20 cells, with HLOW and HWID parameters which will give
you a "nice" histogram with the mean approximately in the center and with small
tails. (This may require a second simulation, with the histogram parameters selected
after observing the results of a preliminary simulation.)

Note: Be sure to specify on the GEN statement that you do NOT want intermediate
results, and that the SUMMARY report is to be printed only after the 1000th run. Also
specify on the INITIALIZE statement that the statistical arrays should not be cleared
between runs. The following should work:
GEN,yourname,RELIABILITY,10/16/96,1000,,N,,N,Y/1000,72;
LIM,,,8;
INIT,,,NO;
NETWORK;

57:022 HW#1 2/1/97 page 33

Solutions

(SLAM II network statements go here)

END;
FIN;

Sol’n: The (edited) SLAM output appears below:

1 GEN,Hansuk Sohn,Reliability,10/9/1996,1000,,N,,N,Y/1000,72;
2 LIM,,,8;
3 INIT,,,NO;
4 NETWORK;
5 CREATE;
6 ACT/1,ERLNG(100,5),,Na; Component A
7 ACT/2,EXPON(200),,Nb; Component B
8 ACT/3,EXPON(200),,Nb; Component B
9 ACT/4,EXPON(300),,Nc; Component C
10 ACT/5,EXPON(300),,Nc; Component C
11 ACT/6,EXPON(300),,Nc; Component C
12 Na GOON,1;
13 ACT,ERLNG(100,5),0.95,FAIL;
14 ACT,0,0.05,FAIL;
15 Nb ACCUM,2; Node "b" will be released when both Bs fail.
16 ACT,,,FAIL;
17 Nc ACCUM,2; Node "c" will be released when 2 out of 3 Cs fail.
18 ACT,,,FAIL;
19 FAIL COLCT,FIRST,TIME_OF_FAILURE,20/0/20;
20 TERM,1;
21 END;
22 FIN;
S L A M I I S U M M A R Y R E P O R T

SIMULATION PROJECT RELIABILITY BY HANSUK SOHN

DATE 10/ 9/1996 RUN NUMBER 1000 OF 1000

CURRENT TIME .1267E+03

STATISTICAL ARRAYS CLEARED AT TIME .0000E+00

**STATISTICS FOR VARIABLES BASED ON OBSERVATION**

MEAN STANDARD COEFF. OF MINIMUM MAXIMUM NO.OF
VALUE DEVIATION VARIATION VALUE VALUE OBS
TIME_OF_FAILURE .174E+03 .116E+03 .669E+00 .224E+01 .666E+03 1000

**REGULAR ACTIVITY STATISTICS**

ACTIVITY AVERAGE STANDARD MAXIMUM CURRENT ENTITY
INDEX/LABEL UTILIZATION DEVIATION UTIL UTIL COUNT
1 COMPONENT A .1315 .3380 1 1 0
2 COMPONENT B .4178 .4932 1 1 0
3 COMPONENT B .3887 .4874 1 0 1
4 COMPONENT C .3057 .4607 1 1 0
5 COMPONENT C .3012 .4588 1 0 1
6 COMPONENT C .3357 .4722 1 0 1

**HISTOGRAM NUMBER 1**

TIME_OF_FAILURE

OBS RELA UPPER

FREQ FREQ CELL LIM 0 20 40 60 80 100
+ + + + + + + + + + +
0 .000 .000E+00 + +
22 .022 .200E+02 +* +

57:022 HW#1 2/1/97 page 34

Solutions

53 .053 .400E+02 +***C +

73 .073 .600E+02 +**** C +
75 .075 .800E+02 +**** C +
85 .085 .100E+03 +**** C +
84 .084 .120E+03 +**** C +
77 .077 .140E+03 +**** C +
70 .070 .160E+03 +**** C +
62 .062 .180E+03 +*** C +
58 .058 .200E+03 +*** C +
41 .041 .220E+03 +** C +
53 .053 .240E+03 +*** C +
51 .051 .260E+03 +*** C +
35 .035 .280E+03 +** C +
25 .025 .300E+03 +* C +
24 .024 .320E+03 +* C +
22 .022 .340E+03 +* C +
15 .015 .360E+03 +* C +
12 .012 .380E+03 +* C +
11 .011 .400E+03 +* C +
52 .052 INF +*** C
--- + + + + + + + + + + +
*** 0 20 40 60 80 100

**STATISTICS FOR VARIABLES BASED ON OBSERVATION**

MEAN STANDARD COEFF. OF MINIMUM MAXIMUM NO.OF

VALUE DEVIATION VARIATION VALUE VALUE OBS
TIME_OF_FAILURE .174E+03 .116E+03 .669E+00 .224E+01 .666E+03 1000

c. What is your estimate of the average lifetime of this system, based upon the simulation
results?
Sol’n: 174 days ( from SLAM output )

d. Suppose that the system is required to survive for a 100-day mission. What is the
estimated reliability of the system, i.e., the probability that the system survives 100 days?
Sol’n: From the histogram above, we see that 308 (the sum of the first six cells) failures
occurred during the first 100 days (and 1000-308=692 survived at least 100 days). Hence,
308
R(100) = 1- = 0.692
1000
e. Suppose your company will offer a warranty on this system, specifying the length of
the warranty period such that 98% of the systems will survive past the warranty period.
What should be the length of the warranty?

Sol’n: The above histogram doesn’t provide sufficient detail at the low end to allow an
accurate estimate, but if we use a linear interpolation we would estimate that the 20th
failure occurred at time 20(20/22) = 18.2 days, leaving at that time 980 survivors (98% of
the systems simulated.) Therefore, according to this estimate, 18.2 days should be the
length of the warranty (which undoubtedly would be rounded to 18 days for the sake of
convenience). Interpolating again to find the reliability of the system for an 18-day
lifetime, we would estimate that 22(18/20) = 19.8 systems, or 1.98% of the total, failed
during the first 18 days, leaving 98.02% surviving.

To get a more accurate estimate of the correct warranty period, we should

• increase the number of runs from 1000, e.g., to as many as 10000
• change the specifications for the histogram so as to get a count of failures each
day, e.g.,

57:022 HW#1 2/1/97 page 35

Solutions

FAIL COLCT,FIRST,TIME_OF_FAILURE,30/0/1;
which would give 30 cells, each of length 1 day.

•••••••••••••••••••••••••••• Homework # 8 ••••••••••••••••••••••••••••••••

1. Project Scheduling. Consider the home-building project:
Predecessor Duration (days)
Activity Description Activities Mean Std Dev
A Walls & ceiling B 5 2
B Foundation none 4 1
C Roof timbers A 2 1
D Roof sheathing C 2 1
E Electrical wiring A 4 2
F Roof shingles D 2 1
G Exterior siding H 5 1
H Windows A 4 1
I Paint F,G,J 3 1
J Inside wall board E,H 3 1

1. Complete the AON network by labeling the nodes:

J I

H G
2. Complete the AOA & the corresponding SLAM networks below by inserting any
"dummy" activities which are necessary, and labeling the nodes.

3. Give numerical values (0, 1, 2, 3, 4, or ∞ ) for parameters "a" - "i" and a statistic
type for “j” on the SLAM network below which would simulate the project and
collect statistics on the completion time. (You needn’t insert the probability
distributions for the durations!)

57:022 HW#1 2/1/97 page 36

Solutions

a b c d e f g h i j
1 1 1 3 2 1 3 1 1 FIRST

5. Complete the ETs (earliest times) & LTs (latest times) in the network below, using
the expected activity durations, as indicated. Don't forget any "dummy" activities
which you entered above!

11 16

9 14 15 21

14
6. What are the critical activities?
Soln : B - A - H - G - I

7. What is the "total slack" or "total float" in activity D? 3 days

8. What is the expected completion time of the project? 21 days

...the standard deviation of the project’s completion time? 2.8284 days

Activity i j µ σ2
B 0 1 4 1
A 1 2 5 22
H 2 5 4 1
G 5 7 5 1
I 7 8 3 1
sum 21 8
σT = 8 = 2.8284

57:022 HW#1 2/1/97 page 37

Solutions

2. Building a Hydroelectric Power Station. Below is a rough breakdown of a

building project into tasks or activities. (In practice, each of the tasks below, e.g. #11
“Dam building”, would itself be divided into many individual jobs.) The task
descriptions and estimated durations (in days) are:
Activity Description Predecessor(s) Duration
A Ecological survey of dam site none 137
B File environmental impact report A 201
and get EPA approval
C Economic feasibility study A 161
D Preliminary design & cost estimation C 93
E Project approval & commitment of funds B,D 225
F Call quotations for electrical E 95
equipment (turbines, generators, etc.)
G Select suppliers for electrical equipment F 69
H Final design of project E 143
I Select construction contractors E 60
J Arrange construction mat’ls supply H,I 115
K Dam building J 546
L Power station building J 405
M Power lines erection G,H 447
N Electrical equipment installation G,L 150
O Build up reservoir water level K 47
P Commission the generators N,O 27
Q Start supplying power M,P 25

a. Draw the AON (activity-on-node) network representing this project.

begin end

F G M
B
Q

A E

L
P

C D I K O

b. Draw the AOA (activity-on-arrow) network representing this project. Explain

the necessity for any "dummy" activities which you have included.

57:022 HW#1 2/1/97 page 38

Solutions

G M Q
F
P

E H N
B

D L O
A

I
C K
J

c. Label the nodes of the AOA network, so that i<j if there is an activity with node i
as its start and node j as its end node.

G M Q
5 8 10 14 15
F
P

E H N
2 4 6 11 13
B

D L O
A
0 1

I
C K
J
3 7 9 12
d. Perform the forward pass through the AOA network to obtain for each node i,
ET(i) = earliest possible time for event i.

57:022 HW#1 2/1/97 page 39

Solutions

711 780 780 1494 1519

M Q
5 G 8 10 14 15
F
P
391 616 759 1279
E H N
2 4 6 11 13
B 1467
D L O
0 137
A
0 1

I
C K
J
3 7 9 12
298 759 874 1420
e. What is the earliest completion time for this project? 1519 days

f. Perform the backward pass through the AOA network to obtain, for each node i,
LT(i) = latest possible time for event i (assuming the project is to be completed in
the time which you have specified in (e).)
978 1047 1047 1494 1519
M Q
5 G 8 10 14 15
F
P
391 616 759 1317 1467
E H N
2 4 6 11 13
B

0 137 D L O
A
0 1

I
C K
J
3 7 9 12
298 759 874 1420
g. For each activity, compute and record below:
ES = earliest start time
LS = latest start time
EF = earliest finish time
LF = latest finish time
TF = total float (slack)
Sol’n:

57:022 HW#1 2/1/97 page 40

Solutions

Critical Activity Activity ES LS EF LF TF

Yes A 0 0 137 137 0
No B 137 190 338 391 53
Yes C 137 137 298 298 0
Yes D 298 298 391 391 0
Yes E 391 391 616 616 0
No F 616 883 711 978 267
No G 711 978 780 1047 267
Yes H 616 616 759 759 0
No I 616 699 676 759 83
Yes J 759 759 874 874 0
Yes K 874 874 1420 1420 0
No L 874 912 1279 1317 38
No M 780 1047 1227 1494 267
No N 1279 1317 1429 1467 38
Yes O 1420 1420 1467 1467 0
Yes P 1467 1467 1494 1494 0
Yes Q 1494 1494 1519 1519 0

h. Which activities are "critical", i.e., have zero float? (Circle the labels A-Q of the
critical activities above.) Indicate the critical path in your AOA network in part
(b).
Soln : A - C - D - E - H - J - K - O - P - Q
M Q
5 G 8 10 14 15
F
P

E H N
2 4 6 11 13
B

D L O
A
0 1

I
C K
J
3 7 9 12
i. Schedule this project by entering the AON network into the MacProject PRO
software Specify that the start time for the project will be November 1, 1996.
What is the earliest completion date for the project? August 28, 2002
(Note that
• stated to the left and right of each node, i.e., acitivity, are the earliest start and
latest finish dates, respectively.
• MacProject PRO assumes 5-day work weeks by default, so that the time between
beginning & completion date isn’t the same as your answer in (e).)

ES EF
ES : Earliest Start Time
EF : Earliest Finish Time

57:022 HW#1 2/1/97 page 41

Solutions

5/13/97 2/17/98 7/26/99 10/28/99 7/25/02 8/28/02

B G Q

10/29/99 7/16/01

3/15/99 7/23/99 M

F
11/1/96 5/12/97

A
9/27/01 4/24/02

N
3/15/99 9/29/99

3/9/00 9/26/01

5/4/98 3/12/99 L

9/30/99 3/8/00
5/13/97 12/23/97 6/18/02 7/24/02
J
C P

3/15/99 6/4/99 4/12/02 6/17/02

I O

12/24/97 5/1/98 3/9/00 4/11/02

D K

•••••••••••••••••••••••••••• Homework # 9 ••••••••••••••••••••••••••••••••

Building a Hydroelectric Power Station. Below is a rough breakdown of a
building project into tasks or activities. (In practice, each of the tasks below, e.g. K
(#11) “Dam building”, would itself be divided into many individual jobs.) The task
descriptions and estimated durations (in days) are:

Activity Description Predecessor(s) Duration

A Ecological survey of dam site none 137 ± 12
B File environmental impact report A 201 ± 18
and get EPA approval
C Economic feasibility study A 161 ± 14
D Preliminary design & cost estimation C 93 ± 5

57:022 HW#1 2/1/97 page 42

Solutions

E Project approval & commitment of funds B,D 225 ± 20

F Call quotations for electrical E 95 ± 10
equipment (turbines, generators, etc.)
G Select suppliers for electrical equipment F 69 ± 5
H Final design of project E 143 ± 12
I Select construction contractors E 60 ± 5
J Arrange construction mat’ls supply H,I 115 ± 15
K Dam building J 546 ± 30
L Power station building J 405 ± 25
M Power lines erection G,H 447 ± 15
N Electrical equipment installation G,L 150 ± 10
O Build up reservoir water level K 47 ± 15
P Commission the generators N,O 27 ± 2
Q Start supplying power M,P 25 ± 3

(This is the same data as given in Homework #8, except that the durations are now
given as intervals centered around the originally-given duration.)

The AOA network representing this project is shown on the next page.

a. Use SLAM to simulate the project 1000 times, with a triangular distribution for the
activity durations, e.g., TRIAG(125,137,149) for activity A. (Note that the mean
value for each duration is the central value which was used in Homework #8, e.g.,
137 days for activity A.)

Warning! In the case of a “dummy” activity with duration 0, or any other

constant, don’t specify a probability distribution such as TRIAG(0,0,0), which
would lead to an error message!

57:022 HW#1 2/1/97 page 43

Solutions

Solutions :
l-ecn015% rslam 1a-1

1 GEN,Hansuk Sohn,Hydroelectric Power Station,11/5/1996,1000,,N,,N,Y/1000,72;

2 LIM,20,50,500;
3 INIT,,,NO;
4 NETWORK;
5 CREATE,,,,1;
6 ACT/1,TRIAG(125,137,149),,N1; Activity A +/- 12
7 N1 ACCUM,,,,2;
8 ACT/2,TRIAG(147,161,175),,N2; Activity C +/- 14
9 ACT/3,TRIAG(183,201,219),,N3; Activity B +/- 18
10 N2 GOON;
11 ACT/4,TRIAG(88,93,98),,N3; Activity D +/- 5
12 N3 ACCUM,2;
13 ACT/5,TRIAG(205,225,245),,N4; Activity E +/- 20
14 N4 ACCUM,,,,3;
15 ACT/6,TRIAG(85,95,105),,N5; Activity F +/- 10
16 ACT/7,TRIAG(131,143,155),,N6; Activity H +/- 12
17 ACT/8,TRIAG(55,60,65),,N7; Activity I +/- 5
18 N5 GOON;
19 ACT/9,TRIAG(64,69,74),,N8; Activity G +/- 5
20 N6 ACCUM,,,,2;
21 ACT/10,,,N7; DummyAct 1
22 ACT/11,,,N10; DummyAct 2
23 N7 ACCUM,2;
24 ACT/12,TRIAG(100,115,130),,N9; Activity J +/- 15
25 N8 GOON,2;
26 ACT/13,,,N10; DummyAct 3
27 ACT/14,,,N11; DummyAct 4
28 N9 GOON,2;
29 ACT/15,TRIAG(516,546,576),,N12; Activity K +/- 30
30 ACT/16,TRIAG(380,405,430),,N11; Activity L +/- 25
31 N10 ACCUM,2;
32 ACT/17,TRIAG(432,447,462),,N14; Activity M +/- 15
33 N11 ACCUM,2;
34 ACT/18,TRIAG(140,150,160),,N13; Activity N +/- 10
35 N12 GOON;
36 ACT/19,TRIAG(32,47,62),,N13; Activity O +/- 15
37 N13 ACCUM,2;
38 ACT/20,TRIAG(25,27,29),,N14; Activity P +/- 2
39 N14 ACCUM,2;
40 ACT/21,TRIAG(22,25,28),,EFT; Activity Q +/- 3
41 EFT COLCT,FIRST,Earliest_Finish_Time,50/1450/3
42 TERM,1;
43 END;
44 FIN;

S L A M I I S U M M A R Y R E P O R T

SIMULATION PROJECT HYDROELECTRIC POWER BY HANSUK SOHN

DATE 11/ 5/1996 RUN NUMBER 1000 OF 1000

CURRENT TIME .1488E+04

STATISTICAL ARRAYS CLEARED AT TIME .0000E+00

**STATISTICS FOR VARIABLES BASED ON OBSERVATION**

57:022 HW#1 2/1/97 page 44

Solutions

MEAN STANDARD COEFF. OF MINIMUM MAXIMUM NO.OF

VALUE DEVIATION VARIATION VALUE VALUE OBS

EARLIEST_FINISH_ .152E+04 .190E+02 .125E-01 .147E+04 .157E+04 1000

**REGULAR ACTIVITY STATISTICS**

ACTIVITY AVERAGE STANDARD MAXIMUM CURRENT ENTITY

INDEX/LABEL UTILIZATION DEVIATION UTIL UTIL COUNT

1 ACTIVITY A .0903 .2866 1 0 1

2 ACTIVITY C .1062 .3080 1 0 1
3 ACTIVITY B .1323 .3388 1 0 1
4 ACTIVITY D .0611 .2395 1 0 1
5 ACTIVITY E .1481 .3552 1 0 1
6 ACTIVITY F .0625 .2420 1 0 1
7 ACTIVITY H .0940 .2919 1 0 1
8 ACTIVITY I .0395 .1948 1 0 1
9 ACTIVITY G .0454 .2081 1 0 1
10 DUMMYACT 1 .0000 .0000 1 0 1
11 DUMMYACT 2 .0000 .0000 1 0 1
12 ACTIVITY J .0756 .2644 1 0 1
13 DUMMYACT 3 .0000 .0000 1 0 1
14 DUMMYACT 4 .0000 .0000 1 0 1
15 ACTIVITY K .3596 .4799 1 0 1
16 ACTIVITY L .2665 .4421 1 0 1
17 ACTIVITY M .2940 .4556 1 0 1
18 ACTIVITY N .0986 .2981 1 0 1
19 ACTIVITY O .0309 .1732 1 0 1
20 ACTIVITY P .0178 .1321 1 0 1
21 ACTIVITY Q .0165 .1272 1 0 1

**HISTOGRAM NUMBER 1**

EARLIEST_FINISH_

OBS RELA UPPER

FREQ FREQ CELL LIM 0 20 40 60 80 100
+ + + + + + + + + + +
0 .000 .1455+04 + +
0 .000 .1459+04 + +
0 .000 .1462+04 + +
0 .000 .1465+04 + +
0 .000 .1468+04 + +
0 .000 .1471+04 + +
0 .000 .1474+04 + +
3 .003 .1477+04 + +
1 .001 .1470+04 + +
3 .003 .1483+04 + +
1 .001 .1486+04 + +
6 .006 .1489+04 +C +
9 .009 .1492+04 +C +
15 .015 .1495+04 +*C +
24 .024 .1498+04 +* C +
38 .038 .1501+04 +** C +
40 .040 .1504+04 +** C +
36 .036 .1507+04 +** C +
37 .037 .1500+04 +** C +
49 .049 .1513+04 +** C +
57 .057 .1516+04 +*** C +

57:022 HW#1 2/1/97 page 45

Solutions

57 .057 .1519+04 +*** C +

47 .047 .1522+04 +** C +
53 .053 .1525+04 +*** C +
78 .078 .1528+04 +**** C +
60 .060 .1531+04 +*** C +
35 .035 .1534+04 +** C +
48 .048 .1537+04 +** C +
55 .055 .1530+04 +*** C +
48 .048 .1543+04 +** C +
39 .039 .1546+04 +** C +
36 .036 .1549+04 +** C +
29 .029 .1552+04 +* C +
31 .031 .155E+04 +** C +
16 .016 .155E+04 +* C +
16 .016 .156E+04 +* C +
13 .013 .156E+04 +* C+
5 .005 .156E+04 + C+
6 .006 .156E+04 + C
4 .004 .157E+04 + C
1 .001 .157E+04 + C
2 .002 .157E+04 + C
2 .002 .158E+04 + C
0 .000 .158E+04 + C
0 .000 .158E+04 + C
0 .000 .159E+04 + C
0 .000 .159E+04 + C
0 .000 .159E+04 + C
0 .000 .159E+04 + C
0 .000 .160E+04 + C
0 .000 .160E+04 + C
0 .000 INF + C
--- + + + + + + + + + + +
*** 0 20 40 60 80 100

b. In your simulation, what are the

• average completion time? __1520_ days
• standard deviation? _19_ days
• minimum completion time? _1470_ days
• maximum completion time? _1570_ days
(Recall that in Homework #8, the duration was determined to be 1519 days.)

c. According to PERT, the critical path which you found in Homework #8, assuming
the durations to be equal to their expected values, namely
[A - C - D - E - H - J - K - O - P - Q],
is always the critical path. Determine whether that is true for the 1000
simulations you have performed. (Because of the relatively large “float” of the
noncritical activities found in Homework #8, it might be true. If necessary,
modify your SLAM model to determine this!) If not true, in how many simulations
was this path not the critical path?

57:022 HW#1 2/1/97 page 46

Solutions

G M Q
5 8 10 14 15
F
P

E H N
2 4 6 11 13
B

D L O
A
0 1

I
C K
J
3 7 9 12

Soln: For the 1000 simulations performed 'A-C-D-E-H-J-K-O-P-Q' is “nearly always” the critical
path. This was determined by making the adjustments below to the previous SLAM II model.

l-ecn008% rslam 1c-ok

1 GEN,Hansuk Sohn,Hydroelectric Power Station,11/5/1996,1000,,N,,N,Y/1000,72;

2 LIM,20,50,500;
3 INIT,,,NO;
4 NETWORK;
5 CREATE,,,1,1;
6 ACT/1,137,,A1; Activity A
7 A1 ASSIGN,ATRIB(1)=1;
8 N1 ACCUM,,,LAST,2;
9 ACT/2,161,,N2; Activity C
10 ACT/3,201,,A3; Activity B
11 N2 GOON;
12 ACT/4,93,,N3; Activity D
13 A3 ASSIGN,ATRIB(1)=0;
14 N3 ACCUM,2,,LAST;
15 ACT/5,225,,N4; Activity E
16 N4 ACCUM,,,LAST,3;
17 ACT/6,95,,A5; Activity F
18 ACT/7,143,,N6; Activity H
19 ACT/8,60,,A7; Activity I
20 A5 ASSIGN,ATRIB(1)=0;
21 N5 GOON;
22 ACT/9,69,,N8; Activity G
23 N6 ACCUM,,,LAST,2;
24 ACT/10,,,N7; DummyAct 1
25 ACT/11,,,A10; DummyAct 2
26 A7 ASSIGN,ATRIB(1)=0;
27 N7 ACCUM,2,,LAST;
28 ACT/12,115,,N9; Activity J
29 N8 GOON,2;
30 ACT/13,,,A10; DummyAct 3
31 ACT/14,,,A11; DummyAct 4
32 N9 GOON,2;
33 ACT/15,546,,N12; Activity K
34 ACT/16,405,,A11; Activity L
35 A10 ASSIGN,ATRIB(1)=0;

57:022 HW#1 2/1/97 page 47

Solutions

36 N10 ACCUM,2,,LAST;
37 ACT/17,447,,N14; Activity M
38 A11 ASSIGN,ATRIB(1)=0;
39 N11 ACCUM,2,,LAST;
40 ACT/18,150,,N13; Activity N
41 N12 GOON;
42 ACT/19,47,,N13; Activity O
43 N13 ACCUM,2,,LAST;
44 ACT/20,27,,N14; Activity P
45 N14 ACCUM,2,,LAST;
46 ACT/21,25,,CHECK; Activity Q
47 CHECK COLCT,ATRIB(1),Number_Same_CP
48 EFT COLCT,FIRST,Earliest_Finish_Time
49 TERM,1;
50 END;
51 FIN;

S L A M I I S U M M A R Y R E P O R T

SIMULATION PROJECT HYDROELECTRIC POWER BY HANSUK SOHN

DATE 11/ 5/1996 RUN NUMBER 1000 OF 1000

CURRENT TIME .1519E+04

STATISTICAL ARRAYS CLEARED AT TIME .0000E+00

**STATISTICS FOR VARIABLES BASED ON OBSERVATION**

MEAN STANDARD COEFF. OF MINIMUM MAXIMUM NO.OF

VALUE DEVIATION VARIATION VALUE VALUE OBS

NUMBER_SAME_CP 100E+01 .000E+00 .000E+00 .100E+01 .100E+01 1000

EARLIEST_FINISH_ .152E+04 .000E+00 .000E+00 .152E+04 .152E+04 1000

A new COLCT node was inserted to collect statistics on the value of ATTRIBUTE #1,
which is equal to 1 if the path of interest [A - C - D - E - H - J - K - O - P - Q] is
critical, and zero otherwise. We see that the minimum value and maximum value of
attribute #1 are both equal to 1, indicating that the path was always the longest
path in the network. Warning: this isn’t true in general! In this particular
example, all noncritical activities had rather large values of slack (“float”), as
found in the previous homework assignment!

Based upon your simulation, if you wish to set a deadline for completion of the project
so that you are 90% certain of meeting this deadline, what should this deadline be?

Sol’n: See the histogram of the project completion time.

36 .036 .1549+04 +** C +
29 .029 .1552+04 +* C +

By summing the counts of the cells, we find that at time=1549 days, 875 (i.e., 87.5%) of
the simulated projects had been completed, while at time=1552 days, 904 (i.e. 90.4%) of
the projects had been completed. Using a linear interpolation, we estimate that 90% of
the simulated projects should be completed at time= 1549+ 3(25/29) = 1551.6 days.

To get a more accurate estimate of the correct warranty period, we should

• increase the number of runs from 1000, e.g., to as many as 10000

57:022 HW#1 2/1/97 page 48

Solutions

• change the specifications for the histogram so as to get a count of failures each
day, e.g.,
EFT COLCT,FIRST,Earliest_Finish_Time,50/1545/0.05
which would give 50 cells, each of length 0.05 days.

e. Suppose that, by using improved procedures, the variability in the duration of

activity J, “Arrange construction mat’ls supply”, which presently is ± 15 days, might
be reduced to ± 5 days. How would this effect the deadline which you set in (d)?

Sol’n: Modify the SLAM statement for activity J, i.e.,

24 ACT/12,TRIAG(110,115,120),,N9; Activity J +/- 5

and re-run the simulation. In the new histogram, we find that at time=1548 days 895
projects had been completed, and at time=1551, 918 projects had been completed:

38 .038 .1548+04 +** C +

23 .023 .1551+04 +* C +

Using linear interpolation as before, we estimate that 900 projects would be complete at
time = 1548 + 3(5/23) = 1548.7 days (2.9 days earlier than estimated in part (d).)

To get a more accurate estimate of the correct warranty period, we should

• increase the number of runs from 1000, e.g., to as many as 10000
• change the specifications for the histogram so as to get a count of failures each
day, e.g.,
EFT COLCT,FIRST,Earliest_Finish_Time,50/1543/0.1
which would give 50 cells, each of length 0.1 days.

f. According to PERT, the project duration should have approximately a normal

distribution. Use the chi-square goodness of fit test to determine whether this
assumption is valid for this project.

Solution: To reduce the computational burden somewhat, the below table was
computed with each interval of 6 days length, rather than the 3 days in the SLAM
histogram. Using normal distribution tables with mean µ = 1520 and standard
deviation σ = 19, we compute the probability pi that one of the 1000 observations
falls into cell #i, and then multiply this by 1000 to get Ei , the expected number of
observations in cell #i:

# Interval Oi Ei Pi Oi - Ei 2
Ei
1 1470 - 1475 2 4.682 0.004682 1.53633575
2 1475 - 1480 3 8.648 0.008648 3.68870305
3 1480 - 1485 3 14.980 0.014980 9.58080107
4 1485 - 1490 12 24.354 0.024354 6.26678640
5 1490 - 1495 25 37.068 0.037068 3.92890428
6 1495 - 1500 55 52.590 0.052590 0.11044115
7 1500 - 1505 59 69.306 0.069306 1.53253161
8 1505 - 1510 68 84.775 0.084775 3.31938219
9 1510 - 1515 92 96.525 0.096525 0.21212769
10 1515 - 1520 84 102.802 0.102802 3.43879695

57:022 HW#1 2/1/97 page 49

Solutions

11 1520 - 1525 99 102.802 0.102802 0.14061209

12 1525 - 1530 112 96.525 0.096525 2.48096996
13 1530 - 1535 66 84.775 0.084775 4.15807284
14 1535 - 1540 85 69.306 0.069306 3.55382847
15 1540 - 1545 74 52.590 0.052590 8.71625975
16 1545 - 1550 53 37.068 0.037068 6.84764821
17 1550 - 1555 50 24.354 0.024354 27.00654170
18 1555 - 1560 25 14.980 0.014980 6.70229640
19 1560 - 1565 17 8.648 0.008648 8.06613136
20 1565 - 1570 9 4.682 0.004682 3.98229902
21 1570 - 1575 3 2.368 0.002368 0.16867568
22 1575 - 1580 4 1.110 0.001110 7.52441441
sum = 1000 D= 112.96256000

Note : pi = F(ti ) − F(ti−1 ) ∀ i ≥ 1 , where t ~ N(1520,19) Thus, for example,

∴ p1 = F(t1 ) − F(t0) = F(1475) - F(1470) = 0.004682
∴ E1 = 1000× P 1 = 4.682
p1 = F(t1 ) − F(t0) = F(1480) - F(1475) = 0.008648
∴ E2 = 1000× P 2 = 8.648

We perform a Chi-Square Goodness of Fit test to determine whether this normal

distribution provides an acceptable fit to the observed data, with α = 5%.
2
Since " D ( 112.96) > χ 5% (= 30.15) ", we can conclude that this distribution does not
provide an acceptable fit to the observed data with α =5%.

•••••••••••••••••••••••••••• Homework # 10 ••••••••••••••••••••••••••••••••

1. Consider an inventory system in which the number of items on the shelf is
checked at the end of each day. The maximum number on the shelf is 8. If 3 or
fewer units are on the shelf, the shelf is refilled overnight. The demand
distribution is as follows:
x 0 1 2 3 4 5 6
P{D=x} .1 .15 .25 .25 .15 .05 .05

The system is modeled as a Markov chain, with the state defined as the number of
units on the shelf at the end of each day. The probability transition matrix is:

57:022 HW#1 2/1/97 page 50

Solutions

a. Explain the derivation of the values P19, P 35, P 51, P 83 above. (Note that
state 1=inventory level 0, etc.)

State Definition

1 SOH = 0
2 SOH = 1
3 SOH = 2
4 SOH = 3
5 SOH = 4
6 SOH = 5
7 SOH = 6
8 SOH = 7
9 SOH = 8

Solution: P ij = P{Xn = j|X n−1 = i}

if i > 4 (SOH >3), no replenishment occurs :
 P{D = (i − j)} for j > 1 (SOH > 0)
P ij = 
 P{D ≥ (i − j)} for j = 1 (SOH = 0)
∴ P 83 = P{D = (8-3) =5 } = 0.05
P 51 = P{D ≥ (5-1)= 4} = 0.15 + 0.05 + 0.05 = 0.25

if i ≤ 5 (SOH ≤ 4), the SOH at the beginning of the next day is 8 :

P ij = P{D = (8-(j-1) )}
∴ P19 = P{D = (8-(9-1) )} = P{D = 0} = 0.1
P 35 = P{D = (8-(5-1) )} = P{D = 4} = 0.15

The steady-state distribution of the above Markov chain is:

57:022 HW#1 2/1/97 page 51

Solutions

b. Write two of the equations which define this steady-state distribution. How
many equations must be solved to yield the solution above?

Solution: 9 equations must be solved to compute the values of the 9

probabilities. One equation must be
π 1 + π 2 + π 3 + π 4 + π5 + π6 + π 7 + π 8 + π 9 = 1
while the remaining 8 are obtained by setting π i equal to the inner product of
column i and π . For example, using column i=1 we would obtain the equation
π 1 = 0.25π 5 + 0.1π 6 + 0.05π 7

c. What is the average number on the shelf at the end of each day?

Solution: Since in state i the stock on hand is (i-1),

9
Average Stock-on-Hand = ∑ (i − 1)π i = 4.968123
i =1
The mean first passage matrix is:

d. If the shelf is full Monday morning, what is the expected number of days
until the shelf is first emptied ("stockout")?
Solution: m 91 = 15.4523 (days)
e. What is the expected time between stockouts?
Solution: m 00 = 15.4523 (days)
f. How frequently will the shelf be restocked? (i.e. what is the average
number of days between restocking?)

57:022 HW#1 2/1/97 page 52

Solutions

Solution: Since the shelf is restocked whenever the state of the system is
1,2,3, or 4, the steadystate probability that the shelf is restocked on a day is
4

∑ π = 0.4076 = 1/2.4534,
i=1
i.e., we expect the shelf to restocked with frequency once each 2.4534 days.

2. Consider a manufacturing process in which raw parts (blanks) are machined on

three machines, and inspected after each machining operation. The relevant data
is as follows:

For example, machine #1 requires 0.5 hrs, at $20/hr., and has a 10% scrap rate. Those
parts completing this operation are inspected, requiring 0.1 hr. at $15/hr. The
inspector scraps 10%, and sends 5% back to machine #1 for rework (after which it
is again inspected, etc.)
The Markov chain model of a part moving through this system has transition
probability matrix:

a. Draw the diagram for this Markov chain and describe each state.
Solution:

57:022 HW#1 2/1/97 page 53

Solutions

0.05 0.03 0.02

1
State Location
0.9 0.85 0.95 0.87 0.98 0.93
1 2 3 4 5 6 7 1 Machine1
2 Inspection station1
3 Machine2
0.1 0.1 0.05 0.1 0.02 0.05 4 Inspection station2
5 Machine3
8 6 Inspection station3
7 Pack-&-Ship Dept.
1 8 Scrap bin

b. Which states are transient? which are absorbing?

Solution: Transient States : 1, 2, 3, 4, 5, and 6
Absorbing States : 7 and 8

The absorption probabilities are:

The matrix E is as follows:

c. Explain how matrix E was computed. Explain how matrix A was computed,
given E.
Solution:

0 0.9 0 0 0 0 0 0.1
0.05 0 0.85 0 0 0 0 0.1
0 0 0 0.95 0 0 0 0.05 Q R
0 0 0.03 0 0.87 0 0 0.1 =
0 0 0 0 0 0.98 0 0.02 0 I
0 0 0 0 0.02 0 0.93 0.05
0 0 0 0 0 0 1 0
0 0 0 0 0 0 0 1

matrices E and A could be computed by equations E = ( I − Q) and A=E× R,

−1

respectively.

57:022 HW#1 2/1/97 page 54

Solutions

d. What percent of the parts which are started are successfully completed?
Solution: A1,OK = 0.6335 = 63.35 %
e. What is the expected number of blanks which should be required to fill an
order for 100 completed parts?
100 100
Solution: = = 157.8532
A1,OK 0.6335
f. What percent of the parts arriving at machine #2 will be successfully
completed?
Solution: A 3,OK = 0.7909 = 79.09 %
g. What is the expected total number of inspections which entering parts will
undergo?
Solution: E12 + E14 + E16 = 2.4069
h. Explain the meaning of the number appearing in row 3, column 2 of the A
matrix.
Solution: A 32 = .2091 indicates that parts arriving at machine 2 have a
20.91 % probability of being failed.
i. Explain the meaning of the number appearing in row 3, column 3 of the E
matrix.
Solution: E 33 =1.029 indicates the expected number of times that parts
arriving at machine 2 returns to machine 2 before successfully finished or
failed.
j. To fill the order for 100 completed parts, what is the expected man-hour
requirement for each machine? for each inspection station?

Solution:
Operation state Man-hour / 100 completed part

Machine 1 1 1.047 × 0.5 × (1/0.6335)× 100 = 82.64

Inspection 1 2 1.047 × 0.1 × (1/0.7039)× 100 = 14.87
Machine 2 3 1.029 × 0.75 × (1/0.7909)× 100 = 97.58
Inspection 2 4 1.029 × 0.2 × (1/0.8325)× 100 = 24.72
Machine 3 5 1.02 × 0.25 × (1/0.9296)× 100 = 27.43
Inspection 3 6 1.02 × 0.25 × (1/0.9486)× 100 = 26.88
Pack-&-Ship 7 1 × 0.1 × (1/1) × 100 = 10

Total = 284.12
k. What are the expected direct costs (row materials + operating costs - scrap
value of rejected parts) per completed part?

Solution:
Each completed part requires an expected 1/0.6335, i.e., 1.5785 entering
parts.
Materials : $ 50 × 1.5785 = $ 78.925
Scrap value of rejected parts : $ 10 × 1.5785 × 0.3665 = $ 5.785
Estimated Man-hour requirements per completed part

57:022 HW#1 2/1/97 page 55

Solutions

Operation state Man-hour / completed part

Machine 1 1 1.047 × 0.5 × (1/0.6335) = 0.8264

Inspection 1 2 1.047 × 0.1 × (1/0.7039) = 0.1487
Machine 2 3 1.029 × 0.75 × (1/0.7909) = 0.9758
Inspection 2 4 1.029 × 0.2 × (1/0.8325) = 0.2472
Machine 3 5 1.02 × 0.25 × (1/0.9296) = 0.2743
Inspection 3 6 1.02 × 0.25 × (1/0.9486) = 0.2688
Pack-&-Ship 7 1 × 0.1 × (1/1) = 0.1

Total = 2.8412
Operation cost :
Operation state Hourly Rate × Man-Hours

Machine 1 1 20 × 0.8264 = 16.528

Inspection 1 2 15 × 0.1487 = 2.2305
Machine 2 3 20 × 0.9758 = 19.516
Inspection 2 4 15 × 0.2472 = 3.708
Machine 3 5 20 × 0.2743 = 5.486
Inspection 3 6 15 × 0.2688 = 4.032
Pack-&-Ship 7 10 × 0.1 = 1

Total = $ 52.5005

Total Direct Cost = row material + operating costs - Scrap value of rejected
parts
= $ 78.925 + $ 52.5005 - $ 5.785
= 125.6405

3. A city's water supply comes from a reservoir. Careful study of this reservoir over
the past 20 years has shown that, if the reservoir was full at the beginning of the
summer, then the probability it would be full at the beginning of the following
summer is 80%. On the other hand, if the reservoir was not full at the beginning of
one summer, the probability it would be full at the beginning of the following
summer is only 25%.
a. Defining the states to be "full" and "not full", draw a transition diagram for a
Markov chain model of this reservoir.

Solution:

0.80 0.75
state definition 0.20
1 full 1 2
2 not full 0.25
b. Write the transition probability matrix.

Solution:
 0.80 0.20 
P=
 0.25 0.75 

57:022 HW#1 2/1/97 page 56

Solutions

c. If the reservoir was full at the beginning of summer 1996, what is the
probability that it will be full at the beginning of summer 1998?
  0.6900 0.3100  
Solution: P112 = 0.69 from P2 = 
  0.3875 0.6125  
d. Write equations which determine the steadystate probability distribution for
this Markov chain, and solve them.
Solution:
 π1 = 0.80π 1 + 0.25π 2  π 2 = 0.20π 1 + 0.75π 2
 or 
 π1 + π 2 = 1  π1 + π 2 = 1
5 4
∴π 1 = , π2 =
9 9
e. Over a long period of time, in what percent of the years would we expect the
reservoir be full at the beginning of the summer, according to this model?

5
Solution: π1 = = 55.56 %
9

57:022 HW#1 2/1/97 page 57