–18-08-2021

CVL 773
IIT DELHI
CIVIL ENGINEERING MATERIALS.

( Lecture 1)
Introduction & Statistical concepts

B. Bhattacharjee
CIVIL ENGINEERING DEPARTMENT
IIT DELHI

B. Bhattacharjee
DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

General Outline
™Data Reduction √
™Normal distribution & examples √
™Distribution of sampling mean √
™“t” and “F” statistic, example of hypothesis
testing
™Binomial distribution and Poisson’s
distribution
™Joint and Conditional probability

DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

2

–1
 –18-08-2021

Data-Discrete-integer/continuous random variable

1. If we have a hotel with 100 beds and we want to provide a
central A.C. All AC s will not operate simultaneously
2. If there are n (say n=16), 32 Ampere outlets in a
construction sites for vibrators, winches etc. Again all will
not be operating simultaneously.
3. We want to know on an average how many AC s or how
many sockets draw current simultaneously ; on an average
4. We go say a 1000 times to a facility with ‘n’ outlets and
carry out a survey,
5. Random variable here is discrete and integer i.e. the
outlets (AC/socket)
f 1 times we found 0 outlet in use
f 2 times we found 1 outlets in use, f 3 , f 4, ….f n

B. Bhattacharjee
3
DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

Data reduction discrete

The results shown in histogram plot
25
Total observation
=Σfi=153
20
Frequency

15

fi
10

5 Relative frequency
0 f1 f2
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
=fi/Σfi
ΣRelative frequency
Sockets

=Σfi/Σfi=1

Frequency Histogram Dmax =16, Dmin =0

B. Bhattacharjee
4
DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

–2
 –18-08-2021

Continuous Data (CI)

fi
f1 f2

B. Bhattacharjee
5
DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

CONCRETE QUALITY (USPV)

Zone Doubtful Medium Good Excellent Total
location location Location
Up to 6th 6 5 282 28 321
Floor
7-13th 26 100 325 3 454
Floor
14th-17th 10 82 221 2 315
Floor
Total 42 187 831 33 1093

DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

–3
 –18-08-2021

NDT results
NO. OF SAMPLE: 321 Up to 6th
floor
MEAN: 4.1 Mode 4.17
MEDIAN: 4.1
STANDARD S P
DEV: 0.375
MAXIMUM: 4.9 Mean-3SD
MINIMUM: 1.7 2.94
RANGE: 3.2
NO. OF
CI(3.3LOGN+1) 9
CLASS INT. 0.357
B. Bhattacharjee
DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

UPVresults Up to 6th
Interval Relative
no. lower upper Frequency frequency
1 1.7 2.06 3 0.01
2 2.06 2.41 0 0
3 2.41 2.77 2 0.01
160
4 2.77 3.13 1 0
140 5 3.13 3.48 5 0.02
6 3.48 3.84 54 0.17
120
7 3.84 4.2 149 0.46
100 8 4.2 4.55 91 0.28
80
9 4.55 4.91 16 0.05

60 Sum 321 1

40

20

0
1 2 3 4 5 6 7 8 9

B. Bhattacharjee
DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

–4
 –18-08-2021

Discrete Uniform distribution

When f1=f2=fi=fn .., the distribution is uniform
The probability mass function f(x)=Pr(x)=constant,
Consider simple case of a die, 6 out comes,
repeat the experiment of throwing the die large
number of times a=1, b=6, random variable
a≤x≤b; f(x)=1/6
The cumulative distribution
function CDF, F(x) is given by

B. Bhattacharjee
9
DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

NATURE OF VARIATIONS
0.18
(uniform distribution)
0.16

0.14 ¾ The variations

0.12
Leads to a uniformly
Relative Frequency

0.1
Distributed population
0.08
fi 1
0.06 relativefrequncy = n
≈
N = ∑ fi
0.04
6
0.02 1

1
Pr( x) = f ( x) =
0
1 2 3 4 5 6
Outcomes
6

DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

–5
 –18-08-2021

NATURE OF VARIATIONS
0.18
(triangular distribution)
¾ Two dies
0.16

0.14

0.12
The variations
Relative frequency

0.1 Leads to a triangular

0.08 Distributed population
0.06 fi
relativefrequncy = n
N = ∑ fi
0.04

0.02
1
0
1 2 3 4 5 6 7 8 9 10 11
1 2 3 4 Outcomes
5 6 7 8 9 10 11
(x) 2 3 4 5 6 7 8 9 10 11 12

NATURE OF VARIATIONS
0.06
(nonlinear distribution)
0.05
¾ Three dies
0.04
The variations
0.03
Leads to a nonlinear
0.02 Distributed population
fi
0.01 relativefrequncy = n

0
N = ∑ fi
1 2 3 4 1

DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

–6
 –18-08-2021

Random variable as sum of

random variables
When a random variable is a sum of number of
random variable, i.e.
X=x1+x2+x3……xn ….
f(x) is non linear ; for large value n dispersion
/spread increases
Curve becomes bell shaped

Gauss fitted normal (Gaussian) distribution

from error function concept; at x=-∞ or +∞,
f(x)=0; at x=0, f(x)=1
B. Bhattacharjee
13
DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

NATURE OF VARIATIONS
¾ For large number of
Factors, i.e. x , e.g.,
activities in a project
Frequency

¾The variations
Leads to a normally
Distributed population

Project duration
All critical activities finishes at
All critical activities finishes at
pessimistic duration
optimistic duration

DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

–7
 –18-08-2021

NATURE OF VARIATIONS
The error function (also called the Gauss error
function), often denoted by erf, is a function of a
variable z is defined as:
Area under the
curve erf (z) is 1
x−μ for z=∞,
Z= ;
σ ⎛ 1 ⎛ x − μ ⎞2 ⎞
1
⎛ x− μ ⎞
−⎜
2

x−μ
f ( x) = exp ⎜⎜ − ⎜ ⎟ ⎟⎟
σ π ⎝ 2⎝ σ ⎠ ⎠
⎟
⎝ σ ⎠
e = 0 for = −∞ or + ∞ 2
σ
∞
1 ⎛ 1 ⎛ x − μ ⎞2 ⎞ F ( 0, μ ,σ ) = 0
F ( ∞, μ , σ ) = ∫ exp ⎜ − ⎜ ⎟ ⎟⎟dx = 1
⎜
0 σ 2π ⎝ 2⎝ σ ⎠ ⎠
0 ≤ F ( x, μ ,σ ) ≤ 1

DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

NATURE OF VARIATIONS

ƒAverage represents mean

ƒSD measures
dispersion

Mean= Σx/n
s2=Σ(xi-x¯)2/(n-1)
Measure of
σ=s
DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI
16

–8
 –18-08-2021

NATURE OF VARIATIONS

Mean= Σx/n
s2=Σ(xi-x¯)2/(n-1)

DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

17

PERT duration
SD of activity duration.

Average is a measure of Mean & varies linearly

n
f i xi
mean μ = ∑ n
i =1
N = ∑ fi
1
For single dice experiment mean =
6
i (6 × 7)
mean μ = ∑ = = 3.5
i =1 N =6 2
For two dice 6
mean μ = ∑
( i + 1) i + 11 ( i + 1) (12 − i) = 7 = 3.5 + 3.5
: ∑
i =1 N = 36 i =7 36

B. Bhattacharjee
18
DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

–9
 –18-08-2021

Example of PERT
Uncertainty in activity duration arises out of
variation of durations of component work
elements of the activity
General pattern of variation of activity duration can
be visualized by considering a situation where
same activity is repeatedly performed several times
i.e., sufficiently large number of times.
For a given activity which is repeated large
number of say the minimum duration is Dmin and
Maximum duration is Dmax, one can divide the
range Dmax, - Dmin in to several class interval
{NCI=3.3log(n)+1}, & plot a histogram.

B. Bhattacharjee
19
DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

PERT (activity duration)

Common experience shown in histogram
250
plot
200
Most often
150
Frequency

Series1
100

50

Frequency Histogram Dmax =17, Dmin =5

5 6 7 8 9 10 11
Duration
12 13 14 15 16 17

B. Bhattacharjee
20
DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

–10
 –18-08-2021

PERT (activity duration)

Common experience shown in histogram plot can
be represented by 2 parameter β distribution
Most often Mean

Beta distribution
two parameter α=2 β=5
B. Bhattacharjee
21
DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

PERT (activity duration)

X axis is (D-Dmin)/(Dmax-Dmin), Y axis is
probability density
1<α <β

Beta distribution
two parameter α=2 β=4

β function
B. Bhattacharjee
22
DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

–11
 –18-08-2021

PERT (activity duration)

X axis is (D-Dmin)/(Dmax-Dmin), Y axis is
probability density 1<α <β

0.25×(17-5)+5=8=mode
1/3×(17-5)+5=9=mean Beta distribution
two parameter α=2 β=4

β function

B. Bhattacharjee
23
DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

PERT duration
Thus expected (mean) activity duration is
calculated from 3 time information in PERT,
pessimistic duration tp, most likely duration tm,
and optimistic duration to.

Network and subsequent calculations follows same

way as done for CPM using expected activity
duration
The project duration calculated is expected
project duration.
PERT deals probabilistic project duration.

B. Bhattacharjee
24
DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

–12
 –18-08-2021

NATURE OF VARIATIONS (more

activities)

a= minimum duration
b=maximum duration

¾ Three activities in series: The variations Leads

to a nonlinearly Distributed population

DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

NATURE OF VARIATIONS (more

activities)

¾ Three parallel activities ending on a node: The

variations Leads to a nonlinearly Distributed
population
DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

–13
 –18-08-2021

PERT duration
SD of activity duration.

Average is a measure of Mean & varies linearly

n
f i xi
mean μ = ∑ n
i =1
N = ∑ fi
1
For single dice experiment mean =
6
i (6 × 7)
mean μ = ∑ = = 3.5
i =1 N =6 2
For two dice 6
mean μ = ∑
( i + 1) i + 11 ( i + 1) (12 − i) = 7 = 3.5 + 3.5
: ∑ i =1 N = 36 i =7 36

B. Bhattacharjee
27
DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

NATURE OF VARIATIONS
¾ For large number of
Factors, i.e. x , e.g.,
activities in a project
Frequency

¾The variations
Leads to a normally
Distributed population

Project duration
All critical activities finishes at
All critical activities finishes at
pessimistic duration
optimistic duration

DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

–14
 –18-08-2021

PERT duration
SD of activity duration.

Mean all activity duration is sum of means of

individual activity durations
mean μ = ∑ μ j ; j s tan dsforall criticalactivities
j

SD cannot be added as it involves square terms

SD square of activities can be added to obtain

project SD

B. Bhattacharjee
29
DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

PERT duration
SD of activity duration.

Using the mean and standard deviations probability

of project completion for a given duration can be
calculated
x−μ
Z= ;
σ
Normal distribution table or graph given in the
next slide can be used.

B. Bhattacharjee
30
DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

–15
 –18-08-2021

PERT duration
x−μ
Z= ;
σ
99.9% of time project
would be completed
by Mean+3×SD of
project duration, i.e.
Z=3
(95% of project will be
completed
Mean+1.65×SD of
project duration, i.e.
Z=1.65
In general p(D) for
D=μproj +Z×σproj from
graph or table

B. Bhattacharjee
31
DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

Distribution of sampling mean

–Take n samples from a large population
Find the mean of n samples, say m (large
number times, [∞] plot histogram and find mean
of means, SD etc.
–Distribution of these means is also normal
–Mean is same as population mean
Standard deviation is small
n n n
x1i x x
means : x1 = ∑ ; x2 = ∑ 2i ....xm = ∑ mi
i =1 n i =1 n i =1 n n

m xj m ∑x ji
Population mean μ = ∑ =∑ 1

j =1 m j =1 mn
DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

–16
 –18-08-2021

Distribution of sampling mean

Consider now n1 and n2 samples from same population
such that n2 is 2n1 , the variation of sampling mean for n1
is likely to be higher than n2
In general sampling mean will be less variable than
single observation
–Mean is same as population mean Standard
deviation is small s2=Σ(xi-x¯)2/(n-1)
σ 2
s2 Measure of σ = s
sn2 = sn2 =
n n proven by statistical
experiment
n 1 4 9 64 100 ∞
sn=f(σ) σ σ/2 σ/3 σ/8 σ/10 0

DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

Distribution of sampling mean

Using second moment concept

σ2
s =
2
n
n
Can be derived
DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

–17
 –18-08-2021

DISTRIBUTION OF SAMPLING MEAN

Mean= Σx/n
sn 2=s2/n
–Sampling mean
distribution
3sn + 3sn

DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

Number of test
• Higher the number of test (repetitions)
less is the error in measurements
• Number of test required to obtain the
mean (true) results within an error E with
a given confidence level (say 99%) is
given by:
For known population standard deviation
2
• n=(Z×σP/E)
• For unknown population standard
deviation
2
• n=(t×s/E)
DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI
36

–18
 –18-08-2021

THANK YOU
FOR HEARING

DEPARTMENT OF CIVIL ENGINEERING, IIT DELHI

37

–19