You are on page 1of 20

Treatment and Interpretation of Engineering Data

1. Graphical representation

 Propose  to show what has happened


 to show a relationships between quantities
 to show distribution

 General types of graphs


 Time series (a); Scatter plot (b); Histogram (c)
6 6
(a) (b) (c)
4 4
B (unit)

B (unit)
2 2

0
0
0 1 2 3 4 5
A B C D E F G H
A (unit)
 What about the axes?
 Horizontal axis (x-axis) => Independent variable, explanatory variable
 Vertical axis (y-axis) => Dependent variable, response variable

 What's interesting?
 Slope of tangent: Rate of change of y with x
 Area under curve: Cumulative product of x and y
y

y2

y = y2-y1
y1
x = x2-x1
Slope = y/x

x
x1
x2
General shape of curve fit families

Logarithm

http://www.ncsu.edu/labwrite/res/gt/gt-menu.html
2. Empirical equations
Example 25
x y 20
0.2 3.2
15

y
0.4 3.7
1.0 4.1
10
y  a  bx
5
2.0 8.1
0
3.0 13.7 0.0 1.0 2.0 3.0 4.0 5.0
4.0 22.6 x

y  ax n

???
y  c  ax n

 3.2  ax n
2.1 Evaluation of constants
 Direct reading from curve  very simple
From curve
x y
0.2 3.2 At x = 0.6 => y-3.2 = 0.4; and
at x = 4.5 => y-3.2 = 25
0.4 3.7
0.4  a0.6
n
1.0 4.1
Then (1)

25  a4.5
2.0 8.1 n
(2)
3.0 13.7
Solving; a = 1.15; n = 2.05
4.0 22.6
or y  3.2  1.15x 2.05
Check
x y ycal R = ycal- y R2
0.2 3.2 3.2 0 0
0.4 3.7 3.5 -0.2 0.04
1.0 4.1 4.4 0.3 0.09
2.0 8.1 8.0 -0.1 0.01 Not good
3.0 13.7 14.1 0.4 0.16
4.0 22.6 22.9 0.3 0.09
R = 0.7  R2 = 0.39
 Method of average R  ycal  y and R  0

Example
Step 1. Classified the data into group: No. of group = No. of constant
x y
Step 2. Define the equation: y  c  ax n
 
0.2 3.2
0.4 3.7
Step 3. Given the residue: R  ycal  y  c  ax n  y
R1  c  a0.2  3.2
n
1.0 4.1 then (1.1)
R2  c  a0.4  3.7
n
2.0 8.1 (1.2)
3.0 13.7 R3  c  a1.0  4.1
n
(2.1)
4.0 22.6 R4  c  a2.0  8.1
n
(2.2)
R5  c  a3.0  13.7 (3.1)
n

R6  c  a4.0  22.6
n
(3.2)
Step 4. For each group: R  0 ; then
Solving: a = 1.07, c = 3.3
2c  a0.2  0.4
n n
  6.9  0 (a) n = 2.08

2c  a 1  2
n
 12.2  0
n
(b) then y  3.3  1.07 x 2.08

2c  a3  4   36.3  0 R = 0.2; R2 = 0.27


n n
(c)
Solving: a = 1.07, c = 3.3, n = 2.08 y  3.3  1.07 x 2.08

x y ycal R = ycal- y R2
0.2 3.2 3.34 0.14 0.018942
0.4 3.7 3.46 -0.24 0.058033
1.0 4.1 4.37 0.27 0.0729
2.0 8.1 7.82 -0.28 0.076155
3.0 13.7 13.81 0.11 0.013151
4.0 22.6 22.43 -0.17 0.029599
R = -0.2  R2 = 0.27
 Method of Least Squares Linear regression

Mean ( y) y 
y i
; i = 1,…, n
n

 y  y
2
St
Standard deviation (Sy) Sy  i

n 1 n 1


 iy  y 2
St
Varian (Sy2) Sy  
2

n 1 n 1
Sy
Coefficient of variation (c.v.) c.v.  100%
y

Linear regression

Empirical equation: ycal  a  bx

Residual or error: R  ycal  y  a  bx   y


n n
Sum of the squares of the residuals: Sr  R i
2
  a  bxi   yi 2

i 1 i 1
S r
Constant a and b can be calculated by  2 a  bxi  yi 
a

S r
 2 a  bxi  yi xi 
b

S r S r
Setting  0 and  0 ; then
a b
S r
 2 a  bxi  yi   0  a  bx  y   0  a   bx   y  0
a i i i i

 a  na na   xi b  y i (*)

S r
 2 a  bxi  yi xi   0  a  bx  y x   0
b i i i

 ax   bx
i
2
i  x y i i
(**)

n xi yi   xi  yi
Solving b  and a  y  bx
n x   xi 
2 2
i
Example

xi yi xiyi xi2
Fit a straight line to the x and y values of the
following data
1 0.5 0.5 1.0
2 2.5 5.0 4.0 From data n  7 x y
i i  119.5
3 2.0 6.0 9.0
4 4.0 16.0 16.0
x 2
i  140 x i  28
28
5 3.5 17.5 25.0  yi  24.0 x 
7
 4
6 6.0 36.0 36.0
24
7 5.5 38.5 49.0 y   3.42857
7
 28 24.0 119.5 140

From n xi yi   xi  yi 7119.5  2824.0


b    0.8393
n x   xi  7140.0  282
2 2
i

a  y  bx  3.42857  0.83934  0.0714

 y  0.0714  0.8393x
Error in linear regression
Sum of squares of difference between data and their average
n 2

St   y
i 1
i  y

 y  y
2
i
St
Sy   i 1
“Standard deviation”
n 1 n 1

Standard error of estimation difference between data and regression model

Sr
Sy/ x 
n2

 a 
n n n n

R  y  yi   y  ai  bxi 
2
when S r  i
2
 cal
2
i  bxi  yi  i
2

i 1 i 1 i 1 i 1

St  S r
Coefficient of determination R2  If…Sr = 0 and R = R2 = 1  ??
St

Correlation coefficient R  R 2 If…Sr = St and R = R2 = 0  ??


Distribution of standard error around regression equation  Normal curve
y y

x x
Distribution of data around mean Distribution of data around the best-fit line
of the dependent variables

Example of linear regression with residual errors


y y

x x
Linear regression with small residual errors Linear regression with large residual errors
Application of linear regression

 Power function
ln y
Empirical equation:
Curve fitting
Slope = B
y  Ax B
ln y  ln A  B ln x
ln x

Linear Intercept = ln A
y
regression

n n n
x n ln xi ln yi    ln xi  ln yi 
b  i 1 i 1 i 1
2
n
 n 
n ln xi     ln xi 
2

i 1  i 1 
n n

 ln y   b ln x 
i i
a  i 1 i 1
 ln y  bln x
n
Where B  b and A  exp(a)
 Exponential function
Empirical equation: ln y
Curve fitting

y  Ae Bx  A exp Bx  ln y  ln A  Bx
Slope = B

Linear
y
regression x
Intercept = ln A

x
n n n
n xi ln yi    xi  ln yi 
b  i 1 i 1 i 1
2
n
 n 
n xi     xi 
2

i 1  i 1 
n n

 ln y   b x 
i i
a  i 1 i 1
 ln y  b x
n

Where B  b and A  exp(a)


 Logarithmic function
Empirical equation: y
Curve fitting
y  a  b ln x y  a  b ln x
Slope = b

y Linear
regression ln x
Intercept = a

x
n n n
n  yi ln xi    ln xi   yi 
b  i 1 i 1 i 1
2
n
 n 
n ln xi     ln xi 
2

i 1  i 1 
n n

  y   b ln x 
i i
a  i 1 i 1
 y  bln x
n
 Saturation growth rate function
Empirical equation: 1/y
Curve fitting
x 1 1 B
y  A  
Bx y A Ax Slope = B/A

Linear
y regression 1/x
Intercept = 1/A

x 1 1 n 1n 1
n
n         
i 1  yi xi  i 1  xi  i 1  yi 
b  2
 1   n  1 
2
n
n        
i 1  xi   i 1  xi  

1
n n
1
  y    x 
 
i 1  i 
 b
i 1  i 
a   1 / y  b1 / x
n

Where A  1/a and B  bA


Example
A chemical engineer is studying the rate at which a reactant is consumed
in a chemical reaction involving the following data. Define the relationship of
the concentration and reaction rate by using the linear regression concept.

Concentration (mole/ft3) 100 80 60 40 20 10 5 1


Reaction rate (mole/s) 2.85 2.00 1.25 0.67 0.22 0.072 0.024 0.0018

Solution
3
It is in the power form
2

y  Ax B ln y  ln A  B ln x
0
Reaction rate (mole/s)

ln(y)

-2
-4
-6
2 -8
0 1 2 3 4 5
ln(x)
n n n
n ln xi ln yi    ln xi  ln yi 
1
b  i 1 i 1 i 1
2
n
 n 
n ln xi     ln xi 
2

0 i 1  i 1 
0 20 40 60 80 100 n n

Concentration (mole/ft3)
 ln y   b ln x 
i i
a  i 1 i 1
 ln y  bln x
n
Where B  b and A  exp (a)
xi yi ln(xi) ln(yi) ln(xi)ln(yi) (ln(xi))2 From data
100 2.85 4.605 1.047 4.823 21.208
n  8
80 2.00 4.382 0.693 3.037 19.202
lnx  lny
i i
 23.6812.63  299.08
60 1.25 4.094 0.223 0.914 16.763
40 0.67 3.689 -0.400 -1.477 13.607  ln x  23.68
i

20 0.22 2.996 -1.514 -4.536 8.974  ln y  12.63


i

 ln x   87.648
2
10 0.072 2.303 -2.631 -6.058 5.302
i
5 0.024 1.609 3.730 -6.003 2.590
 ln x   23.6823.68
i
2
 560.74
1 0.0018 0.000 -6.320 0.000 0.000
23.68  12.63
 316 7.09 23.68 -12.63 -9.300 87.648 ln x   2.96 ln y   1.58
8 8

n xi yi   xi  yi n lnxi lnyi   lnxi  lnyi 8 9.30  23.68 12.63


From b     1.5988
n xi2   xi  n lnxi 2   lnxi  887.648  560.74
2 2

a  y  bx  lny  blnx  1.58  1.59882.96  6.312 y = 0.0018x1.599

Reaction rate (mole/s)


2 R2 = 1

 A  exp a   exp (6.312)  0.00182 1

By Excel
B  b  1.5989 0
0 20 40 60 80 100

Solution y  0.00182 x 1.5989


Concentration (mole/ft3)
 Polynomial function
Empirical equation: y  a0  a1 x  ...  ak x k

 y  a 
n 2

Sr   a1 xi  ...  ak xi
k
Residual i 0
i 1

Partial derivatives (dropping superscripts)


S r
   
n
 2 yi  a0  a1 xi  ...  ak xi
k
0
a0 i 1

S r
   x
n
 2 yi  a0  a1 xi  ...  ak xi  0
k

a1
i
i 1

S r
   x
n
 2 yi  a0  a1 xi  ...  ak xi  0
k k

ak
i
i 1

n n n
Solution a0 n  a1  xi  ...  ak  x k
i  y i
i 1 i 1 i 1

n n n n
a0  xi  a1  x  ...  ak  x 2
i
k 1
i  x y i i
i 1 i 1 i 1 i 1

n n n n
a0  x  a1  x
k
i
k 1
i  ...  ak  x 2k
i  x k
i yi
i 1 i 1 i 1 i 1
Example
Empirical equation: Empirical equation:

y  a0  a1 x y  a0  a1 x  a2 x 2

 y  a 
n 2
Residual Sr   a1 xi  a2 xi
2
Residual i 1
i 0

S r
   
2 n
 2 yi  a0  a1 xi  a2 xi
n

 y  a  a1 xi 
2
Sr  i 0 a0
0
i 1
i 1

S r
   x
n
S r n
 2 yi  a0  a1 xi  a2 xi
 2  yi  a0  a1 xi   0  0
2

a1
i
a0 i 1 i 1

S r
   x
n
S r  2 yi  a0  a1 xi  a2 xi  0
n 2 2
 2  yi  a0  a1 xi  xi  0 a2 i 1
i
a1 i 1

n n n n n
or na0  a1  xi  y i 1 or na0  a1  xi  a2  x 2
i
 y i 1
i 1 i 1 i 1 i 1 i 1
n n n n
n n n a0  xi  a1  x 2
 a2  x  3
x y 2
a0  xi  a1  x x y
i i i i
2
i  i i 2 i 1 i 1 i 1 i 1
i 1 i 1 i 1
n n n n
a0  x  a1  x  a2  x
2
i
3
i
4
i  x 2
i yi 3
i 1 i 1 i 1 i 1

You might also like