Random Errors:
Uncertainty Analysis
Since the total error includes random error, which
are uncertain, errors are usually referred to as
uncertainties.
For any experimental study, uncertainties analysis
must be performed and reported along with the
measurements.
approach
involve
probability
Propagation of Uncertainties
Consider the measurement of an effective resistance of
a circuit comprising three resistors in series
N = f (u1, u2 , u3 ,....., un )
The total uncertainty (in N) n would include uncertainties
of ui ; ie u1, u2, u3, u4, un
Therefore;
f
f
f
N =
u1 +
u2 + .... +
un
u1
u2
un
f
f
N =
u1 +
u2 + ... +
un
u1
u2
un
Example problem
Calculate the uncertainty in head loss hl expressed as
2
flV
hl =
2 gd
Solution
2
hl hl f
hl =
l +
v + d
l v d
fv 2 hl
hl
where
=
= ;
l 2 gd l
hl 2 fvl hl
=
=
v 2 gd 2v
hl
hl fv 2l
=
=
d 2 gd 2
d
hl 2 fvl hl
=
=
v 2 gd 2v
h
fv
hl
=
= l;
l 2 gd l
hl
hl fv 2l
=
=
2
d 2 gd
d
1/ 2
l 1 v d
hl = hl + +
l 4 v d
2
1/ 2
1
2
2
2
1/ 2
hl
1
2
2
2
= [ 0.02] + [ 0.04] + [ 0.01]
hl
4
Total Error = 3 %
= 0.03
Points to be noted:
In many mechanical problems, we assume the value
of g is 9.8.
However, in reality; there is an uncertainty in the
value of g, which has to be accounted where the
situation arises.
So, expect for universal constants, all other
parameters used for estimating a quantity to be
considered while calculating the total uncertainties.
pd 4
Q=
128 l
Grashof Number; Gr
Gr =
g 2 (Ts T )l 3
Solution
Q
Q Q
Q
Q =
C d +
A +
( P) +
C d
A P
2
Q Q Q Q
Q Q
Q
Q
=
;
= ;
=
;
=
C d C d A A p 2 p
2
1/ 2
uC d 2 u A 2
u p 2
u 2
= ( ) + ( ) + ( ) + (
)
Q Cd
A
2
2p
uQ
C d
= 0.005
Cd
A
= 0.002
;
A
p
= 0.004
p
P
=
RT
95000
=
287 X 293
= 1.13 kg / m3
= 0.005
Total uncertainty
u p 2
u 2
uA 2
2
d
= (
) +( ) +( ) +(
)
Q Cd
A
2
2p
uC
uQ
uQ
1/ 2
2 1/ 2
2 P
= 0.63
(0.06) 2
2 157.85
= 0.03 m3 / hr
1.13
The graphs should be plotted in such a way that the reader should
understand the conveyed information with out any difficulties.
2.
The axes should have clear labels i.e. name of quantities, units, and
their symbol if necessary
3.
Axes should be clearly numbered and should have clear tick marks
with significant numerical divisions (Sub division also should be
clearly mentioned)
1.6
1.4
1.2
1
0.8
0.6
MmNi4.6Al0.4
ma=0.4 kg
0.4
Absorption temperature = 27C
Absorption pressure
= 80 bar
0.2
0
0
10
Number of cycles
12
4.
5.
6.
18
16
20
14
Compressor work
12
MmNi4.6Al0.4
Tc = 20C, Ps = 10 bar
ma = 0.4 kg, Th = 85C
10
8
20
15
10
Compressor work
MmNi4.6Al0.4
Tc = 20C, Ps = 10 bar
ma = 0.4 kg, Th = 85C
0
28
30
32
34
36
38
40
42
10
15
20
25
30
35
40
45
100
MmNi4.6Al0.4
Tc = 20C, Ps=5 bar
ma = 0.4kg
3.5
95C
95C
85C
10
85C
75C
75C
2.5
2
0.1
1.5
1
0.01
Hydrogen compressed
Hyderogen compression rate
0.5
0
4.5
0.001
0
50
100
150
200
250
300
350
400
450
500
Time (s)
4.5
MmNi4.6Al0.4
Tc = 20C, Ps=5 bar
ma = 0.4kg
4
3.5
95C
85C
75C
95C
85C
75C
2.5
2
1.5
Hydrogen compressed
Hyderogen compression rate
0.5
0
0
50
100
150
200
250
300
350
400
450
Time (s)
500
8
7
6
5
4
3
2
Experimental values
Numerical simulation
Ps = 5 Bar
Ps = 10 Bar
Ps = 15 Bar
MmNi4.6Al0.4
Tc = 20C
ma = 0.4 kg
0
70
75
80
85
90
95
100
40
=
=
=
=
35
30
MmNi4.6Fe0.4
MmNi4.6Al0.4
10C
15C
20C
25C
25
20
15
10
5
0
0
0.2
0.4
0.6
0.8
Concentration (H/M)
1.2
16
40
U = 1000 W/m K
Tc = 20C
Ps = 10 bar
Th = 85C
Vs = 3.8 litre
35
30
14
12
10
25
8
20
6
15
10
5
0
1600
200
400
600
800
1000
1200
1400
MmNi4.6Al0.4
Time (s)
Choosing X Y Coordinates
Linear - Linear
Liner Logarithmic ( Semi log )
Logarithmic Logarithmic (Full log)
Polar coordinates ( Quantities varies with an angle)
Selection of axis
SS residual
R = 1
SStotal
2
SS residual = i=1 ( Ei Pi )
n
SStotal = ( n 1) 2
n
i =1
(E E)
= standard deviation
Ei = Experimental value
Pi = Predicted value
n = total numberof data
S = i =1 yi ( axi + b )
y = ax + b
y2, x
r = 1 2
y
[ yi yic ]
= i =1
n2
y, x
1/2
[ yi y m ]
y = i =1
n 1
1/2
f (a, b) = a + bx
Vertical least squares fitting proceeds by finding the sum of the squares of the
vertical deviations of a set of data points . The error E is defined as
n
E (a, b) = [ yi (a + bx)]2
2
i =1
n
( E 2 )
= 2 [ yi (a + bxi )] = 0
a
i =1
n
( E 2 )
= 2 [ yi (a + bxi )]xi = 0
b
i =1
i =1
i =1
na + b xi = yi
n
a xi + b xi = xi yi
2
i 1
n
x
i =1
i =1
i =1
n
x
i
i =1
a i =1 yi
n
=
n
2
b
x
x
y
i =1 i i
i =1
n
a
= n
b x
i =1
x
i=1 i
n
2
xi
i =1
1
n
y
i =1 i
n xy
i =1 i i
a
1
=
b n n xi2
i =1
( x )
n
i =1 i
n y n x2 n x n x y
i =1 i i =1 i i=1 i i =1 i i
n n xy n x n y
Observation
1. For a perfect fit y,x = 0; here r = 1. There is no variation
between the estimated values and the values obtained from
the correlation.
2. r = 0 indicates a poor fit or the values are widely scattered
around a straight line. Hence, the value of r should be closer
to 1 for a good fit.
3. It is necessary to observe the behavior of the fit and also
their range of scatter. If the data scatter but still appear to
follow a liner relationship, then a fit is also said to be poor
one.
Example
= A1 exp
RT
n
---------(1)
Q
ln = ln A1 + n ln
RT
Q
Ei = ln ln A1 + n ln +
RT
The minimization of the error is carried out by finding the
partial derivatives of iN=1 Ei2
w.r.t ln(A1), n and Q/R and
setting them equal to zero, where N is the number of
observations. Taking the partial derivatives, the following
equations are obtained
i =1 ln A1 + i =1 n ln i =1 ln
n
( ln ln A ) +
n
i =1
n
i =1
n ( ln )
Q
n
= i =1 ln ( )
RT
n
ln Q
i =1 ln
= i =1 ( ln ln ( ) )
RT
n
ln A1
n n ln
n
n ln ( )
Q
i =1
i =1
+ i =1
= i =1
2
T
T
RT
T
n
N ln
i =1
N 1
i 1
i =1 ln
N
2
(ln
)
i =1
N
ln
i =1
T
N
T ln ( A ) i =1 ln
1
N
N ln
n = i =1 ln ln
i =1
T
N ln
N 1
i =1 T 2 R i =1 T
i =1
N
N ln
i =1
N 1
i 1
T
i =1 ln
N
2
(ln
)
i =1
i =1
N
ln
T
ln
T ln ( A ) i =1
N
N ln
i =1
n = i =1 ln ln m
T
N ln m
N 1
i =1 T 2 R i =1 T
i =1
N
= A1 n exp
[ x ] = [ A] [ b ]
250
Applied pressure vs milliampere
200
150
y = 12.804x - 51.611
R2 = 1
Trendline
100
50
0
0
10
15
20
Milliampere (mA)
25
20
18
16
Compressor work
14
Different
Patterns of fit
12
10
MmNi4.6Al0.4
Tc = 20C, Ps = 10 bar
ma = 0.4 kg, Th = 85C
8
6
15
20
25
30
35
45
40
20
y = 41.718e-0.0362x
R2 = 0.7532
18
16
14
Compressor work
12
10
MmNi4.6Al0.4
Tc = 20C, Ps = 10 bar
ma = 0.4 kg, Th = 85C
8
6
15
20
25
30
35
40
45
Different pattern
of fits
Reference:
J.P.Holman
Different pattern
of fits
Reference:
J.P.Holman
N = f (u1, u2 , u3 ,....., un )
Consider a variable N,
The total uncertainty (in N), N was defined by root sum squares
f
f
N =
u1 +
u2 + ... +
un
u1
u2
un
N
= u1a1u2a 2u3a 3 (ai uiai 1 )..........unan
ui
Dividing by N,
1 N ai
=
N ui ui
2 1/2
N ai ui
=
N
xi
N = a 1 u 1 + a 2 u 2 + . .. ... ... . + a n u n =
N
= ai
ui
1/2
2
N
2
2 1/2
N =
( ui ) = ( ai .ui )
ui
a iu i
Sample Problems
1. The resistance of a copper wire is given as R = R0[1+(T-20)].
Where Ro = 6 0.3% is the resistance at 20C.
= 0.004C 1%, T = 30 1C.
Calculate the uncurtaining in the resistance of the wire.
2. A resistor has an nominal stated value of 10 1%. A voltage is
impressed on the resistor and the power dissipation is to be
calculated in two different ways. (i) Only voltage is measured (ii)
both Voltage and current will be measured. Calculate the uncertainty
in power for each case when the measured values of E and I are
E = 100V 1% (for both case and I= 10A 1%.