You are on page 1of 42

Calibration & Curve Fitting

P M V Subbarao
Professor
Mechanical Engineering Department

How to obtain y=f(x) ???????

We have developed confirmed methods to measure xi & ui


What are the units of xi & ui?
How to convert a Measured variable into Physical
(Derived) Variable?

Instrument & calibration


An instrument is a device that transforms a physical variable of
interest (the measurand) into a form that is suitable for recording
(the measurement).

Simple Instrument Model

Advanced Instrument Model

Sensors
Sensors convert physical variables to signal
variables.
Sensors are often transducers : They are devices
that convert input energy of one form into output
energy of another form.
Sensors can be categorized into two broad classes
depending on how they interact with the
environment they are measuring.
Passive sensors: do not add energy as part of the
measurement process but may remove energy in
their operation.
Active sensors : add energy to the measurement
environment as part of the measurement process.

Interpolation Vs Curve Fitting

Calibration
The process of development of a relationship between the
physical measurement variable input and the signal variable
(output) for a specific sensor is known as the calibration of the
sensor.
Typically, a sensor (or an entire instrument system) is calibrated
by providing a known physical input to the system and
recording the output.
The data are plotted on a calibration curve.

Sensitivity of A Sensor

Saturation Point

Linear range

Sensitivity of Thermistor

Curve Fitting Techniques

Where does this given function


Measured Variable = f (Physical Variable)
come from in the first place?
Analytical models of phenomena (e.g. equations from physics)
Create an equation from observed data
Curve fitting - capturing the trend in the data by assigning a
single function across the entire range.
A straight line is described generically by

f x ax b

The goal is to identify the coefficients a and b such that


f(x) fits the data well

Linear curve fitting (linear regression)

Given the general form of a straight line


How can we pick the coefficients that best fits the
line to the data?
What makes a particular straight line a good fit?

Quantifying error in a curve fit


Assumptions:
positive or negative error have the same value (data point is
above or below the line)
Weight greater errors more heavily
Denote data values as (x, y)
Name points on the fitted line as
(x, f(x)).

The error is available at the


four data points.

Hunting for A Shape & Geometry of A Data Set

Error d i2 y1 f x1 y2 f x2 y3 f x3 y4 f x4
2

f x ax b

Our fit is a straight line, so now substitute


N

Error yi f xi yi axi b
2

i 1

i 1

The best line has minimum error between line and data points
This is called the least squares approach, since square of the error
is minimized.

Minimize Error yi axi b


i 1

Minimize Error yi axi b


i 1

Take the derivative of the error with respect to a and b, set each to zero
N
Error
2 xi yi axi b 0
a
i 1
N
Error
2 yi axi b 0
b
i 1

Solve for the a and b so that the previous two equations both = 0

i 1

i 1

a x b xi xi yi
2
i

i 1

i 1

i 1

a xi bN yi
put these into matrix form

N
x

i 1

b
i 1

N

2 a
xi

i 1

y
i 1

x y
i 1

i 1

i 1

i 1

N xi yi xi yi
N

N x
i 1

2
i

i 1

y x x x
i 1

i 1
N

2
i

N x

i 1

i 1

2
i

i 1

i 1

yi

Is a straight line suitable for each of these cases ?

The Least-Squares mth Degree Polynomials


When using an mth degree polynomial

y a0 a1 x a2 x .........am x
2

to approximate the given set of data, (x1,y1), (x2,y2) (xn,yn),


where n m, the best fitting curve has the least square error,
i.e.,

i 1 yi f xi min
n

i 1 yi a0 a1 xi a x ......a x
n

2
2 i

n
n i

min

To obtain the least square error, the unknown coefficients


a0, a1, . and am must yield zero first derivatives.

Expanding the previous equations, we have

The unknown coefficients can hence be obtained by solving the


above linear equations.

No matter what the order j, we always get equations LINEAR with


respect to the coefficients.
This means we can use the following solution method

Selection of Order of Fit

2nd and 6th order look similar, but 6th has a squiggle to it.
Is it Required or not?

Under Fit or Over Fit: Picking An appropriate Order


Underfit - If the order is too low to capture obvious trends in the
data
Overfit - over-doing the requirement for the fit to match the data
trend (order too high)
Polynomials become more squiggly as their order increases.
A squiggly appearance comes from inflections in function

General rule: pick a polynomial form at least several


orders lower than the number of data points.
Start with linear and add order until trends are matched.

Linear Regression Analysis

Linear curve fitting


Polynomial curve fitting
Power Law curve fitting: y=axb
ln(y) = ln(a)+bln(x)
Exponential curve fitting: y=aebx
ln(y)=ln(a)+bx

Goodness of fit and the correlation coefficient


A measure of how good the regression line as a
representation of the data.
It is possible to fit two lines to data by
(a) treating x as the independent variable : y=ax+b, y
as the dependent variable or by
(b) treating y as the independent variable and x as the
dependent variable.
This is described by a relation of the form x= a'y +b'.
The procedure followed earlier can be followed again.

i 1

i 1

a y b yi xi yi
'

2
i

i 1

'

a yi b N xi
'

'

i 1

i 1

put these into matrix form

i 1

b'
i 1

N
2 a '
yi

i 1

x
i 1

x y
i 1

x a yb a
'

'

'

i 1

i 1

i 1

N xi yi xi yi
N

N y
i 1

2
i

i 1

Recast the second fit line as:

1
a'

1
b'
y ' x '
a
a
is the slope of this second line, which not same as
the first line

The ratio of the slopes of the two lines is a measure of


how good the form of the fit is to the data.
In view of this the correlation coefficient defined
through the relation

slope of first Regression line


'

aa
Slope of second Regression line
2

y ax b a

i 1

i 1

i 1

N xi yi xi yi
N

N x
i 1

x a y b a
'

'

'

2
i

i 1

i 1

i 1

i 1

N xi yi xi yi
N

N y
i 1

2
i

i 1

i 1

i 1

i 1

N xi yi xi yi

N x

i 1

2
i

2
xi N yi

i 1
i 1

i 1

i 1

i 1

i 1

N xi yi xi yi

N x
i 1

2
i

2
xi N yi

i 1
i 1

y
i 1

Correlation Coefficient
The sign of the correlation coefficient is determined by the sign
of the covariance.
If the regression line has a negative slope the correlation
coefficient is negative
while it is positive if the regression line has a positive slope.
The correlation is said to be perfect if = 1.
The correlation is poor if 0.
Absolute value of the correlation coefficient should be greater
than 0.5 to indicate that y and x are related!
In the case of a non-linear fit a quantity known as the index of
correlation is defined to determine the goodness of the fit.
The fit is termed good if the variance of the deviates is much less
than the variance of the ys.
It is required that the index of correlation defined below to be
close to 1 for the fit to be considered good.

y
i 1

f xi

y
i

i 1
y

N
i 1

2=1.000

2=0.821

2=0.991

2=0.493

2=0.904

2=0.0526

Multi-Variable Regression Analysis


Cases considered so far, involved one independent variable and
one dependent variable.
Sometimes the dependent variable may be a function of more than
one variable.
For example, the relation of the form

d orifice

m f p, T , p, A,

d
pipe

is a common type of relationship for flow through an Orifice or


Venturi.
mass flow rate is a dependent variable and others are
independent variables.

Set up a mathematical model as:

d orifice
p
c d
m a
p A

d pipe
RT
b

Taking logarithm both sides

d orifice
p
ln m ln a b ln
c ln p d ln A e ln

d pipe
RT
Simply:

y ln a bl cm dn eo

where y is the dependent variable, l, m, n, o and p are independent


variables and a, b, c, d, e are the fit parameters.

The least square method may be used to determine the fit


parameters.
Let the data be available for set of N values of y, l, m, n,
o, p values.
The quantity to be minimized is given by
N

Error yi a bli cmi dni eoi fpi


i 1

What is the permissible value of N ?

The normal linear equations are obtained by the usual


process of setting the first partial derivatives with respect
to the fit parameters to zero.
N
Error
2 yi a bli cmi dni eoi fpi 0
a
i 1
N
Error
2 li yi a bli cmi dni eoi fpi 0
b
i 1

i 1

i 1

i 1

i 1

i 1

i 1

Na b li c mi d ni e oi f pi yi
N

i 1

i 1

i 1

i 1

i 1

i 1

i 1

a li b li2 c li mi d li ni e li oi f li pi li yi

These equations are solved simultaneously to get the six


fit parameters.
We may also calculate the index of correlation as an
indicator of the quality of the fit.
This calculation is left to you!