MECHANICAL
ENGINEERING SYSTEMS
LABORATORY
Group 02
Asst. Prof. Dr. E. İlhan KONUKSEVEN
PRESENTATION OF
EXPERIMENTAL RESULTS
The presentation of numerical findings of an experiment is as
important as the conduct of the tests. Because all conclusions to
be drawn will be based on this information
THE MOST SUITABLE MANNER DEPENDS UPON
THE NATURE OF THE RESULTS
RESULTS MAY BE :
1. SINGLE VALUE
2. MULTIPLE VALUES
3. CONTINUOUS RELATIONSHIPS
PRESENTATION OF
EXPERIMENTAL RESULTS
There exist three basic ways of presenting the experimental data:
1. Tabular
2. Graphical
3. Empirical Equation
1. In a tabular way of presentation, all individual data sets are presented
with their numerical details; This type of representation is preferred if the
raw data need to be reported and/or if any further numerical data
analysis is to be carried out.
2. In a graphical way of presentation, the major goal is to emphasize on
the existing trends in the data in a visual manner.
3.In an empirical expression way of presentation, an equation is already
fitted to the original data. This type of representation is particularly
convenient if an analytical relationship is needed in a subsequent
analysis.
CONTINUOUS RELATIONSHIP BETWEEN
SEVERAL VARIABLES
THREE VARIABLES CAN BE DISPLAYED
CONVENIENTLY ON ONE DIAGRAM
IF THERE IS A FOURTH VARIABLE
THREE VARIABLE SHOULD BE PRESENTED ON
ONE PLOT AND
A DIFFERENT PLOT SHOULD BE PRODUCED
FOR EACH VALUE OF THE FOURTH VARIABLE
CONTINUOUS RELATIONSHIP BETWEEN
SEVERAL VARIABLES
ALTERNATIVELY A MULTIVARIABLE EMPIRICAL
EQUATION MIGHT BE PRODUCED
PLOTTING DATA
GENERAL POINTS
1. CHOOSE SENSIBLE, EASILY
UNDERSTANDABLE SCALES
2. THERE IS NO POINT IN USING SCALE
DIVISIONS WHICH ARE MUCH
SMALLER THAN THE PROBABLE
UNCERTAINTY OF THE DATA.
PLOTTING DATA
GENERAL POINTS
3. THE USE OF MAJOR AND MINOR
DIVISIONS IN SCALES MAY HELP IN AN
EASIER COMPREHENSION OF SCALES.
4. SUPPRESS ZERO IF IT ALLOWS A
BETTER SCALE TO BE USED,
PROVIDING ZERO SUPPRESSION DOES
NOT MAKE THE SIGNIFICANCE OF THE
PLOT LESS OBVIOUS
PLOTTING DATA
GENERAL POINTS
5. IF THE NUMERICAL RANGE OF THE
DATA SPANS OVER SEVERAL DECADES,
TRY USING A LOGARITHMIC SCALE.
6. IN CASE OF LARGE PROBABLE ERRORS
IN DATA POINTS, THE ERROR RANGE
ABOUT EACH DATA POINT MUST ALSO
BE PLOTTED.
PLOTTING DATA
GENERAL POINTS
7. IF A (STRAIGHT) LINE GRAPH SHOULD
THEORETICALLY PASS THROUGH THE
ORIGIN, IT SHOULD NOT BE
AUTOMATICALLY DRAWN THROUGH
THE ORIGIN.
8. IF IT IS NECESSARY TO EXTRAPOLATE
ONE VARIABLE, X, TOWARDS INFINITY, IT
IS BETTER TO REARRANGE THE
VARIABLE AS 1/X AND EXTRAPOLATE TO
ZERO.
PLOTTING DATA
GENERAL POINTS
9. WHERE THERE IS A SCATTER OF THE
PLOTTED POINTS, AND THE PROBABLE
ERROR IS LARGE
IT IS USUALLY BETTER TO PLOT THE
ERROR RANGE ABOUT EACH POINT, AND
THEN TO PLOT A SMOOTH CURVE
THROUGH EACH ERROR RANGE,
RATHER THAN TO DRAW A OSCILLATING
CURVE THROUGH EACH POINT
PLOTTING DATA
GENERAL POINTS
PLOTTING DATA
GENERAL POINTS
10. WHERE ONE VARIABLE, y, IS
APPROXIMATELY PROPORTIONAL TO A
SECOND VARIABLE x, PLOT y/x versus x
y y/x
x x
RECOMMENDED
PLOTTING MORE THAN TWO VARIABLES
1. Two dependent variables, y and z, versus one
independent variable, x; i.e., y = f (x) and z = g (x)
Use
a) either two separate plots as
y z
x x
b) or one plot with multiple scales as
y z
x
PLOTTING MORE THAN TWO VARIABLES
1. One dependent variable, z, versus two independent
variables, x and y; i.e., z = f (x,y)
Possible representations are:
a) Family of curves
b) Perspective (3D) plots
c) Contour plots
d) Carpet plots
A. FAMILY OF CURVES
One of the independent variables (say y) is
picked up as a parameter and a series of curves
zi = f (x,yi) are drawn for fixed values of yi
(i=1,2,…, n).
y3 y2 y1
x
B. PERSPECTIVE (3D) PLOTS
One of the independent variables (say y) is picked
up as a parameter and a series of curves zi = f (x,yi)
are drawn in a perspective view for fixed values of
yi (i=1,2,…, n).
z
x
y1
y2
y3
y
C. CONTOUR PLOTS
The dependent variable, z, is picked up as a
parameter and a series of curves f (x,y) = zi
(constant) are drawn in y versus x plane for
fixed values of zi (i=1,2,…, n).
z3 z2 z1
x
D. CARPET PLOTS
Both independent variables are treated as
parameters and two sets of curves are drawn
for their fixed values
z
y3
y2
y1
x2 x3
x1
CURVE FITTING
THIS IS THE PROCESS WHEREBY AN
EQUATION IS FOUND THAT FITS EXISTING
EXPERIMENTAL DATA
ADVANTAGES
compared with numerical data, an equation is
a) MORE EASILY COMMUNICATED
b) MORE EASILY MANIPULATED
c) MORE READILY USED TO PREDICT
BEHAIVIOUR IN A SLIGHTLY
DIFFERENT SITUATION
PROCEDURE FOR CURVE FIITING
1. BE FAMILIAR WITH THE FORMS
OF COMMONLY OCCURRING
LAWS
e.g. POWER LAW y Cx m
EXPONENTIAL LAW y Ce mx
SKETCH THE FUNCTION SO THAT
YOU ARE SURE IT FOLLOWS THE
DESIRED TRENDS
2. TRY TO LINEARIZE BY :
a) TRANSPOSITION
b) TAKING LOGARITHMS
3. IF A STRAIGHT LINE IS OBTAINED
THEN APPLY LEAST SQUARES FIT
EXAMPLES
y Cx m
Ln( y ) Ln(C) m .Ln( x )
y’
y A Bx
' '
y ' A Bx'
x’
y Ln( y ) A Ln(C) B m x Ln( x )
' '
EXAMPLES
y Ce mx
Ln( y ) Ln(C) mxLn(e)
Ln( y ) Ln(C) mx
y A Bx
' '
y Ln( y ) A Ln(C) B m x x
' '
EXAMPLES
x
y PLOT (1/y) versus (1/x)
a bx
y a bx cx 2
PLOT
(y-y1)/(x-x1) versus (x)
x
y c PLOT
a bx (x-x1) /(y-y1) versus (x)
EXAMPLES
bx cx
y ae
2
PLOT
y 1
x x1
Ln versus x
y1
EXAMPLES
bx
y 1e
PLOT
1
Ln
1 y
versus x
EXAMPLES
y ab PLOT y versus 1/x
x
y ab x PLOT y versus x
Transformation to Linear Representation:
The followings are plotting methods that may be used to produce straight
lines on graph paper for several different types of functions:
1.y = ax + b y versus x on linear paper
slope = a; y-intercept (@ x=0) = b
2.y = axb z = log y versus log x on log-log paper
slope = b; z-intercept (@ log x = 0) = log a
3.y = aebx z = log y versus x on semi-log paper
slope = b log e; z-intercept (@ x=0) = log a
4.y = x/(a+bx) z=1/y versus 1/x on linear paper
slope = a; z-intercept (@ 1/x=0) = b
5.y = a + bx + cx2 z=(y-y1)/(x-x1) vs. x on linear paper
slope = c; z-intercept (@ x=0) = b+cx1
6.y = c+x/(a+bx) z=(x-x1)/(y-y1) vs. x on linear paper
slope = b+(b2x1/a); z-intercept (@ x=0) = a+bx1
2
7.y = a ebx+cx z = log[(y/y1)1/(x-x1)] vs. x on semi-log p.
slope = c log e; z-intercept (@ x=0)=b+cx1loge
8.y = 1-e-bx z = log[1/(1-y)] vs. x on semi-log paper
slope = b; z-intercept (@ x=0) = 0
9.y = a + b/x y versus 1/x on linear paper
slope = b; y-intercept (@ x=0) = a
10.y = a + b x1/2 y versus on linear paper
slope = b; y-intercept (@ x=0) = a