CHAPTER 6

MATHEMATICAL MODELLING

6.1 Development of a Mathematical Model

A mathematical model is defined as a representation of the essential aspects of
an existing system (or a system to be constructed) which presents knowledge of that
system in usable form. A mathematical model can take many forms, including but not
limited to dynamic systems, statistical models, differential equations, or game
theoretical models. These and other types of models can overlap, with a given model
involving a variety of abstract structures. Often when engineers analyse a system to be
controlled or optimized, they use a mathematical model. A mathematical model is a
description of the behaviour of the system using mathematical language. The process of
developing a mathematical model is termed as mathematical modelling (also written as
modelling). Mathematical models are extensively used in the natural sciences (such as
economics, psychology, sociology and political science) and also by physicists,
engineers, statisticians, operations research analysts and economists. In analysis,
engineers can build a descriptive model of the system as a hypothesis of how the system
could work. Similarly, in control of a system, engineers can try out different control
approaches in simulation.
Actual model is the set of functions that describe the relations between the
different variables. A mathematical model usually describes a system by set of variables
and a set of equations that establish relationships between the variables. The values of
the variables can be practically anything; real or integer numbers. The Theory of
Experimentation as suggested by Hilbert Schenck is a good approach of representing
the response of any phenomenon in terms of proper interaction of various inputs of the
phenomenon. This approach finally establishes an experimental data based models for
any phenomenon. There are many phenomenon which are very complex and it is very
difficult to fabricate large number of experimental set-ups to gather data in a planned
way. The concept of Least Square Multiple Regression curves has been used to develop
the models, hereafter referred as mathematical model.
The objective of any researcher, experimenter or designer of any machine or
manufacturer or entrepreneur is to arrange the optimized inputs to get targeted response.
The objective is only achievable by formulation of such models.
6.2 Formulation of Model
To formulate the mathematical model there are various technique available. In
present experimental work Buckingham’s pi theorem is used for model development.
The dependent and the independent variables are reduced and clubbed in a single
dimension less mathematical equation in which both the sides of equation is balanced.
The Eq.5.19 shows the dimensionless mathematical equation. Dimensional equations so
obtained can be formulated into model using multiple regression analysis. Multiple
Regression analysis is a statistical tool that utilizes the relation between two or more
quantitative variables so that one variable can be predicted from another. The variable to
be predicated is called the output or response/dependent variable. The variable
predicting is called the input or independent variable.
With the use of this procedure the different dimensional equations and models
can be formulated for the different outputs. The formulated models can be evaluated on
the basis of Correlation and Root Mean Square Error between the computed values and
the estimated values.
A mathematical model is the representation of the selected phenomenon of the
system of interest. A model is similar to but simpler than the phenomenon it simulates.
One purpose of a model is to enable the analyst to predict the effect of changes to the
phenomena. On one hand, a model has to be a close approximation to the real system
and incorporates most of its salient features. On the other hand, it should not be so
complex that it is impossible to understand and experiment with it.

6.3 Formulation of Experimental Data Based Model

In the mathematical equation.5.19, there are seven independent pi terms (π02, π03, π04, π05,
π06, π07, π08) and one dependent pi terms (π 01).

Independent π term = (π02, π03, π04, π05, π06, π07, π08)

Dependent π term = (π01)
The dependent π term is the function of independent π terms,
Cutting tool temperature =(π02, π03, π04, π05, π06, π07, π08)
A probable exact mathematical form for this phenomenon could be the empirical
relationships in between dependent dimensionless ratio and independent dimensionless
ratio are assumed to be exponential.
T = K × (π02) a × (π03) b × (π04) c × (π05) d× (π06) e× (π07) f× (π08) g
 The values of exponential a, b, c, d are established, considering exponential
relationship between dependent pi term T and Independent π terms π02, π03, π04, π05, π06,
π07, π08 independently taken one at a time, on the basis of data collected through classical
experimentation.
In Eq.6.1 there are eight unknown values K, a, b, c, d, e, f and g. The curve
fitting is required, for those five sets of values of (π02, π03, π04, π05, π06, π07, π08) required. In
the experimental work all 27 readings were taken. All 27 readings are included in this
model formulation.

6.3.2 Establishment of Relationship for π Term

T (π01) = K × (π02) a × (π03) b × (π04) c × (π05) d× (π06) e× (π07) f× (π08) g
Taking log on both the side of equation for π 04, getting five unknown terms in the
equation.
Log (π01) = Log (C) × Log (π02) a × Log (π03) b × Log (π04) c × Log (π05) d × Log (π06) e ×
Log (π07) f × Log (π08) g
Log (π01) = Log (C) + aLog (π02) + bLog (π03) + cLog (π04) + dLog (π05) + eLog (π06) +
fLog (π07) + gLog (π08) Eq.6.2
Now,
Take Log (π01) = Y, Log C = K, Log (π02) = A, Log (π03) = B, Log (π04) = C,
Log (π05) = D, Log (π06) = E, Log (π07) = F, Log (π08) = G
Putting the values in Eq.6.2, the Eq.6.3 becomes
Y = K + a×A + b×B + c×C + d ×D + e×E + f×F +g×G Eq.6.3
Eq.6.3 is a regression equation of ‘Y’ on A, B, C, D, E, F and G in an n dimensional co-
ordinate system. This represents a regression hyper plane.
To determine the regression hyper plane, determining a, b, c, d, e, f and g in Eq.6.3 so
that,
∑Y = nK + a ∑ A + b ∑ B + c ∑ C + d ∑ D
∑ (Y×A) = nK ∑ A + a ∑A2 + b ∑ (A×B) + c ∑ (A×C) + d ∑ (A×D)
∑ (Y×B) = nK ∑ B + a ∑ (A×B) + b ∑ B2 + c ∑ (B×C) + d ∑ (B×D)
∑ (Y×C) = nK ∑ C + a ∑ (A×C) + b ∑ (B×C) + c ∑ C2 + d ∑ (C×D)
∑ (Y×D) = nK ∑ D + a ∑ (A×D) + b ∑ (B×D) + c ∑ (C×D) + d ∑ D2 Eq.6.4

In the above set of equations the values of the multipliers of K, a, b, c, d, e, f and

g are substituted to compute the values of the unknowns (viz. K, a, b, c, d, e, f and g).
The values of the terms on LHS and the multipliers of K, a, b, c, d, e, f and g in the set
of equations are calculated and tabulated in the Table C-2 to C-4 (Appendix C). After
substituting these values in the Eq.6.4 one will get a set of 8 equations, which are to be
solved simultaneously to get the values of K, a, b, c, d, e, f and g. The above equations
can be verified in the matrix form and further values of K, a, b, c, d, e, f and g can be
obtained by using matrix analysis.
[Y]=[A][X]
The matrix method of solving these equations using ‘MATLAB’ is given below.
A = 8 x 8 matrix of the multipliers of K, a, b, c and b
Y = 8 x 1 matrix of the terms on LHS and
X = 8 x 1 matrix of solutions of values of K, a, b, c and d

∑Y nK A B C D K

∑YA A AA AB AC AD a

∑YB B AB BB BC BD b
= =
∑YC C AC BC CC CD c

∑YD D AD BD CD DD d

[ X ] = [ Y ] [ A ]-1
[X]=[A]\[Y]
The values of the terms in the matrix is shown in the Apendix-II in table no.II.2
After putting all the values in the above matrix, it becomes

3.04648
60 0.156108 2.531153 0.176364 -3.25672 K
0.47399
4
0.156108 0.044145 0.394783 0.016121 -0.5084 a
7.71289
2.531153 0.394783 6.424632 0.428074 -8.24327 b
7 = =
0.176364 0.016121 0.428074 0.066285 -0.57434 c
0.53795
9
-3.25672 -0.5084 -8.24327 -0.57434 10.60623 d
-9.92153
Using MATLAB, [ X ] = [ A ] \ [ Y ], after solving X matrix with K and indices a, b, c,
and d are as follows
 K 0.000
a 0.0222
b 0.2734
c 0.1692
d - 0.7126

In the above matrix value of K is log C, for the value of C, take Antilog (K),
Therefore C = 1.
Hence the Eq.6.1 for independent π term becomes
Nu = K × (π01) a × (π02) b × (π03) c × (π05) d
Uftg a D b
Nu = C × ( ) ×( ) × Rec × Prd
Umf Dp

Uftg 0.0222 D 0.2734

Nu = 1 × ( ) ×( ) × ℜ0.1692 × Pr - 0.7126 Eq.6.5
Umf Dp

Eq.6.5 is the generalized correlation applicable for the particle size ranging from 200
µm to 425 µm in swirl induced fluidized bed when tangential air is supplied.

6.3.2.2 For Individual Bed Particle Size 200 µm

[Y]=[A][X]

∑Y nK A B C D K

∑YA A AA AB AC AD a

∑YB B AB BB BC BD b
= =
∑YC C AC BC CC CD c

∑YD D AD BD CD DD d

[ X ] = [ Y ] [ A ]-1
[X]=[A]\[Y]
The values of the terms in the matrix is shown in the Apendix-II in table no.II.2
After putting all the values in the above matrix, it becomes

3.06234 20 0.152849 2.698970 0.004652 -3.22568 K

 9

0.46632
2 0.152849 0.042825 0.412534 -0.01102 -0.49778 a

8.26518 2.698970 0.412534 7.284439 0.012556 -8.79005 b

= =
8
0.004652 -0.01102 0.012556 0.016453 -0.01513 c
0.01706
2 -3.22568 -0.49778 -8.79005 -0.01513 10.60686 d

-9.97351

Using MATLAB, [ X ] = [ A ] \ [ Y ], after solving X matrix with K and indices a, b, c,

and d are as follows

K 0.000
a 0.0315
b -3.5372
c 0.1973
d - 3.8698

In the above matrix value of K is log C, for the value of C, take Antilog (K),
Therefore C = 1.
Hence the Eq.6.1 for independent π term becomes
Nu = K × (π01) a × (π02) b × (π03) c × (π05) d
Uftg a D b
Nu = C × ( ) ×( ) × Rec × Prd
Umf Dp

Uftg 0.0315 D −3.5372

Nu = 1 × ( ) ×( ) × ℜ0.1973 × Pr −3.8698 Eq.6.6
Umf Dp

Eq.6.6 is the correlation applicable for the bed particle size 200 µm in swirl induced
fluidized bed when tangential air is supplied.

6.3.2.3 For Individual Bed Particle Size 300 µm

[Y]=[A][X]
[ X ] = [ Y ] [ A ]-1
[X]=[A]\[Y]
The values of the terms in the matrix is shown in the Apendix-II in table no.II.2
After putting all the values in the above matrix, it becomes

3.04764 20 0.156233 2.522879 0.184426 -3.25671 K

0.47457 0.156233 0.044165 0.394157 0.017063 -0.50882 a

7.68882 2.522879 0.394157 6.364917 0.465285 -8.21629 b

= =
0.56469 0.184426 0.017063 0.465285 0.050334 -0.60061 c

-9.92529 0.184426 -0.50882 -8.21629 -0.60061 10.60618 d

Using MATLAB, [ X ] = [ A ] \ [ Y ], after solving X matrix with K and indices a, b, c,

and d are as follows

K 0.000
a 0.0393
b 47.8662
c 0.1587
d 36.1556

In the above matrix value of K is log C, for the value of C, take Antilog (K),
Therefore C = 1.
Hence the Eq.6.1 for independent π term becomes
Nu = K × (π01) a × (π02) b × (π03) c × (π05) d
Uftg a D b
Nu = C × ( ) ×( ) × Rec × Prd
Umf Dp

Uftg 0.0393 D 47.8662

Nu = 1 × ( ) ×( ) × Re0.1587 × Pr 36.1556 Eq.6.7
Umf Dp

Eq.6.7 is the correlation applicable for the bed particle size 300 µm in swirl induced
fluidized bed when tangential air is supplied.

6.3.2.4 For Individual Bed Particle Size 425 µm

[Y]=[A][X]
[ X ] = [ Y ] [ A ]-1
[X]=[A]\[Y]
The values of the terms in the matrix is shown in the Appendix-II in table no.II.2
After putting all the values in the above matrix, it becomes

3.02945
1
20 0.159241 2.371611 0.340013 -3.25663 K
0.48108
8 0.159241 0.045445 0.377659 0.042330 -0.51860 a

7.18467 2.371611 0.377659 5.624539 0.806380 -7.72346 b

= =
6
0.340013 0.042330 0.806380 0.132065 -1.10728 c
1.03212
3 -3.25663 -0.51860 -7.72346 -1.10728 10.60564 d

-9.86581

Using MATLAB, [ X ] = [ A ] \ [ Y ], after solving X matrix with K and indices a, b, c,

and d are as follows

K 0.000
a 0.0066
b 59.70
c 0.0860
d 42.5567

In the above matrix value of K is log C, for the value of C, take Antilog (K),
Therefore C = 1.
Hence the Eq.6.1 for independent π term becomes
Nu = K × (π01) a × (π02) b × (π03) c × (π05) d
Uftg a D b
Nu = C × ( ) ×( ) × Rec × Prd
Umf Dp

Uftg 0.0066 D 59.70

Nu = 1 × ( ) ×( ) × Re0.0860 × Pr 42.5567 Eq.6.8
Umf Dp

Eq.6.8 is the correlation applicable for the bed particle size 300 µm in swirl induced
fluidized bed when tangential air is supplied.
All the MATLAB scripts are added Appendix-