Process
monitoring and simulation is an area where
computers are effectively used to optimise
production and maximise profits. There are
innumerable areas where a generalised software
package (or a product) can be effectively used to
yield immediate direct benefits to operating
factories. A software GITA, a Guided
Intelligent Tool for Analysis, has been developed
for this purpose. GITA is a
collection of the following data
correlation packages:
 CURVEFIT: Based on a novel monotonic
transform
 LINEFIT
: Based on linear regression
 QUADFIT
: Based on quadratic
regression
 LOGFIT
: Based on logarithmic
regression
 EXPOFIT
: Based on exponential
regression
 RRMFIT
: Based on RosinRammler
regression
 POLYFIT
: Based on polynomial regression
 POSYFIT
: Based on posynomial regression
 RECIFIT
: Based on reciprocalposynomial regression
 MULTFIT
: Based on multivariable linear and logarithmic
regression
The software module CURVEFIT
in GITA is based on a novel mathematical
technique of monotonic transform developed and
published in international research
journals. It is essentially useful for
determining mathematical correlations between
data, interpolation and extrapolation,
forecasting, developing simple analytical
equations for complex integrals (which can
otherwise be calculated only using advanced
numerical techniques), developing simple
equations for infinite series solutions (which
are encountered in several engineering problems),
parametric studies etc. The software has
excellent graphing capabilities with possible
instant representation in different forms like
Linear, Loglog, Semilog and Inverse Semilog
scales. It also has inbuilt features for
storage and management of data.
The following are some of
the areas where GITA has been successfully
applied:
Particle size
distribution of powders
Classifier
efficiency curves
Cement strength
development curves
Ball mill grinding
rates
Ball mill chamber
sampling
Kiln operation
curves
Coal calorific
value estimation
Valve calibration
curves
Product yield
curves in petroleum isocracking process
Ammonia density
versus pressure and temperature
Drying rate curves
and diffusivity estimation
More accurate
relation than Arrhenius relation for
reaction rate constant versus temperature
Langlier index
estimation for cooling water quality and
scaling
Semivariogram
curves
Release rates of
drugs
Correlations for
psychrometric relations
Filter life
estimation
Compression ratio
of working fluid in heat pumps
Correlations
for predicting the amount in the desired particle size
range in the final product and its moisture content versus
several operating variables
Sale forecasting
for FMCG, pharmaceutical companies etc.
Simpler relation
for Normal probability integral and other
integrals
Simpler relation
for several infinite series solutions
encountered in engineering
A summary of the actual
values of the parameters estimated for typical
applications, which should be found useful by
users in the respective industries, is given
below. The following summary refers to the
predictions obtained with only one of the pack of
seven packages included in GITA, namely CURVEFIT.
The other packages are also useful for
specific applications. It may be noted that
the error in the predictions is extremely
reasonable in all the cases which is indicative
of the power of this software. The actual
graphs showing the predicted curves along with
the data points are also shown for 13 of the 16
applications mentioned below.
The
package includes a module called FASTFIT, which enables
quick determination of the best package from among the
various packages listed earlier. This leads one to the
package most suitable for a particular application.
After running FASTFIT, one can refine the mathematical
equation by going into the best package suitable for the
application.
GITA
also incorporates an automatic inverse determination feature,
wherein mathematical equations can be obtained for predicting
the xvalue for a given yvalue, in other words, x = f ^{1}(y)
is determined.
The
MULTFIT module in GITA is useful for determining
a mathematical equation between several independent variables
and one dependent variable:
y = f (x_{1},x_{2},x_{3},x_{4},...)
The
functional form can be either linear or logarithmic as
follows:
y = a_{0} + a_{1}x_{1}+ a_{2}x_{2}+
a_{3}x_{3}+ a_{4}x_{4}+ ...
y = C x_{1}^{a1} x_{2}^{a2}
x_{3}^{a3} x_{4}^{a4}...
The
latter form mentioned above is especially useful in
dimensional analysis to establish dimensionless correlations
in process studies. Running MULTFIT gives both
the above correlations along with the corresponding average
absolute error values.
