Warning: include_once(../article_type.php): failed to open stream: No such file or directory in /home/suxhorbncfos/public_html/omcij/OMCIJ.MS.ID.555717.php on line 112
Warning: include_once(): Failed opening '../article_type.php' for inclusion (include_path='.:/opt/alt/php56/usr/share/pear:/opt/alt/php56/usr/share/php') in /home/suxhorbncfos/public_html/omcij/OMCIJ.MS.ID.555717.php on line 112
Forecasting the Composition of Catalysts Of Hydrogenization of Oil and Fat
Sattarov KK and Mazhidov KH*
Doctor of Technical Sciences, Russia
Submission: July 28, 2018; Published: August 03, 2018
*Corresponding author: Mazhidov KH, Doctor of Technical Sciences, Russia, Tel: 998905140640; Email: email@example.com
How to cite this article: Sattarov KK, Mazhidov KH. Forecasting the Composition of Catalysts Of Hydrogenization of Oil and Fat. Organic & Medicinal
Chem IJ. 2018; 7(4): 555717. DOI: 10.19080/OMCIJ.2018.07.555717
The technique of the theory of pattern recognition in predicting the chemical-technological characteristics of alloyed promoted nickel-copper-aluminum stationary catalysts was used. The most effective promoting metals are selected. An increase in the activity and other hydrogenating properties of the hydrogenation catalysts for fats and oils has been achieved.
Keywords:Alloy catalysts; Promoting metals; Methods for selecting promoters; The use of the theory of pattern recognition; Activity prediction; Chemical and technological characteristics
Modern concepts of heterogeneous catalysis link the catalytic activity of solids with structure, chemical composition, energy of stabilization of the crystal field, acid-based surface properties, etc., which together constitute the theoretical basis for the selection of catalysts for reactions of a certain type or group of chemical transformations of the same type .The search for suitable catalysts to accelerate certain reactions based on current knowledge of the range of substances that are potential catalysts and general ideas about the mechanisms of their action is attributed to the tasks of predicting catalysts [1-3]. The theoretical direction cannot yet claim a successful prediction of catalysts, but quantum-mechanical concepts of heterogeneous catalysis play an important role in predicting catalysts using the methods of pattern recognition theory [4-6]. Mathematic-heuristic direction is the development of classical correlation methods and in its essence is a set of statistical computer methods of information processing, generalized by the theory of pattern recognition.
The solution of the recognition problem is based on the initial empirical material - the “training sample”.
The simplest problem solved by the methods of the pattern recognition theory [7,8]. is the selection of a promoter or other modifying additive to a known catalyst. The initial information for the creation of the training sample was the literature and patent data on the composition and properties of heterogeneous catalysts for hydration reactions. The latter were subdivided into 5 classes in accordance with the VA Roiter classification
. The results of recognition of the class of catalysts for the
hydrogenation of fats as a function of the counter class (Table 1) are listed in Table 2. It should be noted that if the catalysts of one class from catalysts of another class were less than 70%, they were considered included in the latter. Accordingly, recognition within 70-100% means an independent existence of the class. The analysis of classes carried out in  (Table 1). allowed us to state that all hydrogenation catalysts are practically suitable for the hydrogenation of fats, with the exception of typical catalysts for the hydrogenation of the acetylene bond . For the prediction of effective promoters of alloyed nickel-aluminum catalysts for hydrogenation of fats, the methods of the pattern recognition theory .
were used. The solution of recognition problems is based on
the initial experimental material - the “training sample”, each
object of which xij (j = 1 ... M) is described by a system of signs
(i = 1 ... N). Thus, the training sample can be represented by a
matrix of dimension MxN, where M is the number of objects (in
our case, catalysts), and N is the number of catalyst features:
Each matrix represents a specific array, a class of catalysts.
In the algorithm “Leader” used, the training sample is divided
into 2 arrays |х_ij | and |х ̅ ij |, that is, two classes of catalysts
-A and А ̅. In each case, the selection of specific training sample
objects and their number, as well as the division of the entire
sample into two (in this case) array is carried out in advance,
based on the task and on the basis of studying both literary and
own experimental material. In our case, the data are indicative
of the component (elemental) composition of the catalysts.
And as the sample objects, catalysts of several hydrogenation
classes are used (Table 1). As a recognition algorithm, a binary
linear classifier of images, the so-called threshold logic element,
was used. In this case, the affiliation of catalysts to one of the
two arrays (images) - A or А ̅ - is determined by calculating
the distances from the objects (individual catalysts) to the
centers of alternative classes - standards A and А ̅, represented
in the multidimensional Euclidean space of the characteristics
describing these objects. We denote the volume of the training
sample by the index M. Then the volume of the class A is equal to
MA, and the volume of the class А ̅ is equal to MА ̅, that is:
The set of objects of classes A and А ̅ can be expressed in
terms of the sums:
Then the coordinates of the centers of classes A and А ̅, that
is, the coordinates of the standards А* and А ̅^*are equal to:
Distance between the object and the standard in the general
case is calculated by the formula:
Comparing the pairwise distances from the object to the
standards, determine the belonging of the given object (of the
given catalyst) to one or another class (of two pairwise matched
The object belongs to class A.
The object belongs to class A ̅.
The reliability of recognition (classification by two classes) is
estimated by the proportion of correct estimates (classifications)
in arrays A and A ̅.
Denote +i- the number of correct classifications of objects
in classes A and A ̅, respectively. Then, taking into account the
volumes of the classes MA and MA ̅, we find the fractions of the
correct classifications of objects in these classes:
The average level of image recognition is half the sum of
recognition levels in each alternative class:
If we now consider a new object that is not included in the
training sample - object y (y1 ... yn) then, as in the previous cases,
the new object y belongs to the class A - yi A, if the following
condition is met:
In the opposite case, the object y belongs to the class A ̅ (yi
∈ A ̅).
A measure of the prospects of each of the objects is the
distance to the “Leader” - a point in the multidimensional space
of characteristics describing objects. We denote this distance as
П = рх, л.
Conditions of object leadership:
As already noted, the distance from the object to the standard
allows us to detect the similarity of the object with other
catalysts of this class. To strengthen the definition of an object’s
belonging to a particular class, the conditions for distinguishing
this object from catalysts of another class are also introduced.
These conditions consist in the requirement of proximity to a
non-reference standard of this class. The distance of the leader
from the standards of the opposing classes should be minimum
and maximum, respectively, as expressed by the above formulas.
Obviously, the prospectivity of this object is the greater, the
smaller the distance between the object and the leader П =
рх, л. According to the data of . the predictive ability of the
technique for solving the problems of selecting catalysts is 75-
100%. In this paper, the version of the “Leader-FS” program was
used, which has a blurred classification block. This method was
carried out as follows: At the first stage of the work, we used
the pattern recognition technique to predict the qualitative
and quantitative composition of alloyed nickel-aluminum and
copper catalysts. For the forecasting, the data obtained in our
experiments, as well as the literature and patent data . were
used. Experimental information is presented in Table 3. In which
the indices y1 denote the activity of the catalyst of the given
composition, and the indices x1-x7 the component composition
of the catalyst (alloy) in weight. percentages: x1-nickel, x2-
copper, x3-aluminum, x4-germanium, x5-rhodium, x6-rhenium,
x7-ruthenium. The entire array of catalysts by activity (y) was
divided into two classes. Class A-11 objects, class A ̅-13 objects.
The conditions for the belonging of catalysts to this or that class:
When processing the data in Table 3, the chemical
composition was normalized. Condition of normalization of the
where x*i is a normalized trait; xi is the given (i-th) sign in the
initial scale of measurements, xmax is the maximum value of the
given characteristic in the analyzed series. As a result of solving
pattern recognition problems, the recognition of highly active
catalysts (class A ̅) was 100%, and for anticlass A - 79% (Figure
1). Further, as a result of the evaluation of the perspectivity of the
learning sample objects, a correlation was obtained between the
activity and the calcula ted catalyst prospects. Linear correlation
is shown in Figure 2. In this graph, three catalysts (alloys) were
identified, for which the proposed chemical composition was
compared in Table 4 and the data presented generally indicate
that the activity of the catalyst increases significantly when
the alloy is promoted by germanium, as well as germanium +
rhodium and germanium + ruthenium systems. Obviously, the
last two catalysts of Table 4 are the most “promising” ones. But
the fulfilled forecast is based only on the activity of the catalysts.
The probable efficiency of the use of selected promoters in
the composition of nickel-copper-aluminum alloy stationary
catalysts for the improvement of the chemical-technological
characteristics of the catalysts studied was previously identified
based on the use of elements of the theory of pattern recognition.
The calculated promising calculated (predicted) activity and