Wyniki wyszukiwania - BazTech

1

Application of neural networks in diagnostics of chemical compounds based on their infrared spectra

Macek-Kamińska K., Stemplewski S.

Ecological Chemistry and Engineering. S

|

2017

|

Vol. 24, nr 1

107--118

EN

The paper presents possibilities of using the so-called „finger-print“ identification method and artificial neural network (ANN) for diagnosis of chemical compounds. The construction of a tool specifically developed for this purpose and the ANN, as well as the required conditions for its proper functioning were described. The identification of chemical compounds was tested in two different ways for proving correctness of the assumptions. First of all, initial studies were carried out with the objective to verify the proper functioning of the developed procedure for IR spectrum interpretation. The second research stage was to find out how the properties of artificial neural networks will satisfy identification or differentiation in case of spectra with very similar structures or for mixtures consisting of several chemical compounds. Interpretation of infrared spectra of mono-constituent substances was successfully performed for both - the training and test data. Interpretation process of infrared spectra of bi-component substances, based on the example of structurally related compounds obstructing identification process, should also be described as positive. The model was able to interpret spectra of mixtures, which were previously registered into the database. Unfortunately, the program is not always able to determine which chemical substances reflect their presence in the infrared spectrum of ternary mixtures. During the research tests, it was also noted that the more complex the structure of a substance being present in the mixture was, the more difficult the interpretation of the spectra to be carry out properly by the program was. On the other hand, positive results were obtained for mixtures of compounds with not so complex structure. It must be emphasized that the results so far are promising and more attention should be paid to them in further studies.

2

Identification of Chemical Compounds Using Neural Networks – the Overfitting Problem

Stemplewski S.

Zeszyty Naukowe. Elektryka / Politechnika Opolska

|

2013

|

z. 69

99--100

EN

A method for diagnostics of chemical compounds based on their spectra in infrared is presented and a C# programme which analyses and indentifies chemical compounds using the above method is described. The overfitting problem in artificial neural networks is also discussed.

3

Wykorzystanie sztucznej sieci neuronowej do diagnostyki związków chemicznych przy pomocy ich widm w podczerwieni

Stemplewski S., Macek-Kamińska K.

Pomiary Automatyka Kontrola

|

2013

|

R. 59, nr 1

38--39

PL

W artykule przedstawiono możliwości zastosowania sztucznej sieci neuronowej w identyfikacji związków chemicznych metodą tzw. „odcisku palca” oraz opisano budowę opracowanego specjalnie do tego celu narzędzia z wykorzystaniem SSN, jak też sprecyzowano wymogi, jakie muszą być spełnione do jej poprawnego funkcjonowania.

EN

This paper presents a combination of the “finger-print” identification method and artificial neural networks (ANN) for effective diagnostics of chemical compounds from their infrared spectra. Identification of chemical compounds on the basis of their IR spectra is a serious problem in absorption spectrophotometry, used in practical chemical analysis. Using ANN to diagnose chemical compounds opens up new abilities for effective identification, not only in terms of speeding the process up but also in view of modeling complex nonlinear signals. A programming tool is developed in Microsoft Visual C# and tested on the basis of one hundred chemical compounds used to teach the ANN. The self-learning ability of ANN is used to construct a relationship between input and output parameters. Using AAN is also possible both to ignore redundant data and those whose impact on the phenomenon is negligible, so it is focused on the input data having a major impact on the modeled process. ANN is used to diagnose one hundred chemical compounds and further studies will be focused on possible expanding the database to include some other compounds.

4

Identification of chemical compounds using neural networks

Stemplewski S.

Zeszyty Naukowe. Elektryka / Politechnika Opolska

|

2012

|

z. 67

111-112

EN

A method for identification of chemical compounds based on their spectra in infrared is presented and a C# programme which analyses and indentifies chemical compounds using the above method is described. A neural network designed specially to solve this problem is also introduced.

5

Development of two-dimensional histograms

Stemplewski S.

Zeszyty Naukowe. Elektryka / Politechnika Opolska

|

2012

|

z. 68

61--62

EN

This article providcs a brief description of methods for developing two-dimensional histograms that allow to accelerate the process ofrecognition objects in the image.

6

Mixtures Identification of Chemical Compounds on the Basis of Their IR Spectra by Artificial Intelligence

Macek-Kamińska K., Sudoł M., Stemplewski S.

Ecological Chemistry and Engineering. A

|

2012

|

Vol. 19, nr 8

973-981

EN

Infrared (IR) spectrometric identification of individual chemical compounds from their mixtures is still a challenging process. Therefore, we developed a method in which we use the IR “Fingerprint” spectra of a particular chemical substance followed by artificial intelligence (AI) – based analysis to correctly characterise components of relatively simple chemical mixtures. We describe here the assembly of tools developed especially for this purpose as well as the artificial neural network design together with the requirements that must be met for its proper functioning. To test our approach, we used a mixture of amphetamine and creatinine which are difficult to identify in mixtures by standard “Fingerprint” rules. The advantages of the artificial neural network approach include the generalisation and adaptation of knowledge by fitting parameter values to change the object characteristics. All this renders the effective identification of a mixture of two substances possible.

PL

W artykule przedstawiono możliwości zastosowania w identyfikacji związków chemicznych metody tzw. odcisku palca oraz sztucznej inteligencji na podstawie widm w podczerwieni. Opisano budowę opracowanego specjalnie do tego celu narzędzia i sztuczną sieć neuronową oraz wymogi, jakie muszą być spełnione do jej poprawnego funkcjonowania. Obecnie stosowane programy użytkowe do identyfikacji związków chemicznych na podstawie ich widm w podczerwieni natrafiają na trudności z poprawną identyfikacją w przypadku mieszanin substancji. W przeprowadzonych badaniach testowych wykorzystano mieszaninę kreatyniny oraz amfetaminy - substancje z którymi obecnie wykorzystywane oprogramowania działające wg zasady Finger-print mają duże trudności. Dlatego też zastosowano sztuczną sieć neuronową, której zalety, takie jak uogólnianie zdobytej wiedzy oraz adaptacja, czyli dopasowania wartości parametrów do zmian charakterystyk obiektu, pozwalają na skuteczną identyfikację w mieszaninie dwóch substancji.

7

Identification of chemical compounds on the basis of their chemical spectra in infrared

Stemplewski S.

Zeszyty Naukowe. Elektryka / Politechnika Opolska

|

2010

|

z. 63

53-54

EN

The article presents methods for identifying chemical compounds on the basis of their spectrums in Infrared. Specifically, transformation of the spectra from absorbing to transmitting ones and vice verse and calculation of the so-called base line and the maxima of the spectra are considered.