intechnguВестник НГУ. Серия: Информационные технологииVestnik NSU. Series: Information Technologies1818-79002410-0420НГУ10.25205/1818-7900-2024-22-2-68-78intechngu-272Research ArticleСтатьиИсследование архитектур нейронных сетей для определения концентраций газов по спектрамStudy of Neural Network Architectures for Determining Gas Concentrations by SpectraЩелоковА. Е.ShchelokovA. E.

Щелоков Александр Евгеньевич, магистрант кафедры компьютерных технологий факультета информационных технологий; инженер-программист 

Новосибирск

Alexander E. Shchelokov, Master’s Degree Student, Department of Computer Technologies, Faculty of Information Technology; Software Engineer

Novosibirsk

azarion123@gmail.comБудниковК. И.BudnikovK. I.

Будников Константин Иванович, кандидат технических наук, доцент кафедры компьютерных технологий факультета информационных технологий; старший научный сотрудник, заведующий тематической группы

Новосибирск

Konstantin I. Budnikov, PhD in Computer Science, Associate Professor, Department of Computer Technologies, Faculty of Information Technologies; Senior Researcher, Head of the Thematic Group

Novosibirsk

budnikov@iae.nsk.suНовосибирский государственный университет; Институт автоматики и электрометрии СО РАНРоссияNovosibirsk State University; Institute of Automation and Electrometry SB RASRussian Federation2024101020242226878Copyright © Щелоков А.Е., Будников К.И., 20242024Щелоков А.Е., Будников К.И.Shchelokov A.E., Budnikov K.I.Данная работа распространяется под лицензией Creative Commons Attribution 4.0.This work is licensed under a Creative Commons Attribution 4.0 License.https://intechngu.elpub.ru/jour/article/view/272

Исследованы нейронные сети разных архитектур с целью определения концентраций газов по спектрам, полученным с помощью оптического эмиссионного газоанализатора, который измеряет спектр электромагнитного излучения, испускаемого газами при возбуждении электрическим разрядом. Нейронная сеть обучается на данных из лаборатории по оптической спектроскопии и способна предсказывать концентрации газов по спектрам с высокой скоростью. Исследования касались архитектур глубокой нейронной сети со сверточными и рекуррентными слоями. Сверточные слои выделяют особенности спектров, а рекуррентные слои учитывают последовательную структуру данных. Качество работы нейронной сети оценивали по коэффициенту детерминации R2 , а сравнение между сетями проводили по показателю RMSE между предсказанными и реальными концентрациями газов.

A study of a number of neural networks of different architectures for determining gas concentrations from spectra obtained using an optical emission gas analyzer, which measures the spectrum of electromagnetic radiation emitted by gases when excited by an electric discharge, is presented. The neural network is trained on data from the optical spectroscopy laboratory and is able to predict gas concentrations from spectra at high speed. The research concerned the deep neural network architectures with convolutional and recurrent layers. Convolutional layers highlight the features of the spectra, while recurrent layers take into account the consistent structure of the data. The quality of the neural network is evaluated by the R2 coefficient of determination, and the comparison between networks by the RMSE indicator between the predicted and real gas concentrations.

спектральный анализнейронные сетиглубокое обучениеspectral analysisneural networksdeep learningReferencesЧумаков B. M. Определение выбросов одоранта на узле одоризации газораспределительных станций // Газовая промышленность. Сер.: Подготовка, переработка и использование газа. Экспресс-информация. М.: ВНИИЭгазпром, 1987. С. 15–16.Chumakov B. M. Determination of odorant emissions at the odorization node of gas distribution stations. Gas Industry. Ser.: Preparation, processing and use of gas. Express information. Moscow, VNIIEgazprom publ., 1987, pp. 15−16. (in Russ.)Chen Q. et al. Feasibility study on identification of green, black and Oolong teas using nearinfrared reflectance spectroscopy based on support vector machine (SVM) // Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy. 2007. Т. 66, № 3. С. 568–574.Chen Q. et al. Feasibility study on identification of green, black and Oolong teas using near-infrared reflectance spectroscopy based on support vector machine (SVM). Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, 2007, vol. 66, no. 3, pp. 568−574.Круглов В. В., Борисов В. В. Искусственные нейронные сети. Теория и практика. М.: Горячая линия – Телеком, 2001. Т. 382.Kruglov V. V., Borisov V. V. Artificial neural networks. Theory and practice. Moscow, Hotline– telecom, 2001, vol. 382. (in Russ.)Caceres J. O. et al. Application of laser-induced breakdown spectroscopy (LIBS) and neural networks to olive oils analysis // Applied spectroscopy. 2013. Т. 67, № 9. С. 1064–1072.Caceres J. O. et al. Application of laser-induced breakdown spectroscopy (LIBS) and neural networks to olive oils analysis. Applied spectroscopy, 2013, vol. 67, no. 9, pp. 1064−1072.Chengxu L. U. et al. Detection of K in soil using time-resolved laser-induced breakdown spectroscopy based on convolutional neural networks // Plasma Science and Technology. 2018. Т. 21, № 3. С. 034014.Chengxu L. U. et al. Detection of K in soil using time-resolved laser-induced breakdown spectroscopy based on convolutional neural networks. Plasma Science and Technology, 2018, vol. 21, no. 3, p. 034014.Husam Alkinani, Abo Taleb Tuama et. al. Applications of Artificial Neural Networks in the Petroleum Industry: A Review // Proc.SPE Middle East Oil and Gas Show and Conference, March 2019. DOI: 10.2118/195072-MSHusam Alkinani, Abo Taleb Tuama et. al. Applications of Artificial Neural Networks in the Petroleum Industry: A Review. Proc. SPE Middle East Oil and Gas Show and Conference, March 2019, DOI: 10.2118/195072-MSKarrenbach M., Essenreiter R., Treitel S. Multiple Attenuation With Attribute-based NeuralNetworks // Society of Exploration Geophysicists. January 2000. DOI: 10.1190/1.1815827Karrenbach M., Essenreiter R., & Treitel S. Multiple Attenuation With Attribute-based NeuralNetworks. Society of Exploration Geophysicists, January 2000, DOI: 10.1190/1.1815827Aminzadeh F., Maity D., Tafti T. A., Brouwer, F. Artificial Neural Network Based Autopicker ForMicro-earthquake Data // Society of Exploration Geophysicist. Jan. 2011. DOI: 10.1190/1.3627514Aminzadeh F., Maity D., Tafti T. A., & Brouwer F. Artificial Neural Network Based Autopicker ForMicro-earthquake Data. Society of Exploration Geophysicist, January 2011, DOI: 10.1190/1.3627514Dashevskiy D., Dubinsky V., Macpherson J. D. Application of Neural Networks for Predictive Control in Drilling Dynamics // Society of Petroleum Engineers. Jan. 1999. DOI:10.2118/56442-MSDashevskiy D., Dubinsky V., & Macpherson J. D. Application of Neural Networks for Predictive Control in Drilling Dynamics. Society of Petroleum Engineers, January 1999. DOI:10.2118/56442-MSGhahfarokhi P. K., Carr T., Bhattacharya S., Elliott J., Shahkarami A., Martin K. A FiberOpticAssisted Multilayer Perceptron Reservoir Production Modeling: A Machine Learning Approach in Prediction of GasProduction // Proc. Unconventional Resources Technology Conference, August 9, 2018. DOI: 10.15530/urtec-2018-2902641Ghahfarokhi P. K., Carr T., Bhattacharya S., Elliott J., Shahkarami A., & Martin K. A Fiber-OpticAssisted Multilayer Perceptron Reservoir Production Modeling: A Machine Learning Approach in Prediction of GasProduction. Proc. Unconventional Resources Technology Conference, August 9, 2018, DOI: 10.15530/urtec-2018-2902641Gholamreza Moradi, Majid Mohadesi, Mohammad Reza Moradi. Prediction of wax disappearance temperature usingartificial neural networks// Journal of Petroleum Science and Engineering. 2013. Vol. 108. P. 74–81. http://dx.doi.org/10.1016/j.petrol.2013.06.003Gholamreza Moradi, Majid Mohadesi, Mohammad Reza Moradi Prediction of wax disappearance temperature usingartificial neural networks. Journal of Petroleum Science and Engineering, 2013, vol. 108, pp. 74–81, http://dx.doi.org/10.1016/j.petrol.2013.06.003LeCun Y., Bengio Y., Hinton G. Deep learning // Nature. 2015. Т. 521, № 7553. С. 436–444.Lecun Y., Bengio Y., Hinton G. Deep learning. Nature, 2015, vol. 521, no. 7553, pp. 436−444. (in Russ.)Аггаравал Ч. Нейронные сети и глубокое обучение: учебный курс. СПб.: Диалектика, 2020. 752 с.Aggarwal Ch. Neural networks and deep learning: a training course. Dialectics LLC, St. Petersburg, 2020, 752 p.

The authors declare that there are no conflicts of interest present.