Показано, что одной из важнейших характеристик взрывозащищенного оборудования является его температурный класс. Установлено, что на определение температурного класса влияет температура самовоспламенения. Ввиду сложности экспериментального нахождения этого показателя пожаровзрывобезопасности предложено использовать методику прогнозирования пожароопасных свойств продуктов нефтепереработки на основе молекулярных дескрипторов и искусственных нейронных сетей. Верификацию предлагаемой методики проводили на примере прогнозирования температуры самовоспламенения кислородсодержащих предельных углеводородов, в частности сложных эфиров пропионовой кислоты. Установлено, что среднее значение абсолютной погрешности при прогнозировании составило 6,25 °С, а относительной погрешности - 2 %. Сделан вывод, что методика прогнозирования пожароопасных свойств продуктов нефтепереработки позволяет определять основные показатели пожаровзрывоопасности свойств веществ с приемлемой точностью, что дает возможность установить температурный класс взрывозащищенного электрооборудования.
The article raises the topical issue - the lack of physical and chemical properties of new synthesized substances. These properties will allow supervisors to develop fire safety systems at security facilities. The efficiency of such systems is achieved by eliminating the combustible environment or the ignition source. Using the example of esters of butyric acid, which are used practically in all areas of industry and produced according to reference data in the amount of more than several tens of millions of tons per year, it was possible to predict one of the most important fire hazard properties of a substance - the self-ignition temperature, using the technique for predicting the fire hazard properties of refined products based on molecular descriptors and artificial neural networks. The proposed methodology is implemented with the help of the author’s computer program “NeuroPacket KDS 1.0”. The program “NeuroPacket KDS 1.0” allows you to: download and view databases containing chemical compound structures and their properties; to correlate the input data; statistically evaluate the resulting models; use the obtained neuronet models to predict the properties of substances without conducting a complex experiment. This approach to predicting the fire hazard property of oil refining products describes the structure of the molecule with the help of molecular descriptors and establishes quantitative correlations between the values found using artificial neural networks. Based on some reference data, data was verified. Analysis of the results obtained showed that the average relative error does not exceed 3 %, which is a good indicator. In addition, the autoignition temperature of esters of butyric acid was predicted, information on which is absent in the reference and regulatory literature. This makes it possible to build on the values obtained in the development of fire safety systems. Based on the obtained results on the self-ignition temperature of the substance, there were determined the temperature classes of the explosion-proof electrical equipment, which, on the whole, ensures fulfillment of item 4, article 50 of the “Technical Regulations on Fire Safety Requirements” (Federal Law No. 123). It is also worth noting that the methodology for predicting the fire-hazardous properties of oil refining products based on the use of molecular descriptors and artificial neural networks allows us to conclude that this technique can be used to predict other fire-hazardous properties of organic substances.
The authors declare that there are no conflicts of interest present.