Прогнозирование основных показателей пожаровзрывоопасности органических соединений с помощью дескрипторов и искусственных нейронных сетей, используемых в расчете пожарного риска

Abstract

According to Government Resolution No. 290 On the Federal State Fire Supervision dated 12.04.2012 for the purposes of monitoring compliance with the legal entity, individual entrepreneur, etc. of fire safety are carried out routine checks. This type of oversight activities conducted after three years from the date of registration in a tax body or three years since the last scheduled inspection. To avoid this event in accordance with article 6 of Federal Law No. 123 regulations on fire safety requirements (hereinafter - the Technical regulation) is permitted to carry out the calculation of the assessment of fire risk. If the calculated value does not exceed the allowable value set by the Technical regulations, it is considered that the facility meets fire safety requirements. Every year the number of organic compounds is increased by 250-300 thousand, details of which there is no, and the calculation of the magnitude of fire risk requires knowledge of the properties of substances as lower heating value, specific speed of burnout, the linear speed of burnout. Experimental determination of physico-chemical properties of substances is associated with significant technical difficulties, economic and time costs. Therefore, a promising method for determination of the fire performance method is based on the use of descriptors and artificial neural networks. The program KDS 1.0 handles pre-computed descriptors of the substance and predicts the required property.