ResearchBib Share Your Research, Maximize Your Social Impacts
Sign for Notice Everyday Sign up >> Login

Probabilistic approach to the modeling of fire development on open land by the use percolation process and function of the neural network

Journal: Pozharovzryvobezopastnost/Fire and Explosion Safety (Vol.26, No. 2)

Publication Date:

Authors : ;

Page : 44-53

Keywords : ;

Source : Downloadexternal Find it from : Google Scholarexternal

Abstract

Introduction. Prevention of fires in rural settlements bordering forests, is becoming more and more urgent task. There is a problem of their prevention, containment and suppression. Fires in open territories are large in scale, the need to consider the weather conditions, design features of buildings and structures, the distance of fire stations, availability and location of water sources, etc. Purpose. Development of models describing the development of the fires to assess the fire danger of objects located in open areas. Tasks: 1. The development of the percolation model describing fire development. 2. The creation of a model fire hazard analysis facilities located in open areas, with the use of neural networks, based on expert data. Methods. Stochastic analysis, mathematical modeling, percolation theory, neural networks. Theory. To obtain a model of the combustion process, was used the theory of percolation, which is used to calculate the dimension and density of a space-filling (specific) rural settlement. In the percolation model, the territory on which the rural settlement is represented as a system of combustible (flammable at external heat exposure) stations (nodes) distributed in space. Plan rural area modeled in the form of a lattice. The nodes on the grid is a fire load that corresponds to the test object. Nodes are interconnected by links. These relationships can be of two classes. Some of them connect the stations between which there is a probability that the fire will be transferred by the transfer of heat. When the other class connect a couple of sections, between which the fire will spread by means of sparks. Calculations. The procedure of calculations was the following: § construction facilities were located on a square lattice, which corresponds to rural settlements. Construction of objects, flammability class was divided into wood (IV and V degree of fire resistance), brick buildings and structures (I and II degree of fire resistance) and brick construction with wood trim (III degree of fire resistance); § defining the dimension of the space of Hausdorff - Besicovitch, by generating random points Monte; § if the ratio of the fire load (the total space) total area of construction facilities in the territory of that space is greater than or equal to the percolation limit (the ratio of the number of points to the total number of hit points to construction sites), determine the area S of flame propagation using a computer simulation; § if not, calculated the maximum cluster. Results. As a result, was modeled with the real fires that occurred in rural settlements of the Leningrad region. Discussion. For the determination of fire hazard properties are located in open areas, two questions arose. The first question was associated with external environmental factors influencing the development and spread of fires. The second was associated with the accuracy of data processing. The solution to these questions obtained by the use of neural networks. As the input characteristics there were taken statistics of fires in rural areas that occurred in Leningrad region: wind speed, temperature, distance from the inverter to the fire, journey time, time of following of the first barrels, while localized, the distance to the water source. Conclusions. Thus, the study showed that the prediction of the development of fires in rural settlements on the basis of percolation process with the use of neural networks, will allow to assess the fire danger of objects located in open areas. These guidelines can be used for planning of fire- prevention actions and tactics of fire suppression.

Last modified: 2018-10-17 20:45:49