Research On Fire Source Location In Large Warehouse Environment

With the modernization of the national economy,the number of large space buildings and warehouses in provinces and cities have become larger.The building structure has also become to high level,large scale and comprehensive direction.Due to the large space structure and function of the warehouse building,it will cause a lot of loss once the fire accident happens.So that the state has made the strict requirements and put special emphasis on large space fire prevention of the warehouse.If the fire source point can be find in the early of the fire,the fire control and people trapped in the fire safety evacuation work will greatly reduce the loss caused and damage.This paper is aiming at fire positioning system research and the problems existing in the actual application that based on the large space building warehouse under the sensor detection system.Using flue gas diffusion model,cluster analysis,neural network and deep learning theory as the means for the wind and no wind two conditions of large space fire warehouse location method for the purpose of research.So that the research of this paper has certain significance and application prospect.It’s research results will help to meet the demand of the modern large space warehouse for precise location of fire source points.The specific research contents are as follows:(1)When there is no wind environment,this paper uses the theory of turbulence under static plume model to simulate the indoor fire smoke diffusion equation and according to gas concentration sensor information with the weighted least square method for the fire source positioning.Finally using the K-means clustering method to real-time dynamic selection of the dense regions of positioning results and get the location of the fire source point estimates(2)When there is wind environment in the warehouse,the influence of wind speed and wind direction is taken into consideration and the diffusion equation in wind environment is improved.But the system equation is nonlinear.Aimed at this situation,by using the particle filter method to estimate the fire source position and compared with the least squares method.It can be seen than the results of this paper is accuracy and effectiveness.(3)When the warehouse environment is complex and the diffusion model is unknown.There is no reliable model for positioning calculation,so the data obtained by the sensorneeds to be modeled and then positioned.This paper proposes a unknown propagation model based on deep learning.Using deep auto-encoder for sensor data dimension down and extract the data characteristics.Then using the extreme learning machine algorithm to training out the smoke concentration and the relationship between the fire point location to location.