Research On Monitoring Method Of Fire In Repair Stage Of Ancient Buildings Based On YOLO-BP Neural Network

Compared with general building,the building materials of ancient buildings are mostly wood,with more combustible materials and large fire load,which makes fire easily occur.In the repair stage,the risk of fire is even greater,because of construction technology,poor management,illegal operation of construction personnel and environmental factors,which may lead to fire.Therefore,in order to improve the fire monitoring level in the repair stage of ancient buildings,this paper combines the existing fire monitoring technology and risk assessment ability,identifies the monitoring indicators,constructs the fire monitoring model in the repair stage of ancient buildings,and studies the fire monitoring methods,which has guiding significance for the effective prevention of fire accidents.(1)By analyzing the fire risk of ancient buildings in the repair stage,combining with the repair site hazard sources and considering the feasibility of the technology,three fire monitoring indexes of ancient buildings in the repair stage are summarized: temperature of fire source,distance between fire source and combustible,wind speed.By selecting monitoring instruments and arranging detection points,the fire monitoring information is collected and processed on the repair site of ancient buildings,and the index data of fire source temperature and wind speed are obtained,which makes the index quantitative process reasonable.(2)Analyze the fire state evolution of ancient buildings in the repair stage,and sort out the combustion situation of indoor combustibles in each stage.Aiming at the fire source and combustibles in the initial stage of fire,a fire monitoring image target detection algorithm based on YOLO was proposed.The image features were identified,the preprocessed monitoring image was input for target detection and positioning,and the distance index data between the fire source and combustibles was obtained through the coordinate transformation algorithm.It solves the complicated operation problem of the traditional target detection algorithm and makes the detection task more convenient and quick than the traditional algorithm.(3)The fire monitoring and analysis in the repair stage of ancient buildings are carried out,and the fire safety grade is defined,and the numerical interval corresponding to the safety grade is defined.The BP neural network fire monitoring model is constructed,and then a large number of monitoring data are used to train and test the model,and the error convergence curve is obtained.The model has good fitting effect and high accuracy.Through the analysis of the output data,the conclusion of the safe distance of fire operation in the repair stage of ancient buildings is obtained.(4)Using FDS software PyroSim to model an ancient building,simulate the rules of fire heat release rate,temperature and visibility change in the temple under different wind speeds,and obtain the conditions most conducive to fire spread.The fire parameters are set up to simulate the fire development situation when the fire source is at a safe distance,and the safe distance of fire operation is verified.It also proves the applicability of fire monitoring method.It provides a theoretical basis for fire control safety guidance in the repair stage of ancient buildings.