Research On The Prediction Of Critical Velocity Of Highway Tunnel Fire Based On Artificial Neural Network

With the increasing number of highway tunnels in China,fire is the biggest safety hazard of the tunnel,and the probability of its occurrence is increasing year by year.The critical velocity is a key parameter in the longitudinal ventilation and smoke extraction design of tunnel fires,its size directly determines whether it can effectively exhaust smoke from the tunnel,which has attracted widespread attention from scholars at home and abroad.At present,the study on the prediction of the critical velocity of tunnel fires is mostly based on a single influencing factor.However,the study on the prediction of the critical velocity of tunnel fire under various influencing factors is rare,but there are many factors affecting the critical velocity in the process of fire,and it is difficult to express it accurately with the existing mathematical function formula.Therefore,based on the traditional numerical simulation method,this paper introduces BP neural network technology to study the prediction of the critical velocity of tunnel fire under various influencing factors.Firstly,this thesis,focusing on two-lane highway tunnel,established a full-size horseshoe tunnel model according to the design specification of the highway tunnel.The influence of fire source power,fire source width,fire source location,tunnel blocking ratio and slope on the critical velocity is simulated by FDS.Moreover,the variation of the critical velocity with the influence factors was obtained.The reliability of simulated results was verified by comparison with the existing research findings.Meanwhile,the importance of the influence of various influencing factors on the critical velocity is clarified.Secondly,five main factors related to the critical velocity,such as fire source power,fire source width,fire source location,blocking ratio and slope,were selected as input parameters,and numerical simulation experiment data were used as training samples to construct the BP neural network prediction model of the critical velocity of highway tunnel fire.Based on the shortcomings of the standard BP algorithm,several improved BP algorithms were used to train the network to determine the optimal training algorithm of the model,so as to establish a neural network model with good prediction performance for the critical velocity of the highway tunnel fire.Finally,through test samples,the performance of the neural network prediction model of the critical velocity of highway tunnel fire established in this paper is tested.The results show that the maximum relative error between the predicted value and the expected value of this model is 0.0211 in all prediction points,which can well meet the precision requirements of fire engineering.Therefore,the neural network prediction model established in this paper can better predict the critical velocity under various influencing factors,and it can provide a new method for the development of engineering calculation model to quickly predict the critical velocity of highway tunnel fire.