State parameters prediction of fire using generalized regression neural network

Fire Dynamics Simulation (FDS) can be used to simulate the fire process accurately, but it is time-consuming and unable to meet real-time computing needs. Therefore, it is difficult to be used in fire rescue scheduling and guidance of crowd evacuation in case of emergency. To solve the above mentioned problem, a fire surrogate model using generalized regression neural network (GRNN) is proposed to predict the key state parameters of fire spread process. In the proposed surrogate model, datasets produced by the FDS are employed to train and test the generalized regression neural network. Experimental results indicate that the proposed fire surrogate model not only can satisfy the prediction precision, but also can reduce the computational time significantly. Therefore, the proposed fire surrogate model is a promising tool to assist fire rescue and crowd evacuation.