APPROXIMATE MODEL OF SIMULATED SMOKE IN AIRCRAFT CARGO HOLD BASED ON NARX NEURAL NETWORK

To solve the problem of over-dependence on research resources and excessive consumption of computational fluid dynamics CFD simulation tools in the study of the diffusion law of the simulated smoke flow field in the cargo hold of an aircraft, a new NARX neural network model is proposed to the CFD model, and the time factor and the boundary conditions of the flow field are used as the influencing conditions to make predictions on the diffusion law of the smoke flow field. The smoke concentration at a point in the CFD model and the boundary conditions of the flow field were used as the input of the neural network model to train it and obtain the neural network agent model. The model training and testing results show that the model can effectively replace the CFD model for related studies, and the approximation calculation is effective and the simulation time is greatly reduced.