Prediction of FAE concentration based on Physics-Informed Neural Network

The dispersion concentration distribution of Fuel-Air Explosives (FAE) directly affects fuse warhead matching, thereby influencing the FAE weapon's effectiveness. Therefore, real-time simulation of the concentration distribution during the fuel dispersion process is a key technology for the design of FAE weapon’s fuse warhead matching. This paper simplifies the FAE fuel dispersion process, constructs a model of the fuel dispersion process, establishes fuel diffusion control equations, designs the training process of Physics Informed Neural Networks(PINN) to fit the fuel diffusion process, and predicts of concentration of fuel clouds at different spatial coordinates at different times. The global relative error of the predicted values for the three physical fields is approximately 15% to 20%, providing a new method for simulating experiments on the dispersion process of FAE.