PHYSICS-INFORMED NEURAL NETWORK (PINN) AND ITS APPLICATION TO POLLUTANT TRANSPORT

• Main Author: Yuki Sutanto, Christopher
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/83041
• Institution: Institut Teknologi Bandung

Water pollution is one of the environmental problems that can have a significant negative impact on river ecology. To control and address this issue, it is necessary to study the movement of pollutants. The mathematical model of this pollutant movement is formulated in the form of partial differential equations. The advection-diffusion equation is the mathematical model used in this research. To obtain solutions to this equation, both analytical methods and the Physics-Informed Neural Network (PINN) method are employed. Simulations of the advection-diffusion equation results using both the analytical method and PINN are conducted, and the mean-squared error (MSE) value is calculated. The simulation results and MSE calculations show that PINN can be an effective method for solving partial differential equations, particularly the advection-diffusion equation. Furthermore, after comparing the performance of PINN against the analytical method, the solution of the two-dimensional advection-diffusion equation with a pollutant source using the Physics-Informed Neural Network (PINN) method has been studied. The simulation results indicate that the pollutant will move in the direction of advection. The concentration of the pollutant will increase and reach saturation at a certain distance from the pollutant source.