Abstract Submission: Water pollutants present a critical issue in open channels and sewer pipe networks. Computational Fluid Dynamics (CFD) can simulate the spread of various pollutants over time, allowing predictions about future concentrations and the trajectory of these pollutants. For certain pollutants, such as viruses and toxic chemicals, identifying the pollution source is crucial to implement effective precautions and address the problem at its origin. The main challenge in this field is the vast number of possible initial conditions; finding the set that corresponds to the observed final distribution is computationally intensive and challenging. This research aims to utilize automatic differentiation (AD) tools and artificial intelligence (AI) to devise a more cost-effective and rapid solution. Additionally, we develop an open-source solver adaptable to a range of pollutants and channel types that inversely identify source locations.
