Eventos Anais de eventos
COBEM 2023
27th International Congress of Mechanical Engineering
Optimizing Contaminant Source Identification with MLP Neural Network
Submission Author:
Guido Fraga Mares Guia de Carvalho , RJ , Brazil
Co-Authors:
Guido Fraga Mares Guia de Carvalho, Antônio Silva Neto, Diego Knupp, David Pelta
Presenter: Guido Fraga Mares Guia de Carvalho
doi://10.26678/ABCM.COBEM2023.COB2023-1625
Abstract
Identifying the source of contaminants in the atmosphere is of great importance across diverse fields such as agriculture, industry, ecology, and security. This study proposes a faster solution to the source identification inverse problem by leveraging the use of a Multi-Layer Perceptron (MLP) neural network as an efficient solver of the contaminant dispersion problem. The MLP network was trained using data from a numerical solution to the two-dimensional advection-diffusion equation, considering different contaminant source locations. This approach presented a significantly reduced computational time, when compared to the conventional numerical methods. The results demonstrated promising performance with an average distance of approximately 10^-2 for 40 estimated cases. This method has the potential to improve the efficiency of contaminant source identification systems, making it valuable for various applications, including real-time monitoring systems.
Keywords
Inverse problem, Source Identification, Artificial neural networks, Advection-Diffusion Equation, Optimization
