Project Info
Title: “Advanced Geophysical Experiments: The Role of Physics-Informed Neural Networks in Solving the Wave Equation
Manika Prasad
mprasad@mines.edu
Project Goals and Description:
This project introduces a novel method to tackle the wave equation using Physics-Informed Neural Networks (PINNs). Departing from traditional techniques that necessitate discretizing equations, this approach directly integrates the wave equation and its boundary/initial conditions into the training of a deep neural network. PINNs leverage the strengths of deep learning, automatic differentiation, and optimization to closely approximate solutions of complex partial differential equations (PDEs). The aim is to develop a PINN model proficient in accurately solving and forecasting wave behavior under various complex scenarios. This method promises to exceed the capabilities of classical numerical solvers, particularly in scenarios involving intricate geometries or challenging boundary conditions. This project will serve as both a practical application of PINNs in geophysical experiments and a foundational guide for employing this technique across different PDE challenges.
More Information:
Grand Challenge: Develop carbon sequestration methods.
Detailed protocols and guides on applying Physics-Informed Neural Networks in solving the wave equation will be provided, alongside resources for deepening understanding of PINN’s role in enhancing geophysical exploration and analysis. See project-specific intormation at: https://edx.netl.doe.gov/sites/smart/
Primary Contacts:
Manika Prasad, mprasad@mines.edu | Athos Nathanail, athanasios.nathanail@mines.edu
Student Preparation
Qualifications
This project seeks individuals with a curiosity and eagerness to delve into the intersection of geophysics and advanced computational methods. A background in programming or experience with machine learning models can be beneficial but is not required.
TIME COMMITMENT (HRS/WK)
15 - 20
SKILLS/TECHNIQUES GAINED
Introduction to data gathering, data analysis, and machine learning; creation and quality control of models and presentation of results; working in a team
MENTORING PLAN
Depending on student preference:
(1) Initially via one-on-one meetings and discussions with either or both contact persons listed here. Once the student is more familiar with the topic, then working within a group. OR
(2) Initial exploratory meetings to discuss student interest and strengths. Subsequently, working closely with either contact persons listed here or working within a group to address smaller chunks of tasks.
Preferred Student Status
Junior
Senior