STM October Webinar - "AI Surrogate Modeling for Thermal Simulation"

October 16, 2025 10:00 AM - 11:30 AM (EDT)

Description

Mehdi Vahab, Academic Manager for Mechanical and Aerospace Engineering at MathWorks

Mehdi Vahab is the Academic Manager for Mechanical and Aerospace Engineering at MathWorks. His academic background is in physical modeling, specifically for fluid and thermal systems, and applications of computational modeling in different engineering problems. Before joining MathWorks, he developed numerical methods for the computational modeling of multiphase/multimaterial systems and phase-change dynamics throughout his research career. He applied those methods to study thermofluidics phenomena at different scales and design thermal management systems. At MathWorks, he helps researchers, faculty, and students in Mechanical and Aerospace Engineering with their research and teaching challenges by collaborating and consulting to find better and more accessible solutions.

Title of talk: AI Surrogate Modeling for Thermal Simulation

Talk details:

Recent advances in scientific and physics-informed machine learning have introduced a range of neural network architectures capable of approximating solutions to partial differential equations (PDEs), offering new opportunities for accelerating thermal simulations in medicine and engineering. This talk explores several AI-based surrogate modeling techniques, including Physics-Informed Neural Networks (PINNs), Fourier Neural Operators (FNOs), Graph Neural Networks (GNNs), and transformer-based approaches, for solving heat transfer problems governed by the heat equation. Using MATLAB, we demonstrate how these methods can be applied to simulate temperature distributions in 2D and 3D. We will show how to build and train these models in MATLAB, using training data generated from high-fidelity finite element simulations via Partial Differential Equation Toolbox.

* PINNs incorporate the governing PDE directly into the loss function, enabling mesh-free learning from sparse data. In addition to forward modeling, we show how PINNs can be used for inverse problems, such as estimating thermal conductivity from limited temperature measurements, which is necessary when direct measurement is impractical.

*FNOs learn maps between function spaces, offering fast inference on unseen conditions.

*GNNs operate on graph-structured data, making them well-suited for predicting temperature on complex geometries.

Time

The Webinar will run for approximately 90 minutes.

Location