I investigate nonlinear continuum mechanics and multiscale material behavior, leveraging machine learning architectures to build fast, high-fidelity surrogate models for solid mechanics simulations.
Scroll To Explore ↓This project explores a coupled cDDPM and DeepONet framework for predicting the response of hyperelastic materials.
In this project, we explore SOTA Neural Operators to predict the strain trajectories of 2D Digital Composites
In this project, we showcase the efficacy of using FNO for predicting full field stress and strain tensors.
Assisted in graduate-level instruction on constitutive laws, viscoelasticity, nonlinear mechanics, and plane stress and strain formulations for structural applications.
Undergraduate-level instruction on static equilibrium, free-body diagrams, centroids, moments of inertia, and stress-strain analysis of axial, torsional, and bending members.
Led weekly TA sessions, recitations, and grading for determinate and indeterminate truss and beam analysis, Castigliano’s energy methods, influence lines for rolling loads, and cable-arch structures.
Building..... blog page.
A hybrid cDDPM-DeepONet surrogate model to predict full-resolution von Mises($\sigma_{vM}$) fields in hyperelastic materials by decoupling spatial morphology from magnitude scaling
On using Neural Operators to predict the evolution of full-field strain tensors across 2D digital composites under quasi-static loading paths.
Using FNO to predict the full field $\varepsilon$ , $\sigma$ tensors in 2D Digital Composites
Department of Civil and Systems Engineering
Johns Hopkins University
303 Latrobe Hall, 3400 N. Charles Street
Baltimore, MD 21218, United States
Email: mrashi12@jhu.edu