Projects

Projects

Materials Science

Materials in extreme environments (Ongoing)

Collaborators: Army Research Lab (ARL)

Energetic materials such as propellants, explosives, and pyrotechnics are key performance components in a wide range of defense applications, including solid rocket motors and munitions. Sensitivity of energetic materials to loading (e.g., shock) is critically dependent on the heterogeneities of the microstructure. In this project, we develop manifold learning-based surrogate models that will accelerate UQ that will help to to understand, predict and control the shock-to-detonation or shock-to-deflagration transition in RDX explosives.

Amorphous Machine Learning (Ongoing)

Collaborators: Depts. Computational Materials and Data Science (Sandia National Laboratory), Materials Science and Engineering (JHU), Chemical and Biomolecular Engineering,(JHU)

In this project we use machine learning to build a robust bridge between atomic-scale data and the development of meso-scale models suitable for predicting the response of metallic glasses. Specifically, we deploy manifold learning for nonlinear dimension reduction of atomistic data to identify machine learned descriptors of their glassy microstructure and link them to physically motivated descriptors.

Optimal Control of Molecular Systems under Electromagnetic Radiation (Ongoing)

Collaborators: Dept. Materials Science and Engineering, University of California, Riverside

In this project we develop methods for optimizing/controlling the time evolution of a molecular system under the influence of electromagnetic radiation (i.e., a laser pulse or an external light source).

Data-driven UQ in Structural Ceramics (Ongoing)

Collaborators: HEMI (JHU)

In this project we XXX.

Natural Hazards

NHERI: Computational Modeling and Simulation Center (Ongoing)

The goal of the SimCenter is to promote advanced modeling and simulation technologies for natural hazards engineering researchers and practitioners to understand and quantify the effects of earthquakes, hurricanes, tsunamis, and other natural hazards on buildings, lifelines, and communities. The objective of this project is to integrate state-of-the-art uncertainty quantification methods with SimCenter tools for hazard modelers.

Manifold Learning for Rapid Post-Wildfire Debris Flow Hazard Assessment (Ongoing)

Wild fires significantly increase the susceptibility of steep terrain to rainfall-induced debris flows. These fast moving, dense flows are initiated during short, intense rainfall and are often triggered with little-to-no warning, making them especially dangerous for communities situated down slope of recently burned terrain. Uncertainties in soil properties,rainfall,and triggering mechanisms play a crucial role in risk analysis. We develop a mechanics-based and data-driven approach for assessing post-wildfire debris flow hazard both in terms of triggering and areal extent.

Manifold Learning for UQ in Performance-based earthquake engineering (Ongoing)

In this project we utilize machine learning to accelerate UQ in performance-based earthquake engineering.

Aerospace

Manifold learning-based surrogate modeling for UQ and design in single- and multi-disciplinary simulations in aerospace (Ongoing)

Collaborators: Rensellaer Polytechnic Institute, University of Darmstadt (Germany), FEAC Engineering (Greece)

In this project we utilize machine learning to develop surrogate models for accelerating UQ in single- and multi-disciplinary simulations in aerospace.

Other projects

Low-dimensional manifold learning for uncertainty quantification in complex multi-scale stochastic sytems (Ongoing)

This work is primarily focused on the mathematical development of dimension reduction methods for uncertainty quantification. Within a multiscale setting, active learning methods are employed at each length-scale to develop physically-informed surrogate models that can be used for prediction at each length-scale and across length scales. The surrogates at each scale are trained to be predictive while maintaining the physical interpretation of the solution.

OptiSimCVD: a data-driven framework for prediction, sensitivity analysis and uncertainty quantification in Chemical Vapor Deposition (CVD) reactors (Ongoing)

The project OptiSimCVD proposes a data-driven framework for prediction, sensitivity analysis and uncertainty quantification in industrial-scale processes used to produce hard coatings and wear protection. The core of the production process is Chemical Vapor Deposition (CVD) reactors with different set up but with a common goal: uniform coatings with strict limits of variability.

MSEE: A University Research Alliance for Materials Science in Extreme Environments (Ongoing)

Funding Source: Defence Threat Reduction Agency

An alliance of 18 research institutions led by the Johns Hopkins University in collaboration with the Defense Threat Reduction Agency (DTRA) to advance fundamental science to reduce the threat of weapons of mass destruction.

Data-driven Uncertainty Quantification in Computational Human Head Models (Ongoing)

Computational models of the human head are promising tools for estimating the impact-induced response of brain, and thus play an important role in the prediction of traumatic brain injury. Modern biofidelic head model simulations are associated with very high computational cost, and high-dimensional inputs and outputs, which limits the applicability of traditional uncertainty quantification methods on these systems.

Completed projects

An initial investigation of structural reliability from sparse data (2018-2019)

In this project we applied methods for imprecise probability in the context of structural reliability in order to realistically assess the uncertainty in probability of failure estimates. More specifically, we developed a framework that couples Subset simulation (SuS) and First-order reliability method with Bayesian/information theoretic multi-model inference.

Efficient stochastic simulation-based computational modeling for structural design, reliability and life-cycle assessment (2016-2019)

In this project we developed novel adaptive stochastic collocation methods. The proposed method is based on the concept of multi-element stochastic collocation methods and is capable of dealing with very high-dimensional models whose solutions are expressed as a vector, a matrix, or a tensor.

Mastering the Computational Challenges in Numerical Modeling and Optimum Design of Carbon Nanotube Reinforced Materials (2012-2017)

In this project we applied artificial neural networks in the framework of multiscale analysis of composite materials with Carbon nanotubes (CNTs).