Taciroglu Research Group Natural hazards engineering at regional scales

Software and data

We develop our tools in the open. Most of our software is released alongside the papers it supports, so that results can be reproduced and methods reused.

An octree mesh refining toward an earthquake source, with wavefronts radiating outward to a recording station.

Regional building inventories

BRAILS

Building Recognition using AI at Large Scale. An open-source framework for automatically generating regional building inventories from satellite and street-level imagery — the step that usually limits how far a risk assessment can scale.

Developed within the NHERI SimCenter · github.com/NHERI-SimCenter/BRAILS

Earthquake simulation

Hercules

An octree-based finite element earthquake simulator for large-scale wave propagation problems, originally developed by the Quake Group at Carnegie Mellon and extended by our group.

C · github.com/TRG-NHM/hercules

Hercules Toolbox

Python post-processing for Hercules: binary-to-HDF5 conversion, wave propagation animations, and visualization of station time histories and 3D velocity profiles.

Python · github.com/TRG-NHM/Hercules-Toolbox

DRM and PML for Abaqus

Implementation of the Domain Reduction Method and Perfectly Matched Layers for Abaqus, for soil-structure interaction and wave propagation analysis.

C, Python · github.com/TRG-NHM/DRM-and-PML

Hercules documentation

Build, installation, and usage documentation for Hercules.

github.com/TRG-NHM/HerculesDoc

Structural health monitoring

Zero-shot transfer learning for SHM

Code and resources accompanying our work on zero-shot transfer learning for structural health monitoring.

Jupyter Notebook · MIT · doi.org/10.1016/j.ymssp.2023.110404

Machine-to-structure transfer learning

Transferring damage detection knowledge across rotating machines and framed structures.

Jupyter Notebook · MIT · github.com/TRG-SHM/Machine-To-Structure-TL

Wildfire and applied AI

FUELVISION

A multimodal data fusion and multi-model ensemble algorithm for wildfire fuels mapping.

github.com/TRG-AI4Good/FUELVISION

Lahaina generalizability

Computer vision tools for early post-disaster assessment, with an emphasis on generalizing across events.

Jupyter Notebook · github.com/TRG-AI4Good/Lahaina_Generalizability

De-identifier

PII de-identification for CSV data using two named-entity recognizers, spaCy and Flair.

Python · MIT · github.com/TRG-AI4Good/de-identifier

Datasets

IRIS-Wildfire: 2025 Los Angeles Wildfires

Field data on infrastructure performance in the Eaton and Palisades fire zones, collected under the NSF RAPID award “Infrastructure Resilience and Impact Study for Wildfire-affected Infrastructure in Los Angeles.” Recipient of the 2026 NHERI DesignSafe Dataset Award.

Hosted on DesignSafe

Regional building damage survey, 2023 Kahramanmaraş earthquakes

Survey data documenting building damage across the affected region.

ASCE Open, 2(1), 04024009

Centrifuge testing of embedded structures

Circular and rectangular embedded structures under seismic loading.

Earthquake Spectra, 35(3), 1485–1505