gt mined - experts in materials science data

The GT MINED research team is focused on finding new pathways to dramatically reducing the cost and time required for the design, manufacture, and deployment of new/improved materials in advanced technologies. MINED uses revolutionary cross-disciplinary research direction that merges materials science and engineering, manufacturing sciences, computational sciences and engineering, applied mathematics, digital signal processing, systems theory, data and information sciences, machine learning, and advanced statistics.

RESEARCH team

CROSS CUTTING interdisciplinary workflows

MINED designs experiments that produce the large statistical datasets which are required by multiscale models. The datasets are derived from recently designed multi-modal materials characterization methods to probe novel sample and testing geometries that contain gradients in both structure and processing.

data GENERATION

COLLABORATION

data PROCESSING

multimodal spatiotemporal infoMINED is building a suite of open science tools to segment, quantify, & visualize Big Data. There is nodataset too complicated.

Al-Cu SolidificationVoorhees (16TB)

hierarchical fiber compositesTailored analytics, visualization, and algorithms designed weave & fiber scale evolving 3-D datasets.

nonlinear materials analyticsComplicated data requires complicated toolsthat account for long-range order and nonEuclidean measures.

Fuel Cell MembranesKumbur (Drexel)

Ablator PanelsMansour (NASA)

3-D OIMGumbsch (KIT)

Polymer MDJacobs (GT)

Brazed WiresNRL

high-throughput experimentsWith an ICME mindset, experimental information and datasets are designed to feed directly into complicated multiscale simulation routines.

compressed sensingSophisticated sampling patterns produce high pedigree empirical information at a fraction of the time and cost.

19m 4.5s30s

nonlinear simulationNovel high-throughput experimental designs are validated homegrown numerical simulations on Ductile metals

in-situ structure-response imagingExtract in-situ SPP relationships using mulitple SEM detectors coupled with an in-situ nanoindenter

Sample

Indenter tip

SEM pole piece

EBSD detector

data GENERATION

Double Cone Specimen

Jominy Bar Tests

FEM Crystal Plasticity

Spherical Nanoindentation

LOCALIZATION relationshipsMachine learning techniques provide datadriven means to inverting simulation data toflow top-down & predict extreme behaviors.

data ANALYTICS

STRUCTURE identificationMINED pushes the limits of statistical structure measures bybuilding algorithms, workflows, and ideas to take the fullest advantage of large, expensive materials data.

α-β Titanium

Aluminum MD

4-D OrganicSolar Cells

Hazelnut CT

materials INFORMATICSMINED combines extensive knowledge of materials & manufacturingSPP relationships, advanced statistics, and modern data science to rapidly provide high-value, objective knowledge of material systems.

Spatial Statistics

HOMOGENIZATION SPPImproved models to extract Bottom-up knowledge from large materials datasets.

MINED has depth and breadth in large multimodal spatiotemporal materials dataset analysis. MINED develops advanced statistical algorithms, feature identification, graph methods, and visualization techniques to explore enormous volumetric images.

data PROCESSING

MINED combines extensive materials knowledge with machine learning algorithms that can effectively explore the growing Materials Big Data ecosystem. They provide insight into structure comparison along with objective top-down and bottom-up scale bridging techniques.

data ANALYTICS

Real-time collaboration toolsMINED has built a global collaboration network across metals, ceramics, and composite materials domains. MINED realizes that “half the time, half the cost” starts with moving information at the speed of the modern world wide web, and aims to do so by contributing software and datasets to the open source community in real time. MINED’s “Lab to the Cloud” gives our collaborators immediate access to experimental datasets using cloud storage. Similarly, in silico datasets and their underlying codes are often provided, with consent, as accompaniments to research publications. As such, MINED promotes a focused and transparent research process with limited time wasted in redundant explorations.