Session: 04-24-05: Materials Processing and Characterization V

Paper Number: 171436

Inverse Reconstruction of Dislocation Dynamics From High-Frequency Surface Velocity Measurements 

Plastic strain localization is a critical phenomenon that precedes failure in structural metallic materials, often manifesting as slip bands, shear bands, or localized hot spots. It results from the complex, collective dynamics of dislocations interacting with microstructural features such as grain boundaries, precipitates, and solute atoms. Understanding the mechanistic origins of plastic strain localization remains a long-standing challenge due to the extreme spatio-temporal scales at which dislocation activities unfold, typically at sub-micron length scales and sub-nanosecond time scales. Traditional experimental methods, such as transmission electron microscopy (TEM), X-ray diffraction (XRD), and electron channeling contrast imaging (ECCI), provide valuable insights into dislocation structures but suffer from trade-offs between temporal resolution, spatial coverage, and subsurface accessibility.

Emerging surface-based diagnostic technologies, including acoustic emission and laser interferometry, offer high temporal resolution and have shown promise for probing dynamic processes during plastic deformation. However, they capture only sparse, one-dimensional signals on material surfaces, which limits their ability to resolve the underlying three-dimensional dislocation network that drives the observed surface responses. Establishing a quantitative and mechanistically grounded link between surface measurements and internal dislocation dynamics remains a major scientific and technological bottleneck.

In this work, we propose a novel computational framework that bridges this gap by reconstructing four-dimensional plastic strain localization from high-frequency surface velocity measurements. Our approach leverages ensemble variational data assimilation (EnVar) combined with three-dimensional discrete dislocation elastodynamic (3D DDeD) simulations to formulate and solve a high-dimensional inverse problem. By treating the unknown dislocation source positions as the control variables, we significantly reduce the dimensionality of the problem, making it tractable while retaining physical accuracy.

We validate our method using synthetic datasets generated from 3D DDeD simulations that replicate the signals from laser interferometry. Our results demonstrate that the proposed framework can accurately reconstruct the locations and dynamics of dislocation sources, as well as the evolving patterns of plastic strain localization, even with partial surface sensor coverage. The reconstruction error remains below 10% in most cases, with some instances achieving errors as low as 1.67%. This approach opens a new pathway for integrating high-resolution surface diagnostics with physics-informed modeling, enabling the inference of dislocation behavior at spatio-temporal scales that are currently beyond the reach of direct experimental observation.

This framework not only enhances our ability to decode the underlying mechanisms of plastic deformation but also lays the groundwork for real-time diagnostics and predictive modeling in structural materials. By extending this approach to multi-slip systems, polycrystalline materials, and larger geometries, future work could unlock new insights into damage initiation and fatigue life prediction in engineering alloys.

Presenting Author: Junjie Yang Johns Hopkins University

Presenting Author Biography: Junjie Yang graduated with his PhD from Johns Hopkins University in 2025.

Authors:

Junjie Yang Johns Hopkins University
Jaafar El-Awady Johns Hopkins University
Tamer Zaki Johns Hopkins University