Fast Adaptive Mesh Augmented Lagrangian Digital Image Correlation

Abstract

Background: Digital image correlation (DIC) is a widely used experimental method to measure full-field displacements and strains. This technique compares images of speckle patterns before and after deformation to computationally infer the displacement and strain fields. Of particular interest are complex mechanical phenomena where strains are far from uniform. Objective: In such situations, it would be desirable to use an adaptive technique with higher resolution in the regions of rapidly changing strain and lower resolution elsewhere. Methods: This paper builds on the recently proposed augmented Lagrangian digital image correlation (ALDIC) method to incorporate mesh adaptivity. We call the resulting approach adapt-ALDIC. Results: We show that the structure of ALDIC makes it easy to incorporate adaptive resolution. We demonstrate through both synthetic and experimental examples that adapt-ALDIC is robust and saves significant computational time with almost no loss in accuracy. Among two types of adaptive mesh strategies, we find that adaptive quadtree mesh outperforms Kuhn triangulation mesh both in accuracy and computational cost. Indeed, we demonstrate that quadtree adapt-ALDIC provides compatible deformation and noise insensitivity typical of global DIC at the cost of local DIC. Conclusions Adapt-ALDIC with adaptive quadtree mesh can analyze heterogeneous deformations accurately and efficiently. An open-source Matlab code is freely available through GitHub and Caltech DATA.

Files

Additional details