High Resolution Interferometric Imaging of Liquid-Solid Interfaces with HOTNNET

Abstract

Background A variety of imaging methods are available to obtain kinematic data at an interface, with a widely varying range of spatial and temporal resolution. These methods require a trade-off between imaging rate and resolution. Objective A deep learning framework trained on synchronous profilometry data acquired using two imaging modalities at two different spatial resolutions to enhance spatial resolution while maintaining temporal resolution is desired. Methods Fizeau interferometry (FIF) and frustrated total internal reflection (FTIR) are used to overcome the resolution-rate trade-off via a deep learning framework. The FTIR imaging data are recorded at high resolution, while the FIF imaging data are recorded with a lesser resolved, larger field of view. We apply a deep learning framework using a multi-layer convolutional neural network to enhance the FIF image resolution. Results With the deep learning framework, we achieve the high spatial resolution of measurements obtained by FTIR imaging in all three dimensions from the lower resolution FIF data. A high-order overset technique ultimately yields full up-scaled images from the network outputs without losing precision. The accuracy of the super-resolved image is evaluated using test data. Conclusions This hybrid framework, called HOTNNET, is implemented in its entirety on high-speed imaging profilometry data acquired in the study of droplet impacts on a smooth, solid surface, and is used to recover full, high-resolution images at high rates by unwrapping the phase of the interferometry. This framework can be readily adapted to other paired datasets by retraining the network on the novel data.

3D characterization of kinematic fields and poroelastic swellingnear the tip of a propagating crack in a hydrogel

“…After successfully tracking the particles to form trajectories that can be traced back to the reference frame, full-field 3D displacements from the reference frame are directly calculated. The deformation gradient tensor F at each particle position is then estimated by a local least squares routine [58], using the displacement vectors of nearest-neighbor particles. By performing the polar decomposition of F, the near-crack-tip rotation tensor R and stretch tensor U are determined.…”

“…To elevate the reliability in linking particles, an intermediate DIC step is incorporated. Once the 3D displacement data is obtained, deformation gradient tensor is estimated at each particle location by a least-squares approach [58], and further the kinematic data, such as stretch, rotation, and volumetric change. (b) The histogram of the fractional part of particle coordinates in the reference frame, determined in the particle locating process.…”

“…In this work, we use PAAm hydrogel with an oftused composition as a model material for poroelastic solids, and embed passive micro particles into the hydrogel as material tracers [57]. 3D image stacks of a slowly propagating mode I crack are obtained using an optical microscope, and the particle trajectories are derived surrounding the crack with subpixel accuracy [58][59][60], using a hybrid 3D particle tracking algorithm developed based on open-source software TrackPy [61] and Ncorr [62]. Based on the particle trajectories, 3D kinematic fields, including the displacement fields, deformation gradient tensor fields, rotation fields, stretch fields, strain fields, and swelling fields, are characterized near the crack tip [58].…”

“…3D image stacks of a slowly propagating mode I crack are obtained using an optical microscope, and the particle trajectories are derived surrounding the crack with subpixel accuracy [58][59][60], using a hybrid 3D particle tracking algorithm developed based on open-source software TrackPy [61] and Ncorr [62]. Based on the particle trajectories, 3D kinematic fields, including the displacement fields, deformation gradient tensor fields, rotation fields, stretch fields, strain fields, and swelling fields, are characterized near the crack tip [58]. The in-plane displacement components and strain components are compared with LEFM predictions.…”

Recent Advances and Applications of Machine Learning in Experimental Solid Mechanics: A Review

“…For example, Jin et al [53] employed a CNN-based DL framework to extract dynamic cohesive properties and fracture toughness of polyurea directly from image-shearing interferometric fringes. Kaviani and Kolinski [54] developed a CNNbased DL framework to convert fringes from Fizeau interferometry with low resolution into frustrated total internal reflection (FTIR) images with high resolution while studying droplet impact. Another important application for CNN in experimental mechanics is to analyze DIC data.…”