We put forth a modular approach for distilling hidden flow physics from discrete and sparse observations. To address functional expressiblity, a key limitation of the black-box machine learning methods, we have exploited the use of symbolic regression as a principle for identifying relations and operators that are related to the underlying processes. This approach combines evolutionary computation with feature engineering to provide a tool for discovering hidden parameterizations embedded in the trajectory of fluid flows in the Eulerian frame of reference. Our approach in this study mainly involves gene expression programming (GEP) and sequential threshold ridge regression (STRidge) algorithms. We demonstrate our results in three different applications: (i) equation discovery, (ii) truncation error analysis, and (iii) hidden physics discovery, for which we include both predicting unknown source terms from a set of sparse observations and discovering subgrid scale closure models. We illustrate that both GEP and STRidge algorithms are able to distill the Smagorinsky model from an array of tailored features in solving the Kraichnan turbulence problem. Our results demonstrate the huge potential of these techniques in complex physics problems, and reveal the importance of feature selection and feature engineering in model discovery approaches.
20 March 2024 in Drug Delivery and Translational Research
Diabetes currently affects approximately 500 million people worldwide and is one of the most common causes of mortality in the United States. To diagnose and monitor diabetes, finger-prick blood glucose testing has long been used as the clinical gold standard. For diabetes treatment, insulin is typically delivered subcutaneously through cannula-based syringes, pens, or pumps in almost all type 1 diabetic (T1D) patients and some type 2 diabetic (T2D) patients. These painful, invasive approaches can cause non-adherence to glucose testing and insulin therapy. To address these problems, researchers have developed miniaturized blood glucose testing devices as well as microfluidic platforms for non-invasive glucose testing through other body fluids. In addition, glycated hemoglobin (HbA1c), insulin levels, and cellular biomechanics-related metrics have also been considered for microfluidic-based diabetes diagnosis. For the treatment of diabetes, insulin has been delivered transdermally through microdevices, mostly through microneedle array-based, minimally invasive injections. Researchers have also developed microfluidic platforms for oral, intraperitoneal, and inhalation-based delivery of insulin. For T2D patients, metformin, glucagon-like peptide 1 (GLP-1), and GLP-1 receptor agonists have also been delivered using microfluidic technologies. Thus far, clinical studies have been widely performed on microfluidic-based diabetes monitoring, especially glucose sensing, yet technologies for the delivery of insulin and other drugs to diabetic patients with microfluidics are still mostly in the preclinical stage. This article provides a concise review of the role of microfluidic devices in the diagnosis and monitoring of diabetes, as well as the delivery of pharmaceuticals to treat diabetes using microfluidic technologies in the recent literature. Graphical abstract
Local flow dynamics play a central role in physiological processes like respiration and nutrient uptake in coral reefs. Despite the importance of corals as hosts to a quarter of all marine life, and the pervasive threats facing corals, characterizing the hydrodynamics between the branches of scleractinian corals has remained a significant challenge. Here, we investigate the effects of colony branch density and surface structure on the local flow field using three-dimensional immersed boundary, large-eddy simulations for four different colony geometries under unidirectional oncoming flow conditions. The first two colonies were from the Pocillopora genus, one with a densely branched geometry, and one with a comparatively loosely branched geometry. The second pair of geometries were derived from a scan of a single Montipora capitata colony, one with the roughness elements called verrucae covering the surface intact, and one with the verrucae removed. For the Pocillopora corals, we found that the mean velocity profiles changed substantially in the center of the dense colony, becoming significantly reduced at middle heights where flow penetration was poor, while the mean velocity profiles in the loosely branched colony remained similar in character from the front to the back of the colony. For the Montipora corals, somewhat counterintuitively, the colony without verrucae produced almost double the maximum Reynolds stress magnitude above the colony compared to the colony without verrucae. This implies that the smooth colony will have enhanced mass transport and higher bed shear stress and friction velocity values relative to the colony with verrucae.
Polydimethylsiloxane (PDMS) is a synthetic elastomer widely used in biomedical and industrial applications. Despite its widespread use, the natural evolution of its mechanical and surface properties over time remains poorly understood. In this study, we fabricated PDMS samples with base-to-curing agent mixing ratios from 5:1 to 30:1 and aged them for up to 8 weeks under six non-harsh conditions at room temperature. Contact angle measurements revealed increasing hydrophobicity with aging, with maximum increases up to 16.5°. Mechanical testing showed up to 130% increases in Young’s modulus and 60% changes in flexibility after 5 weeks. Storage in mineral oil best preserved surface hydrophilicity, while storage in water best maintained mechanical integrity. These results provide a framework for optimizing PDMS storage conditions in microfluidic and biomedical device applications.
Large eddy simulations were performed to characterize the flow and mass transport mechanisms in the interior of two Pocillopora coral colonies with different geometries, one with a relatively loosely branched morphology (P. eydouxi), and the other with a relatively densely branched structure (P. meandrina). Detailed velocity vector and streamline fields were obtained inside both corals for the same unidirectional oncoming flow, and significant differences were found between their flow profiles and mass transport mechanisms. For the densely branched P. meandrina colony, a significant number of vortices were shed from individual branches, which passively stirred the water column and enhanced the mass transport rate inside the colony. In contrast, vortices were mostly absent within the more loosely branched P. eydouxi colony. To further understand the impact of the branch density on internal mass transport processes, the non-dimensional Stanton number for mass transfer, St, was calculated based on the local flow time scale and compared between the colonies. The results showed up to a increase in St when the mean vortex diameter was used to calculate St, compared to calculations based on the mean branch diameter. Turbulent flow statistics, including the fluctuating velocity components, the mean Reynolds stress, and the variance of the velocity components were calculated and compared along the height of the flow domain. The comparison of turbulent flow statistics showed similar Reynolds stress profiles for both corals, but higher velocity variations, in the interior of the densely branched coral, P. meandrina.
Corals exchange nutrients and dissolved gases with the surrounding environment for metabolic purposes. A recent study demonstrated that corals can actively stir quiescent water columns and produce vortical flows that enhance mass transfer rates by up to 400%. Here, three-dimensional immersed-boundary simulations of the flow through a Pocillopora meandrina colony demonstrate that the passive geometric features of the branching colony produce highly vortical internal flows that enhance mass transfer at the interior of the colony, compensating almost exactly for flows speed reductions there of up to 64% so that the advection time scale remains roughly constant throughout the colony.