Abstract:The control of droplet formation and size using microfluidic devices is a critical operation for both laboratory and industrial applications, e.g. in micro-dosage. Surfactants can be added to improve the...
“…Resampled latent variables z ′ can be characterized by a normal distribution with mean μ and variance σ denoted as ( μ , σ 2 ), with ε sampled from a Gaussian distribution asTo sum up, the total loss function of VAE 56 presents as = MSE ( x , x̂ ) + KLD ( q ( z|x )‖ p ( z|x ))VAE can be used not only like AEs for dimensionality reduction, but also to generate “consistent” data that are different from inputs because of their latent state representation. 29 Here, by ‘consistent’, we refer to the generated data whose statistical properties mimic those of the original dataset. However, it is also an unsupervised learning method and cannot handle condition-specific samples 57 ( e.g.…”
Predicting drop coalescence based on process parameters is crucial for experiment design in chemical engineering. However, predictive models can suffer from the lack of training data and more importantly, the...
“…Resampled latent variables z ′ can be characterized by a normal distribution with mean μ and variance σ denoted as ( μ , σ 2 ), with ε sampled from a Gaussian distribution asTo sum up, the total loss function of VAE 56 presents as = MSE ( x , x̂ ) + KLD ( q ( z|x )‖ p ( z|x ))VAE can be used not only like AEs for dimensionality reduction, but also to generate “consistent” data that are different from inputs because of their latent state representation. 29 Here, by ‘consistent’, we refer to the generated data whose statistical properties mimic those of the original dataset. However, it is also an unsupervised learning method and cannot handle condition-specific samples 57 ( e.g.…”
Predicting drop coalescence based on process parameters is crucial for experiment design in chemical engineering. However, predictive models can suffer from the lack of training data and more importantly, the...
“…86 In addition, other parameters, including cross-junction tilt angles, flow rates, and surfactant concentration, can also be correlated to the droplet size, and multiple droplet properties other than droplet sizes, such as generation frequency and flow regime, can be accurately predicted. [87][88][89] 90,91 To further increase the model training efficiency, Siemenn et al designed a Bayesian optimization and computer vision feedback loop to quickly discover the control parameters. 92 A full optimization procedure can be completed using 60 samples within 2.3 h. Moreover, Wang et al demonstrated that experimental/numerical data from previous publications can also be used as training data, thus significantly increasing the training efficiency and prediction accuracy.…”
Section: Prediction Of Droplet Propertiesmentioning
confidence: 99%
“…86 In addition, other parameters, including cross-junction tilt angles, flow rates, and surfactant concentration, can also be correlated to the droplet size, and multiple droplet properties other than droplet sizes, such as generation frequency and flow regime, can be accurately predicted. 87–89 To speed up the droplet size measurement, Zhang et al developed a mask R-CNN model for feature extraction and data acquisition. 90,91 To further increase the model training efficiency, Siemenn et al designed a Bayesian optimization and computer vision feedback loop to quickly discover the control parameters.…”
Droplet microfluidics emerges as a powerful technology to perform high-throughput experiment, while artificial intelligence (AI) serves as a functional tool to analyze a large set of multiplex data. Their convergence...
“…Figure 3.Effect of TX100 concentration on dimensionless drop ( a ) diameter () and ( b ) formation time () using HSI, PIV, CFD and the Bayesian regularised artificial neural network (BRANN) model from Chagot et al. (2022). …”
Section: Numerical Formulation Scaling and Validationmentioning
confidence: 99%
“…Moreover, a prediction of the drop size was obtained by using a data-driven model (Chagot et al. 2022). As discussed by Kalli et al.…”
Section: Numerical Formulation Scaling and Validationmentioning
The effect of surfactants on the flow characteristics during rapid drop formation in a microchannel is investigated using high-speed imaging, micro-particle image velocimetry and numerical simulations; the latter are performed using a three- dimensional multiphase solver that accounts for the transport of soluble surfactants in the bulk and at the interface. Drops are generated in a flow-focusing microchannel, using silicone oil (
$4.6$
mPa s) as the continuous phase and a 52 % w/w glycerol solution as the dispersed phase. A non-ionic surfactant (Triton X-100) is dissolved in the dispersed phase at concentrations below and above the critical micelle concentration. Good agreement is found between experimental and numerical data for the drop size, drop formation time and circulation patterns. The results reveal strong circulation patterns in the forming drop in the absence of surfactants, whose intensity decreases with increasing surfactant concentration. The surfactant concentration profiles in the bulk and at the interface are shown for all stages of drop formation. The surfactant interfacial concentration is large at the front and the back of the forming drop, while the neck region is almost surfactant free. Marangoni stresses develop away from the neck, contributing to changes in the velocity profile inside the drop.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
