Flow Sculpting Research Articles

Flow sculpting enabled anaerobic digester for energy recovery from low-solid content waste

Traditionally, energy recovery from low-solid-content wastes occurs in Continuously Stirred Tank Reactors, whereas Plug Flow Reactors (PFR) are used to treat high-solid-content wastes. In comparison, this study uses a special configuration of anaerobic PFR (AnPFR), consisting of a coiled tubular structure, for energy recovery from a mixture of Food Waste and Wastewater, fed at a loading rate of 3 gVS.L−1.d−1 and a solids content of 2.5%. The AnPFR was upgraded into a Flow Sculpting enabled Anaerobic Digester (FSAD), an innovative plug flow design relying on flow sculpting via a sequence of pillars to provide passive mixing. The purpose of the FSAD design is to optimize operational performance while maintaining minimum mixing energy requirements. Computational fluid dynamics simulations revealed that pillars induce local vorticity in the fluid and contribute to the inertial deformation of the flow to enhance mixing. Coherently, experimental results proved that upgrading the AnPFR to FSAD resulted in a better stability (VFA dropped from 4433 to 2034 mg L−1) and a higher efficiency (removal efficiencies of COD and volatile solids increased from 75% to 77%–88% and 91%, respectively). Equally important, the methane yield, indicative of energy generation potential, increased from 181 L kg VSfed−1 to 291 L kg VSfed−1.

A Case Study of Deep Reinforcement Learning for Engineering Design: Application to Microfluidic Devices for Flow Sculpting

AbstractEfficient exploration of design spaces is highly sought after in engineering applications. A spectrum of tools has been proposed to deal with the computational difficulties associated with such problems. In the context of our case study, these tools can be broadly classified into optimization and supervised learning approaches. Optimization approaches, while successful, are inherently data inefficient, with evolutionary optimization-based methods being a good example. This inefficiency stems from data not being reused from previous design explorations. Alternately, supervised learning-based design paradigms are data efficient. However, the quality of ensuing solutions depends heavily on the quality of data available. Furthermore, it is difficult to incorporate physics models and domain knowledge aspects of design exploration into pure-learning-based methods. In this work, we formulate a reinforcement learning (RL)-based design framework that mitigates disadvantages of both approaches. Our framework simultaneously finds solutions that are more efficient compared with supervised learning approaches while using data more efficiently compared with genetic algorithm (GA)-based optimization approaches. We illustrate our framework on a problem of microfluidic device design for flow sculpting, and our results show that a single generic RL agent is capable of exploring the solution space to achieve multiple design objectives. Additionally, we demonstrate that the RL agent can be used to solve more complex problems using a targeted refinement step. Thus, we address the data efficiency limitation of optimization-based methods and the limited data problem of supervised learning-based methods. The versatility of our framework is illustrated by utilizing it to gain domain insights and to incorporate domain knowledge. We envision such RL frameworks to have an impact on design science.

FlowSculpt: software for efficient design of inertial flow sculpting devices.

Flow sculpting is a powerful method for passive flow control that uses a sequence of bluff-body structures to engineer the structure of inertially flowing microfluidic streams. A variety of cross-sectional flow shapes can be created through this method, offering a new platform for flow manipulation or material fabrication useful in bioengineering, manufacturing, and chemistry applications. However, the inverse problem in flow sculpting - designing a device that produces a target fluid flow shape - remains challenging due to the complex, diverse, and enormous design space. Solutions to the inverse problem have been constrained to single-material fluid streams that are shaped into top-bottom symmetric shapes due to the bluff-body structures available in current libraries (pillars) that span the height of the channel. In this work, we introduce multi-material design and symmetry-breaking flow deformations enabled by half-height pillars, presented within an extremely fast simulation method for flow sculpting yielding a 34-fold reduction in runtime. The framework is deployed freely as a cross-platform application called "FlowSculpt". We detail its implementation and usage, and discuss the addition of enhanced search operations, which enable users to more easily design flow shapes that replicate their input drawings. With FlowSculpt, the microfluidics community can now quickly design flow shaping microfluidic devices on modest hardware, and easily integrate these complex physics into their research toolkit.

UFlow: software for rational engineering of secondary flows in inertial microfluidic devices

Approaches to abstract and modularize models of fluid flow in microfluidic devices can enable predictive and rational engineering of microfluidic circuits with rapid designer feedback. The shape of co-flowing streams in the inertial flow regime has become of particular importance for new developments in high throughput microscale manufacturing, biological, and chemical research. In a process known as flow sculpting, the cross-sectional distribution of fluid elements is deformed due to the combined effects of diffusion and transverse advection, which are brought on by interaction with velocity gradients induced by sequences of pillar structures. However, the difficulty in solving the Navier–Stokes equations for complex flow-deforming geometries makes design in this space unintuitive, time-consuming, and costly. To mitigate these issues, we have efficiently embedded flow deformation operations previously relegated to high-performance computing into a free, user-friendly, and cross-platform framework called “uFlow”, to bring flow sculpting to the broader community. uFlow computes flow deformation including both advection and diffusion effects from a single pillar in 25 ms on modern consumer hardware, enabling real-time manual design and exploration of microfluidic devices, and fast visualization of 3D particles fabricated via stop flow lithography or optical transient liquid molding. Advanced numerical routines give instant access to a practically infinite set of flow transformations. We showcase uFlow’s design models, describe their implementation and usage, and validate the algorithms which allow real-time feedback with confocal imaging and cutting-edge microfluidic particle fabrication.

A deep learning framework for causal shape transformation

Recurrent neural network (RNN) and Long Short-term Memory (LSTM) networks are the common go-to architecture for exploiting sequential information where the output is dependent on a sequence of inputs. However, in most considered problems, the dependencies typically lie in the latent domain which may not be suitable for applications involving the prediction of a step-wise transformation sequence that is dependent on the previous states only in the visible domain with a known terminal state. We propose a hybrid architecture of convolution neural networks (CNN) and stacked autoencoders (SAE) to learn a sequence of causal actions that nonlinearly transform an input visual pattern or distribution into a target visual pattern or distribution with the same support and demonstrated its practicality in a real-world engineering problem involving the physics of fluids. We solved a high-dimensional one-to-many inverse mapping problem concerning microfluidic flow sculpting, where the use of deep learning methods as an inverse map is very seldom explored. This work serves as a fruitful use-case to applied scientists and engineers in how deep learning can be beneficial as a solution for high-dimensional physical problems, and potentially opening doors to impactful advance in fields such as material sciences and medical biology where multistep topological transformations is a key element.

Deep Learning for Flow Sculpting: Insights into Efficient Learning using Scientific Simulation Data

A new technique for shaping microfluid flow, known as flow sculpting, offers an unprecedented level of passive fluid flow control, with potential breakthrough applications in advancing manufacturing, biology, and chemistry research at the microscale. However, efficiently solving the inverse problem of designing a flow sculpting device for a desired fluid flow shape remains a challenge. Current approaches struggle with the many-to-one design space, requiring substantial user interaction and the necessity of building intuition, all of which are time and resource intensive. Deep learning has emerged as an efficient function approximation technique for high-dimensional spaces, and presents a fast solution to the inverse problem, yet the science of its implementation in similarly defined problems remains largely unexplored. We propose that deep learning methods can completely outpace current approaches for scientific inverse problems while delivering comparable designs. To this end, we show how intelligent sampling of the design space inputs can make deep learning methods more competitive in accuracy, while illustrating their generalization capability to out-of-sample predictions.

Automated Design for Microfluid Flow Sculpting: Multiresolution Approaches, Efficient Encoding, and CUDA Implementation

Sculpting inertial fluid flow using sequences of pillars is a powerful method for flow control in microfluidic devices. Since its recent debut, flow sculpting has been used in novel manufacturing approaches such as microfiber and microparticle design, flow cytometry, and biomedical applications. Most flow sculpting applications can be formulated as an inverse problem of finding a pillar sequence that results in a desired fluid transformation. Manual exploration and design of pillar sequences, while useful, have proven infeasible for finding complex flow transformations. In this work, we extend our automated optimization framework based on genetic algorithms (GAs) to rapidly design micropillar sequences that can generate arbitrary user-defined fluid flow transformations. We design the framework with the following properties: (a) a parameter encoding that respects locality to ensure fast convergence and (b) a multiresolution approach that accelerates convergence while maintaining accuracy. The framework also utilizes graphics processing unit (GPU) architecture via NVIDIA's CUDA for function evaluations. We package this framework in a user-friendly and freely available software suite that enables the larger microfluidics community to utilize these developments. We also demonstrate the framework's capability to rapidly design arbitrary fluid flow shapes across multiple microchannel aspect ratios.

Publisher’s Note: “Optimization of micropillar sequences for fluid flow sculpting” [Phys. Fluids 28, 012003 (2016)

Publisher’s Note: “Optimization of micropillar sequences for fluid flow sculpting” [Phys. Fluids 28, 012003 (2016)

Optimization of micropillar sequences for fluid flow sculpting

Inertial fluid flow deformation around pillars in a microchannel is a new method for controlling fluid flow. Sequences of pillars have been shown to produce a rich phase space with a wide variety of flow transformations. Previous work has successfully demonstrated manual design of pillar sequences to achieve desired transformations of the flow cross section, with experimental validation. However, such a method is not ideal for seeking out complex sculpted shapes as the search space quickly becomes too large for efficient manual discovery. We explore fast, automated optimization methods to solve this problem. We formulate the inertial flow physics in microchannels with different micropillar configurations as a set of state transition matrix operations. These state transition matrices are constructed from experimentally validated streamtraces for a fixed channel length per pillar. This facilitates modeling the effect of a sequence of micropillars as nested matrix-matrix products, which have very efficient numerical implementations. With this new forward model, arbitrary micropillar sequences can be rapidly simulated with various inlet configurations, allowing optimization routines quick access to a large search space. We integrate this framework with the genetic algorithm and showcase its applicability by designing micropillar sequences for various useful transformations. We computationally discover micropillar sequences for complex transformations that are substantially shorter than manually designed sequences. We also determine sequences for novel transformations that were difficult to manually design. Finally, we experimentally validate these computational designs by fabricating devices and comparing predictions with the results from confocal microscopy.