Search

The cyber-agricultural system (CAS) integrates cybersystems with the physical world of agriculture via sensing, modeling, and actuation, and leverages the three pillars of functional cyber-physical systems (CPSs): computation, control, and communication: Advances in computation (i.e., ubiquitous, multimodal sensing, modeling/reasoning enabled by complex computation capabilities, and off-the-shelf deep learning models) have opened up numerous opportunities in CAS. Progress in control/actuation is characterized by advanced agricultural machinery and the rise of agricultural robotics (e.g., dexterous manipulation and harvesting, interactive sensing, precision spraying, mechanical operations, and weed culling). Advanced communication is enabling sensors, actuators, and platforms to coordinate and collaborate using internet of things (IoT) principles/tools. The cyber-agricultural system (CAS) represents an overarching framework of agriculture that leverages recent advances in ubiquitous sensing, artificial intelligence, smart actuators, and scalable cyberinfrastructure (CI) in both breeding and production agriculture. We discuss the recent progress and perspective of the three fundamental components of CAS – sensing, modeling, and actuation – and the emerging concept of agricultural digital twins (DTs). We also discuss how scalable CI is becoming a key enabler of smart agriculture. In this review we shed light on the significance of CAS in revolutionizing crop breeding and production by enhancing efficiency, productivity, sustainability, and resilience to changing climate. Finally, we identify underexplored and promising future directions for CAS research and development. [ABSTRACT FROM AUTHOR]

Abstract Sensors in buildings are used for a wide variety of applications such as monitoring air quality, contaminants, indoor temperature, and relative humidity. These are used for accessing and ensuring indoor air quality, and also for ensuring safety in the event of chemical and biological attacks. It follows that optimal placement of sensors become important to accurately monitor contaminant levels in the indoor environment. However, contaminant transport inside the indoor environment is governed by the indoor flow conditions which are affected by various uncertainties associated with the building systems including occupancy and boundary fluxes. Therefore, it is important to account for all associated uncertainties while designing the sensor layout. The transfer operator based framework provides an effective way to identify optimal placement of sensors. Previous work has been limited to sensor placements under deterministic scenarios. In this work we extend the transfer operator based approach for optimal sensor placement while accounting for building systems uncertainties. The methodology provides a probabilistic metric to gauge coverage under uncertain conditions. We illustrate the capabilities of the framework with examples exhibiting boundary flux uncertainty. Highlights • A methodology accounting for building uncertainties during optimal sensor placement is presented. • This is based on contaminant transport analysis using the fast and accurate Perron-Frobenious operator. • The method is illustrated for two and three dimensional building geometries while accounting for various constraints associated with sensor placements. • This framework is useful for applications involving indoor air quality, chemical and biological warfare and transmission of infectious diseases. [ABSTRACT FROM AUTHOR]

Abstract Attic specific energy modeling tools are used by engineers and researchers to evaluate the impact of different design features, products, and materials systems on an attic's energy, thermal, and hygrothermal performance. These frameworks and their simulation component models have been carefully validated over the last 40–50 years. A fundamental assumption in these modeling tools is that the air within the attic is well mixed. However, various experimental and numerical studies have shown that the flow inside a triangular attic is not well mixed and an asymmetrical heating of enclosure can develop flow instabilities. Under these flow regimes the calculation of interior convective fluxes using empirical correlations can be error prone. To investigate the impact of this assumption on the performance of a benchmark attic problem, we develop a coupling approach between the Fraunhofer Attic Thermal Model (FATM) and computational fluid dynamics (CFD). Instead of directly coupling a CFD solver to FATM, this paper presents an offline surrogate modeling coupling strategy. The generalized method established in this paper allows for future extension of the approach for different attic geometries, as well as whole building simulations. The work flow uses open-source tools for constructing the surrogate and making it easily integrable with any commercial/non-commercial whole building or attic simulation framework. Since the majority of the computations are moved offline, this approach provides substantial computational speedup during online simulations in comparison to earlier coupling approaches. An additional benefit of the approach is the surrogate can also be used for flow visualization inside the attic. [ABSTRACT FROM AUTHOR]

Deep learning (DL), a subset of machine learning approaches, has emerged as a versatile tool to assimilate large amounts of heterogeneous data and provide reliable predictions of complex and uncertain phenomena. These tools are increasingly being used by the plant science community to make sense of the large datasets now regularly collected via high-throughput phenotyping and genotyping. We review recent work where DL principles have been utilized for digital image–based plant stress phenotyping. We provide a comparative assessment of DL tools against other existing techniques, with respect to decision accuracy, data size requirement, and applicability in various scenarios. Finally, we outline several avenues of research leveraging current and future DL tools in plant science. Highlights Review of DL techniques applied to plant stress (biotic and abiotic) phenotyping to drive transformational changes in agricultural sciences. Comparative assessment of DL strategies across a wide range of plant species for plant stress identification, classification, quantification, and prediction (ICQP), specifically focusing on digital image–based phenotyping. Best practices, future avenues, and potential applications of DL techniques in plant sciences with a focus on plant stress phenotyping, including deployment of DL tools, image data fusion at multiple scales to enable accurate and reliable plant stress ICQP, and use of novel strategies to circumvent the need for accurately labeled data for training the DL tools. [ABSTRACT FROM AUTHOR]

Volatile organic compounds, particulate matter, airborne infectious disease, and harmful chemical or biological agents are examples of gaseous and particulate contaminants affecting human health in indoor environments. Fast and accurate methods are needed for detection, predictive transport, and contaminant source identification. Markov matrices have shown promise for these applications. However, current (Lagrangian and flux based) Markov methods are limited to small time steps and steady-flow fields. We extend the application of Markov matrices by developing a methodology based on Eulerian approaches. This allows construction of Markov matrices with time steps corresponding to very large Courant numbers. We generalize this framework for steady and transient flow fields with constant and time varying contaminant sources. We illustrate this methodology using three published flow fields. The Markov methods show excellent agreement with conventional PDE methods and are up to 100 times faster than the PDE methods. These methods show promise for developing real-time evacuation and containment strategies, demand response control and estimation of contaminant fields of potential harmful particulate or gaseous contaminants in the indoor environment. [ABSTRACT FROM AUTHOR]

Most people spend approximately 90% of their lives indoors. Thus, designing effective ventilation systems is essential to mitigating problems with indoor air quality. The measures of mechanical ventilation design performance considered in this study are age of air, air residual life time, air residence time, and ventilation effectiveness. This paper presents two different methods to help quantify these measures. The first method is based on transport equations, where a continuous representation of these quantities are calculated. The second method is based on Markov matrices, where a discrete representation of these quantities are calculated. We show 1) how both the continuous and discrete methods are related and 2) that the age of air and residual life time are adjoints. A new transport equation for residual life time along with methods for these quantities using Markov matrices are established. The two approaches are validated and compared using previously established experimental data. The results show that both approaches provide similar results. Using these techniques allows for the quantities of residual life time and residence time to be integrated into the design processes. This paper provides a simple framework that enables designers to get a comprehensive picture of the ventilation systems they design. [ABSTRACT FROM AUTHOR]

Air quality has been an important issue in public health for many years. Sensing the level and distributions of impurities help in the control of building systems and mitigate long term health risks. Rapid detection of infectious diseases in large public areas like airports and train stations may help limit exposure and aid in reducing the spread of the disease. Complete coverage by sensors to account for any release scenario of chemical or biological warfare agents may provide the opportunity to develop isolation and evacuation plans that mitigate the impact of the attack. All these scenarios involve strategic placement of sensors to promptly detect and rapidly respond. This paper presents a data driven sensor placement algorithm based on a dynamical systems approach. The approach utilizes the finite dimensional Perron-Frobenius (PF) concept. The PF operator (or the Markov matrix) is used to construct an observability gramian that naturally incorporates sensor accuracy, location constraints, and sensing constraints. The algorithm determines the response times, sensor coverage maps, and the number of sensors needed. The utility of the procedure is illustrated using four examples: a literature example of the flow field inside an aircraft cabin and three air flow fields in different geometries. The effect of the constraints on the response times for different sensor placement scenarios is investigated. Knowledge of the response time and coverage of the multiple sensors aides in the design of mechanical systems and response mechanisms. The methodology provides a simple process for place sensors in a building, analyze the sensor coverage maps and response time necessary during extreme events, as well as evaluate indoor air quality. The theory established in this paper also allows for future work in topics related to construction of classical estimator problems for the sensors, real-time contaminant transport, and development of agent dispersion, contaminant isolation/removal, and evacuation strategies. [ABSTRACT FROM AUTHOR]

Advances in automated and high-throughput imaging technologies have resulted in a deluge of high-resolution images and sensor data of plants. However, extracting patterns and features from this large corpus of data requires the use of machine learning (ML) tools to enable data assimilation and feature identification for stress phenotyping. Four stages of the decision cycle in plant stress phenotyping and plant breeding activities where different ML approaches can be deployed are (i) identification, (ii) classification, (iii) quantification, and (iv) prediction (ICQP). We provide here a comprehensive overview and user-friendly taxonomy of ML tools to enable the plant community to correctly and easily apply the appropriate ML tools and best-practice guidelines for various biotic and abiotic stress traits. [ABSTRACT FROM AUTHOR]

Predicting the movement of contaminants in the indoor environment has applications in tracking airborne infectious disease, ventilation of gaseous contaminants, and the isolation of spaces during biological attacks. Markov matrices provide a convenient way to perform contaminant transport analysis. However, no standardized method exists for calculating these matrices. A methodology based on set theory is developed for calculating contaminant transport in real-time utilizing Markov matrices from CFD flow data (or discrete flow field data). The methodology provides a rigorous yet simple strategy for determining the number and size of the Markov states, the time step associated with the Markov matrix, and calculation of individual entries of the Markov matrix. The procedure is benchmarked against scalar transport of validated airflow fields in enclosed and ventilated spaces. The approach can be applied to any general airflow field, and is shown to calculate contaminant transport over 3000 times faster than solving the corresponding scalar transport partial differential equation. This near real-time methodology allows for the development of more robust sensing and control procedures of critical care environments (clean rooms and hospital wards), small enclosed spaces (like airplane cabins) and high traffic public areas (train stations and airports). [ABSTRACT FROM AUTHOR]

Abstract: Organic solar cells (OSC) show great potential as a low-cost energy source. In addition, their mechanical flexibility allows the added advantage of use on a wide variety of surfaces. In recent years, progress in experimental strategies and modeling approaches have enabled enhancing the power conversion efficiencies of OSC. In particular, simulation based analysis has played a significant role in improving our understanding of the charge transport phenomena in the active layer of these devices. The excitonic drift-diffusion (EDD) model has been used widely to simulate the generation and transport properties of bulk heterojunction (BHJ) solar cells. The EDD model – which is derived from the Boltzmann transport is dependent on a number of input parameters such as (1) material properties (mobility and permittivity), (2) operating conditions (illumination and device thickness), and (3) active layer morphology. A comprehensive sensitivity analysis of the short-circuit current, J sc , to the input parameters is performed. This helps in rank ordering the input parameters and operating conditions – by strength and relevance – on their impact on J sc . We particularly focus our investigations on understanding how the active layer morphology affects the sensitivity of J sc . To accomplish this we analyze three classes of morphologies: bilayer, BHJ, and sawtooth. The results show significant differences in sensitivities between BHJ, sawtooth, and bilayer morphologies. Short-circuit current in BHJ structure shows higher sensitivity to material properties than either sawtooth and bilayer structure, suggesting that the necessity for finer control of material properties to counteract the increased disorder in the active layer morphology. The electrode current is found to be most sensitive to illumination intensity for all three morphologies. We report some interesting trends that may help choose the most sensitive parameters to vary for designing OSC''s with better performance. [Copyright &y& Elsevier]

Abstract: Solidification from the melt to near net shape is a commonly used manufacturing technique. The fluid flow patterns in the melt affect the quality of the final product. By controlling the flow behavior, the final solidified material can be suitably affected. Most of the magnetic field approaches to melt flow control rely on the application of a constant magnetic field. A constant magnetic field results in the Lorentz force which is used to damp and control the flow. However, simultaneous application of a magnetic gradient results in the Kelvin force along with the Lorentz force. This can be used for better control of the melt flow resulting in higher crystal quality. In the present work, a computational method for the design of solidification of a conducting material is addressed. The control parameter in the design problem is the time history of the imposed magnetic field. A steady, constant magnetic gradient is also maintained during the process. The design problem is posed as an unconstrained optimization problem. The adjoint method for the inverse design of continuum processes is adopted. Examples of designing the time history of the imposed magnetic field for the directional growth of various materials are presented to demonstrate the developed formulation. [Copyright &y& Elsevier]

Abstract: The structure of fluid flow in a solidifying melt plays a critical role in the quality/properties of the solid. It follows that by controlling the flow behavior, the final solidified material can be suitably affected. Most of the magnetic control approaches used depend on the variation of the Lorentz force for suppression of flow and are limited to conducting materials alone. The application of a magnetic gradient gives rise to an additional force that can be used to affect the melt flow of any material. In this work, a computational method for the design of solidification of a non-conducting material is addressed such that diffusion-dominated growth is achieved by the suppression of convection. The control parameter in the design problem is the time history of the imposed magnetic field. The design problem is posed as an unconstrained optimization problem. The adjoint method for the inverse design of continuum processes is adopted. Examples of designing the time history of the imposed magnetic field for the directional growth of various non-conducting materials are presented to demonstrate the developed formulation. [Copyright &y& Elsevier]

Accurate and rapid monitoring of indoor air quality is critical to ensure occupant safety in the built environment. This is especially important in events where hazardous substances are released, and prompt estimation of contaminant distribution will facilitate quick evacuation and control response. The built environment is usually equipped with a finite set of sensors that measure local concentration of contaminants. The goal is to use this streaming dataset to estimate the contaminant concentration distribution in the complete domain. We accomplish this by integrating two powerful concepts. We utilize an operator theoretic approach – specifically the Perron-Frobenius (P-F) operator – to model the contaminant transport. Previous work has shown that the PF approach is a fast, effective, and accurate paradigm for sensor placement and contaminant transport prediction. The PF approach is integrated with an Ensemble Kalman Filter to rapidly estimate contaminant distribution under unknown release scenarios, given minimal sensor data. The framework is illustrated for two scenarios: a 2D problem involving an office space, and a 3D problem involving a furnished hotel room. Both examples show that the contaminant distribution is accurately predicted within a few sensor measurement cycles. The general applicability of the framework is illustrated by testing the framework for multiple, unknown release locations. This approach provides a unified, extendable framework for rapid contaminant estimation. Estimating contaminant distribution using Perron Frobenious operator in association with EnkF Estimator. Image 1 • A methodology for estimating contaminant distribution using limited sensor data. • A operator-theoretic framework using Perron-Frobenious operator approach for fast, robust and accurate contaminant transport analysis. • Using the Ensemble Kalman Filtering integrated with the linear Perron-Frobenius (PF) operator approach for designing an estimator. • The approach is illustrated for two and three dimensional problems. • We also showcase the versatility of once constructed PF operator for designing the sensor monitoring network as well. • Application and suitability of this work belongs to problems associated with indoor air quality, chemical and biological warfare and transmission of infectious diseases. [ABSTRACT FROM AUTHOR]

Plant stress phenotyping is essential to select stress-resistant varieties and develop better stress-management strategies. Standardization of visual assessments and deployment of imaging techniques have improved the accuracy and reliability of stress assessment in comparison with unaided visual measurement. The growing capabilities of machine learning (ML) methods in conjunction with image-based phenotyping can extract new insights from curated, annotated, and high-dimensional datasets across varied crops and stresses. We propose an overarching strategy for utilizing ML techniques that methodically enables the application of plant stress phenotyping at multiple scales across different types of stresses, program goals, and environments. Plant stress phenotyping is challenging to implement at multiple organizational scales (leaf, canopy, field). There is a need to improve the speed, accuracy, reliability, and scalability of stress phenotyping while allowing flexibility for highly variable program goals. Advances in ML algorithms create opportunities for augmented plant stress phenotyping to address these challenges. [ABSTRACT FROM AUTHOR]

In a recent work, we studied donor-acceptor blend based organic photovoltaics and, by utilizing a combination of morphology simulations and device modeling, demonstrated that block copolymer compatibilizers with appropriately selected energy levels can be used in such systems to give rise to highly efficient devices that, in some cases, can even outperform idealized morphologies. In the present study, we probe whether morphological simulations can be used in conjunction with simple morphological descriptors as a means to screen the performance characteristics of such energy cascade based ternary blend devices as predicted using device-level simulations. Towards this objective, we present results from different parameter combinations to demonstrate that the domain size, percolation ratio, tortuosity of domains, and concentration gradient at the interface between donors and acceptors correlate strongly with the device performance of such ternary blend systems. Subsequently, we present extensive parameter studies where we simultaneously vary the blend composition, the degree of polymerization of the donor homopolymer, and the acceptor composition of the donor- b -acceptor block copolymer to identify blend formulations that give rise to such optimal morphological and device characteristics. Finally, we demonstrate that, while the overall device performance depends on a combination of morphological factors, the morphological descriptors identified in our work may help identify promising blend formulations. [ABSTRACT FROM AUTHOR]

To build materials knowledge, rigorous description of the material structure and associated tools to explore and exploit information encoded in the structure are needed. These enable recognition, categorization and identification of different classes of microstructure and ultimately enable to link structure with properties of materials. Particular interest lies in the protocols capable of mining the essential information in large microstructure datasets and building robust knowledge systems that can be easily accessed, searched, and shared by the broader materials community. In this paper, we develop a protocol based on automated tools to classify microstructure taxonomies in the context of coarsening behavior which is important for long term stability of materials. Our new concepts for enhanced description of the local microstructure state provide flexibility of description. The mathematical description of microstructure that capture crucial attributes of the material, although central to building materials knowledge, is still elusive. The new description captures important higher order spatial information, but at the same time, allows down sampling if less information is needed. We showcase the classification protocol by studying coarsening of binary polymer blends and classifying steady state structures. We study several microstructure descriptions by changing the microstructure local state order and discretization and critically evaluate their efficacy. Our analysis revealed the superior properties of microstructure representation is based on the first order-gradient of the atomic fraction. [ABSTRACT FROM AUTHOR]

Neural network-based approaches for solving partial differential equations (PDEs) have recently received special attention. However, most neural PDE solvers only apply to rectilinear domains and do not systematically address the imposition of boundary conditions over irregular domain boundaries. In this paper, we present a neural framework to solve partial differential equations over domains with irregularly shaped (non-rectilinear) geometric boundaries. Given the shape of the domain as an input (represented as a binary mask), our network is able to predict the solution field, and can generalize to novel (unseen) irregular domains; the key technical ingredient to realizing this model is a physics-informed loss function that directly incorporates the interior-exterior information of the geometry. We also perform a careful error analysis which reveals theoretical insights into several sources of error incurred in the model-building process. Finally, we showcase various applications in 2D and 3D, along with favorable comparisons with ground truth solutions. • Irregular boundary network (IBN) predicts field PDE solution over arbitrary domains. • PDE loss to learn watertight boundary conditions imposed by complex geometries. • Single trained IBN can produce solutions to a PDE across different arbitrary shapes. • Analysis of convergence and generalization error bounds of the PDE-based loss. • Illustrate the approach on Poisson's and Navier–Stokes PDE for different geometries. [ABSTRACT FROM AUTHOR]

Recent advances in generative modeling, namely Diffusion models, have revolutionized generative modeling, enabling high-quality image generation tailored to user needs. This paper proposes a framework for the generative design of structural components. Specifically, we employ a Latent Diffusion model to generate potential designs of a component that can satisfy a set of problem-specific loading conditions. One of the distinct advantages our approach offers over other generative approaches is the editing of existing designs. We train our model using a dataset of geometries obtained from structural topology optimization utilizing the SIMP algorithm. Consequently, our framework generates inherently near-optimal designs. Our work presents quantitative results that support the structural performance of the generated designs and the variability in potential candidate designs. Furthermore, we provide evidence of the scalability of our framework by operating over voxel domains with resolutions varying from 3 2 3 to 12 8 3 . Our framework can be used as a starting point for generating novel near-optimal designs similar to topology-optimized designs. • Latent diffusion model for generating 3D structural component designs. • Framework for generating component designs consistent with topology optimization. • Generated designs have similar (near-optimal) strain energy to SIMP designs. • Large scale 3D voxel dataset for structural topology optimization. [ABSTRACT FROM AUTHOR]

Highlights: [•] Explore and formulate a stochastic approach to analyze the dynamics of natural ventilation systems. [•] Illustrate the emergence of bi-stable behavior under stochastic conditions, which is not seen in the deterministic case. [•] Construct the sensitivity of the systems to stochastic inputs. [Copyright &y& Elsevier]

• The mechanism of surface charge-based sensing in asymmetrically functionalized nanopores is investigated using "Virtual EIS" tool and experiments. • Virtual EIS tool can generate impedance spectrum from electric currents obtained by applying a step voltage input using a high-throughput code making it computationally efficient. • A way of enhancing sensitivity of the device is numerically explored and validated experimentally using thrombin and its aptamer system. [Display omitted] Monitoring electrochemical impedance changes across an asymmetrically functionalized nanochannel array provides an attractive mechanism for chemical and biological sensors. Specific binding of the receptor molecules with their analyte leads to changes in charge distribution on the nanochannel surfaces modifying the ionic transport across them. The magnitude of impedance change due to receptor/ligand binding or sensor sensitivity depends on a large number of parameters and consequently, identification of parameters that result in sensitive and specific sensing performance is extremely tedious and cost-intensive. We rely on a 'virtual EIS' procedure that models the transient ionic current due to a step-change in voltage to determine the frequency-dependent impedance of an asymmetrically functionalized nanochannel. This procedure is used to predict the impedance changes due to the specific binding of thrombin on nanochannel surfaces. Surface charge changes associated with the binding of thrombin protein on the aptamer coated surface result in a decrease of the membrane impedance and computational results suggest that a reduction in the ionic strength of the electrolyte leads to an increase in the magnitude of binding induced impedance reduction. Sensing experiments with thrombin binding aptamer are performed to evaluate the trends from the high-throughput computations. The agreement between model predictions and experimental observations suggests that the present modeling approach may be utilized to computationally evaluate sensor performance for a range of parameters and rapidly identify sensor configurations that enable point-of-care diagnostic devices with improved sensitivities. [ABSTRACT FROM AUTHOR]

• Deployed VMS with weak BCs on massively parallel octree-based adaptive meshes. • Developed octree parallelization, efficient matrix assembly, and rapid in-out tests. • Demonstrated the ability to capture drag crisis without any wall treatment. • Demonstrated scalability of the framework up to O (32 K) processors. • Deployed for industrial scale study of the platooning effect of semi-trucks. We present a variant of the immersed boundary method integrated with octree meshes for highly efficient and accurate Large Eddy Simulations (LES) of flows around complex geometries. We demonstrate the scalability of the proposed method up to O (32 K) processors. This is achieved by (a) rapid in-out tests; (b) adaptive quadrature for an accurate evaluation of forces; (c) tensorized evaluation during matrix assembly. We showcase this method on two non-trivial applications: accurately computing the drag coefficient of a sphere across Reynolds numbers 1 − 10 6 encompassing the drag crisis regime; simulating flow features across a semi-truck for investigating the effect of platooning on efficiency. [ABSTRACT FROM AUTHOR]

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 VS fed −1 to 291 L kg VS fed −1. • A novel Flow Sculpting enabled Anaerobic Digester for treatment of low-solid waste. • Validation of coiled structure and passive (pillar-induced) mixing using CFD modeling. • Improved performance compared to conventional plug flow reactor. • Decreased VFA level to about half; Increased methane yield by 1.6 times. • Increased COD and VS removal from 75% to 88% and from 78% to 91%, respectively. [ABSTRACT FROM AUTHOR]

Structural topology optimization is a compute-intensive process due to several iterations of simulations required to evaluate the performance of the component during optimization. Deep learning (DL) based approaches can address this challenge, but these methods were demonstrated mainly using 2D shapes and, at best, in low-resolution 3D geometries (typically 3 2 3 ). Further, due to non-manufacturable geometric features, the predicted optimal geometries from DL may not be manufacturable, even using additive manufacturing. In this paper, we develop a DL framework using a multigrid convolutional neural network (CNN) to generate high-resolution topology-optimized 3D geometries with additional checks on the manufacturability of the predicted shapes. Our framework predicts the final optimal topology using the initial strain energy (objective function of structural topology optimization) and target volume fraction (material fraction to be preserved after optimization) as input. We train the network using a multigrid approach, which enables topology optimization at 12 8 3 resolution, which was previously computationally challenging. We first train the multigrid CNN at a lower resolution and then transfer the learned network to continue training at higher resolutions. We use a distributed deep learning framework on a GPU supercomputing cluster to further speed up the training time. Distributed DL significantly speeds up the training time by more than 4 × while achieving similar model performance. Finally, we check the optimal geometries for manufacturability using fused deposition modeling (FDM)-specific manufacturability constraints. The large training dataset (> 60,000 high-resolution topology optimization examples) will be released with the paper to enable further research on this topic. • Multigrid approach to train the neural network at fine resolution of 1283. • Demonstrate the manufacturability by 3D printing several sample models predicted our framework. • A data-parallel distributed deep learning framework to accelerate the training process. • A comprehensive high-resolution data set consisting of more than 60k optimal shapes. [ABSTRACT FROM AUTHOR]

Model ecosystems could provide significant insight into the evolution and behavior of real ecosystems. We discuss the advantages and limitations of common approaches like mesocosms. In this context, we highlight recent breakthroughs that allow for the creation of networks of organisms with independently controlled environments and rates of chemical exchange. [ABSTRACT FROM AUTHOR]

Mixtures of conjugated polymers and quantum dot nanocrystals present an interesting solution-processable materials system for active layers in optoelectronic devices, including solar cells. We use scanning transmission electron microscopy to investigate the effects of exchanging the capping ligand of quantum dots on the three-dimensional morphology of the film. We created 3D reconstructions for blends of poly((4,8-bis(octyloxy)benzo(1,2-b:4,5-b’)-dithiophene-2,6-diyl)(2-((dodecyloxy)carbonyl)thieno (3,4-b)-thiophenediyl)) (PTB1) and PbS quantum dots capped with oleic acid (OA), butylamine (BA), OA to 3-mercaptopropionic acid (MPA), and BA to MPA. We use these reconstructed volumes to evaluate differences in exciton dissociation and charge transport as a function of ligand processing. We show that the MPA exchange without an intermediate BA treatment results in severe changes to the film structure and a non-ideal morphology for an effective device. We also show that with a BA exchange, the morphology remains largely unchanged with the additional MPA treatment. This quantitative characterization elucidates previously reported device performance changes caused by ligand exchange and should inform future device fabrication protocols. [ABSTRACT FROM AUTHOR]

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. [ABSTRACT FROM AUTHOR]

The performance of organic solar cells is strongly dependent on the morphology of the bulk heterojunction active layer. There has been intense efforts to identify and quantify morphological traits that correlate with various stages of the photo physics. While it is generally accepted that donor domain size affects exciton dissociation efficiency and connectivity affects charge collection, identifying morphology trait(s) that correlate with fill factor and total efficiency have remained elusive. In this work, we utilize correlation analysis on a large set of two dimensional bulk heterojunction morphologies to identify traits that are correlated with fill factor and efficiency. A large dataset of bulk heterojunction morphologies using a phase-field model of phase separation was first created. A comprehensive suite of morphology descriptors were evaluated for each of these morphologies using a recently developed graph based approach. Following this, a morphology aware excitonic-drift-diffusion based device model was used to compute current-voltage curves, fill factors, efficiencies as well as spatial distributions of exciton generation, dissociation, and charge collection for each of the morphologies. We find that (for a given material system with a specified HOMO-LUMO gap, and assuming perfect contact with electrodes) device efficiency primarily depends on the short circuit current, and has almost no dependence on the fill factor. Interestingly, we find that the fill factor is largely insensitive to many of the investigated descriptors. It is only weakly dependent on the contact area mismatch – the difference between the fraction of anode in direct contact with donor and the fraction of cathode in direct contact with acceptor. The fill factor is maximized when this quantity is nearly balanced. Since morphologies with a higher fraction of the electrodes in contact with the desirable material show higher short circuit current, we conclude that designing morphologies for a high short circuit current will necessarily lead to reasonably high fill factors. [ABSTRACT FROM AUTHOR]

Despite having achieved the long sought-after performance of 10% power conversion efficiency, high performance organic photovoltaics (OPVs) are still mostly constrained to lab scale devices fabricated by spin coating. Efforts to produce printed OPVs lag considerably behind, and the sensitivity to different fabrication methods highlights the need to develop a comprehensive understanding of the processing-morphology relationship in printing methods. Here we present a systematic experimental investigation of a model low bandgap polymer/fullerene system, poly-isoindigo thienothiophene/PC 61 BM, using a lab-scale analogue to roll-to-roll coating as the fabrication tool in order to understand the impact of processing parameters on morphological evolution. We report that domain size and polymer crystallinity can be tuned by a factor of two by controlling the temperature and coating speed. Lower fabrication temperature simultaneously decreased the phase separation domain size and increased the relative degree of crystallinity in those domains, leading to improved photocurrent. We conclude that domain size in isoindigo/PCBM is dictated by spontaneous phase separation rather than crystal nucleation and growth. Furthermore we present a model to describe the temperature dependence of domain size formation in our system, which demonstrates that morphology is not necessarily strictly dependent on the evaporation rate, but rather on the interplay between evaporation and diffusion during the printing process. [ABSTRACT FROM AUTHOR]