Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “feature reduction”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 235 records · Page 13
All sourcesNASA NTRSDOE OSTI
DOE OSTI2023

Bimolecular Reductive Elimination of Ethane from Pyridine(diimine) Iron Methyl Complexes: Mechanism, Electronic Structure, and Entry into [2+2] Cycloaddition Catalysis

The application of bimolecular reductive elimination to the activation of iron catalysts for alkene–diene cycloaddition is described. Key to this approach was the synthesis, characterization, electronic structure determination, and ultimately solution stability of a family of pyridine(diimine) iron methyl complexes with diverse steric properties and electronic ground states. Both the aryl-substituted, ( Me PDI)FeCH 3 and ( Et PDI)FeCH 3 ( R PDI = 2,6-(2,6-R 2 -C 6 H 3 N=CMe) 2 C 5 H 3 N), and the alkyl-substituted examples, ( Cy APDI)FeCH 3 ( Cy APDI = 2,6-(C 6 H 11 N=CMe) 2 C 5 H 3 N), have molecular structures significantly distorted from planarity and S = 3/2 ground states. The related N-arylated derivative bearing 2,6-di-isopropyl aryl substituents, ( iPr PDI)FeCH 3 , has an idealized planar geometry and exhibits spin crossover behavior from S = 1/2 to S = 3/2 states. At 23 °C under an N 2 atmosphere, both ( Me PDI)FeCH 3 and ( Et PDI)FeCH 3 underwent reductive elimination of ethane to form the iron dinitrogen precatalysts, [( Me PDI)Fe(N 2 )] 2 (μ-N 2 ) and [( Et PDI)Fe(N 2 )] 2 (μ-N 2 ), respectively, while ( iPr PDI)FeCH 3 proved inert to C–C bond formation. By contrast, addition of butadiene to all three iron methyl complexes induced ethane formation and generated the corresponding iron butadiene complexes, ( R PDI)Fe(η 4 -C 4 H 6 ) (R = Me, Et, i Pr), known precatalysts for the [2+2] cycloaddition of olefins and dienes. Kinetic, crossover experiments, and structural studies were combined with magnetic measurements and Mössbauer spectroscopy to elucidate the electronic and steric features of the iron complexes that enable this unusual reductive elimination and precatalyst activation pathway. Furthermore, transmetalation of methyl groups between iron centers was fast at ambient temperature and independent of steric environment or spin state, while the intermediate dimer underwent the sterically controlled rate-determining reaction with either N 2 or butadiene to access a catalytically active iron compound.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2021

Synthesis and Characterization of Magnesium Vanadates as Potential Magnesium-Ion Cathode Materials through an Ab Initio Guided Carbothermal Reduction Approach**

Many technologically relevant materials for advanced energy storage and catalysis feature reduced transition-metal (TM) oxides that are often nontrivial to prepare because of the need to control the reducing nature of the atmosphere in which they are synthesized. In this report, we show that an ab initio predictive synthesis strategy can be used to produce multi-gram-scale products of various MgV x O y -type phases (δ-MgV 2 O 5 , spinel MgV 2 O 4 , and MgVO 3 ) containing V 3+ or V 4+ relevant for Mg-ion battery cathodes. Characterization of these phases using 25 Mg solid-state NMR spectroscopy illustrates the potential of 25 Mg NMR for studying reversible magnesiation and local charge distributions. Rotor-assisted population transfer (RAPT) is used as a much-needed signal-to-noise enhancement technique. The ab initio guided synthesis method is seen as a step forward towards a predictive synthesis strategy for targeting specific complex TM oxides with variable oxidation states of technological importance.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
NASA NTRS1982

Intensity changes in the UV spectrum of Beta Cephei

Fifty eight short wavelength International Ultraviolet Explorer spectra were taken over a 152 day period. Variations in the absorption lines of C IV (1550 A), Si IV (1393 and 1404 A) and N V (1239 and 1243 A) and a Cr III feature (1208 to 1211 A) were observed. A data reduction technique was developed in order to compare the changes in the equivalent widths of these absorption lines. Periods were determined by least squares fit to a sinusoid using several combinations of the data. In one case eighteen spectra for each ion taken at approximately one day intervals were used; in another, all 58 measurements of each were used; and in a third all measurements from all the ions wee combined. Using both variable metric and interactive algorithms it was determined that for all combinations mentioned a period of 6.10 + or - .06 days is the best fit to the vector of equivalent widths versus time. No other period, with or without the 6.10 day period removed from the data, was found. Analysis of the Cr III feature near 1210 A gives the same period as the others, but in antiphase.

Fahey, R. P.↗
NASA NTRS1995

Traction Drives for Zero Stick-Slip Robots, and Reaction Free, Momentum Balanced Systems

Two differential (dual input, single output) drives (a roller-gear and a pure roller), and a momentum balanced (single input, dual output) drive (pure roller ) were designed, fabricated, and tested. The differential drives are each rated at 295 rad/sec (2800 rpm) input speed, 450 N-m (4,000 in-lbf) output torque. The momentum balanced drive is rated at 302 rad/sec (2880 rpm) input speed, and dual output torques of 434N-m (3840 in-lbf). The Dual Input Differential Roller-Gear Drive (DC-700) has a planetary roller-gear system with a reduction ratio (one input driving the output with the second input fixed) of 29.23: 1. The Dual Input Differential Roller Drive (DC-500) has a planetary roller system with a reduction ratio of approximately 24:1. Each of the differential drives features dual roller-gear or roller arrangements consisting of a sun, four first row planets, four second row planets, and a ring. The Momentum Balanced (Grounded Ring) Drive (DC-400) has a planetary roller system with a reduction ratio of 24:1 with both outputs counterrotating at equal speed. Its single roller cluster consists of a sun, five first and five second row planets, a roller cage or spider and a ring. Outputs are taken from both the roller cage and the ring which counterrotate. Test results reported for all three drives include angular and torque ripple (linearity and cogging), viscous and Coulomb friction, and forward and reverse power efficiency. Of the two differential drives, the Differential Roller Drive had better linearity and less cogging than did the Differential Roller-Gear Drive, but it had higher friction and lower efficiency (particularly at low power throughput levels). Use of full preloading rather than a variable preload system in the Differential Roller Drive assessed a heavy penalty in part load efficiency. Maximum measured efficiency (ratio of power out to power in) was 95% for the Differential Roller-Gear Drive and 86% for the Differential Roller Drive. The Momentum Balanced (Grounded Ring) Drive performed as expected kinematically. Reduction r-atios to the two counterrotating outputs (design nominal=24:1) were measured to be 23.98:1 and 24.12:1 at zero load.. At 25ONm (2200 in-lbf) output torque the ratio changed 2% due to roller creep. This drive was the smoothest of all three as determined from linearity and cogging tests, and maximum measured efficiency (ratio of power out to power in) was 95%. The disadvantages of full preloading as comvared to variable preload were apparent in this drive as in the Differential Roller Drive. Efficiencies at part load were low, but improved dramatically with increases in torque. These were consistent with friction measurements which indicated losses primarily from Coulomb friction. The initial preload level setting was low so roller slip was encountered at higher torques during testing.

Anderson, William J.↗
DOE OSTI2026

In Situ EXAFS Study of Mo–P Bond Dynamics in Molybdenum Phosphide during CO 2 Reduction and Hydrogen Evolution

Transition Metal Phosphides (TMPs) are widely studied as catalysts in reactions like hydrogen evolution, oxygen reduction, and hydrodesulfurization. TMPs have demonstrated remarkable efficiency in accelerating reactions by providing active sites for adsorption and facilitating the transformation of reaction intermediates, a feature that has increased the interest of TMPs in the electrochemical carbon dioxide reduction reaction (eCO 2 RR). Despite their growing use, the potential-dependent structural behavior of TMPs under operating electrochemical conditions remains insufficiently understood. This work investigates the dynamics of the Mo−P bond during eCO 2 RR and hydrogen evolution reaction (HER) using in situ extended X-ray absorption spectroscopy (EXAFS) studies at the Mo−K edge. Structural characterization of the synthesized MoP nanoparticles was performed by powder X-ray diffraction (XRD) and transmission electron microscopy (TEM). Results of the overall electrochemical activity of MoP are presented through linear sweep voltammetry (LSV) experiments in the potential range of 0 to −1.5 V versus Ag/AgCl and chronoamperometry (CA). Liquid products were analyzed by 1 H NMR spectroscopy. The synthesized catalyst showed increasing activity with the highest Faradaic efficiency of 11.75% for C 2+ products at −1.3 V vs Ag/AgCl. In situ EXAFS revealed a change in the average Mo−P bond distance during eCO 2 RR, whereas no such change was observed during the hydrogen evolution reaction (HER). This finding suggests that phosphorus, particularly through modulation of the Mo−P bond length, plays an important role in the eCO 2 RR and demonstrates how in situ EXAFS can track local structural perturbations in MoP catalysts during electrochemical operation, providing complementary structural insight that correlates with electrochemical behavior.

Catalysts↗
NASA NTRS2001

Analysis of the Westland Data Set

The "Westland" set of empirical accelerometer helicopter data with seeded and labeled faults is analyzed with the aim of condition monitoring. The autoregressive (AR) coefficients from a simple linear model encapsulate a great deal of information in a relatively few measurements; and it has also been found that augmentation of these by harmonic and other parameters call improve classification significantly. Several techniques have been explored, among these restricted Coulomb energy (RCE) networks, learning vector quantization (LVQ), Gaussian mixture classifiers and decision trees. A problem with these approaches, and in common with many classification paradigms, is that augmentation of the feature dimension can degrade classification ability. Thus, we also introduce the Bayesian data reduction algorithm (BDRA), which imposes a Dirichlet prior oil training data and is thus able to quantify probability of error in all exact manner, such that features may be discarded or coarsened appropriately.

Wen, Fang↗
NASA NTRS2000

Condition Monitoring for Helicopter Data

In this paper the classical "Westland" set of empirical accelerometer helicopter data is analyzed with the aim of condition monitoring for diagnostic purposes. The goal is to determine features for failure events from these data, via a proprietary signal processing toolbox, and to weigh these according to a variety of classification algorithms. As regards signal processing, it appears that the autoregressive (AR) coefficients from a simple linear model encapsulate a great deal of information in a relatively few measurements; it has also been found that augmentation of these by harmonic and other parameters can improve classification significantly. As regards classification, several techniques have been explored, among these restricted Coulomb energy (RCE) networks, learning vector quantization (LVQ), Gaussian mixture classifiers and decision trees. A problem with these approaches, and in common with many classification paradigms, is that augmentation of the feature dimension can degrade classification ability. Thus, we also introduce the Bayesian data reduction algorithm (BDRA), which imposes a Dirichlet prior on training data and is thus able to quantify probability of error in an exact manner, such that features may be discarded or coarsened appropriately.

Wen, Fang↗
NASA NTRS1994

A new algorithm for five-hole probe calibration, data reduction, and uncertainty analysis

A new algorithm for five-hole probe calibration and data reduction using a non-nulling method is developed. The significant features of the algorithm are: (1) two components of the unit vector in the flow direction replace pitch and yaw angles as flow direction variables; and (2) symmetry rules are developed that greatly simplify Taylor's series representations of the calibration data. In data reduction, four pressure coefficients allow total pressure, static pressure, and flow direction to be calculated directly. The new algorithm's simplicity permits an analytical treatment of the propagation of uncertainty in five-hole probe measurement. The objectives of the uncertainty analysis are to quantify uncertainty of five-hole results (e.g., total pressure, static pressure, and flow direction) and determine the dependence of the result uncertainty on the uncertainty of all underlying experimental and calibration measurands. This study outlines a general procedure that other researchers may use to determine five-hole probe result uncertainty and provides guidance to improve measurement technique. The new algorithm is applied to calibrate and reduce data from a rake of five-hole probes. Here, ten individual probes are mounted on a single probe shaft and used simultaneously. Use of this probe is made practical by the simplicity afforded by this algorithm.

Reichert, Bruce A.↗
DOE OSTI2021

Ce IV 70 Oxosulfate Rings, Frameworks, Supramolecular Assembly, and Redox Activity**

Abstract M IV molecular oxo‐clusters (M=Zr, Hf, Ce, Th, U, Np, Pu) are prolific in bottoms‐up material design, catalysis, and elucidating reaction pathways in nature and in synthesis. Here we introduce Ce 70 , a wheel‐shaped oxo‐cluster, [Ce IV 70 (OH) 36 (O) 64 (SO 4 ) 60 (H 2 O) 10 ] 4− . Ce 70 crystallizes into intricate high pore volume frameworks with divalent transition metals and Ce‐monomer linkers. Eight crystal‐structures feature four framework types in which the Ce 70 ‐rings are linked as propellers, in offset‐stacks, in a tartan pattern, and as isolated rings. Small‐angle X‐ray scattering of Ce 70 dissolved in butylamine, with and without added cations (Ce IV , alkaline earths, Mn II ), shows the metals‘ differentiating roles in ring linking, leading to supramolecular assemblies. The large acidic pores and abundant terminal sulfates provide ion‐exchange behavior, demonstrated with U IV and Nd III . Frameworks featuring Ce III/IV ‐monomer linkers demonstrate both oxidation and reduction. This study opens the door to mixed‐metal, highly porous framework catalysts, and new clusters for metal‐organic framework design

Colliard, Ian↗
DOE OSTI2021

Ce IV 70 oxosulfate rings, frameworks, supramolecular assembly and redox activity

M IV molecular oxo-clusters (M=Zr, Hf, Ce, Th, U, Np, Pu) are prolific in bottoms-up material design, catalysis, and elucidating reaction pathways in nature and in synthesis. Here we introduce Ce 70 , a wheel-shaped oxo-cluster, [Ce IV 70 -(OH) 36 (O) 64 (SO 4 ) 60 (H 2 O) 10 ] 4- . Ce 70 crystallizes into intricate high pore volume frameworks with divalent transition metals and Ce-monomer linkers. Eight crystal-structures feature four framework types in which the Ce 70 -rings are linked as propellers, in offset-stacks, in a tartan pattern, and as isolated rings. In this work, small-angle X-ray scattering of Ce 70 dissolved in butylamine, with and without added cations (Ce IV , alkaline earths, Mn II ), shows the metals‘ differentiating roles in ring linking, leading to supramolecular assemblies. The large acidic pores and abundant terminal sulfates provide ion-exchange behavior, demonstrated with U IV and Nd III . Frameworks featuring Ce III/IV -monomer linkers demonstrate both oxidation and reduction. Furthermore, this study opens the door to mixed-metal, highly porous framework catalysts, and new clusters for metal-organic framework design.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2024

Electrocatalytic Effects of Au Nanoclusters in its Electrochemiluminescence Enhanced by Coreactant Tertiary Amines

Abstract Atomically precise metal nanoclusters are increasingly exploited in electrocatalysis such as water electrolysis and CO 2 reduction. The catalytic effects are generally evaluated by conventional electrochemical methods or end‐product analysis. Herein, near infrared electrochemiluminescence (ECL) of the nanoclusters is introduced as a new signaling mechanism to complement classical voltammetric analysis for elucidating essential electrocatalytic parameters. Coreactant ECL from an aqueous soluble Au 22 nanocluster (NCs) enhanced by a common pH buffer materials 4‐(2‐hydroxyethyl)‐1‐piperazineethanesulfonic acid (HEPES) and by a piperazine drug hydroxyzine (HDZ) is studied under oxidative reduction pathways. Foot of the wave analysis on voltametric features and the kinetics profiles in potential‐step experiments are studied in different nanocluster and coreactant concentrations. Benchmark parameters such as rate constants are determined under EC (Electron‐transfer‐Chemical‐reaction) mechanism as feasibility validations. Because the luminescence properties of AuNCs are sensitive to the surface ligands or adsorbents, ECL is proposed as a new signal readout for operando studies providing insights for governing reaction mechanism and kinetics during catalytic reactions. Retrospectively, mechanistic understanding of the complex multi‐step reactions during ECL generation can provide guidance for the evaluation and optimization of ECL property for better quantitative applications.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2025

Projection-based multifidelity linear regression for data-scarce applications

Surrogate modeling for systems with high-dimensional quantities of interest remains challenging, particularly when training data are costly to acquire. This work develops multifidelity methods for multiple-input multiple-output linear regression targeting data-limited applications with high-dimensional outputs. Multifidelity methods integrate many inexpensive low-fidelity model evaluations with limited, costly high-fidelity evaluations. We introduce two projection-based multifidelity linear regression approaches with linear and nonlinear features that leverage principal component basis vectors for dimensionality reduction and combine multifidelity data through: (i) a direct data augmentation using low-fidelity data, and (ii) a data augmentation incorporating explicit linear corrections between low-fidelity and high-fidelity data. The data augmentation approaches combine high-fidelity and low-fidelity data into a unified training set and train the linear regression model through weighted least squares with fidelity-specific weights. We introduce a proximity-based weighting scheme with automatic weight selection strategy through cross-validation. Here, the proposed multifidelity linear regression methods are demonstrated on approximating the surface pressure field of a hypersonic vehicle in flight and the temperature field on an aircraft disc braking system. In an ultra low-data regime of no more than twelve high-fidelity samples, multifidelity linear regression achieves approximately 2% – 12% improvement in median accuracy and a higher R 2 score relative to single-fidelity methods at comparable computational cost.

data augmentation↗
DOE OSTI2024

Machine learning in materials research: Developments over the last decade and challenges for the future

The number of studies that apply machine learning (ML) to materials science has been growing at a rate of approximately 1.67 times per year over the past decade. In this review, I examine this growth in various contexts. First, I present an analysis of the most commonly used tools (software, databases, materials science methods, and ML methods) used within papers that apply ML to materials science. The analysis demonstrates that despite the growth of deep learning techniques, the use of classical machine learning is still dominant as a whole. It also demonstrates how new research can effectively build upon past research, particular in the domain of ML models trained on density functional theory calculation data. Next, I present the progression of best scores as a function of time on the matbench materials science benchmark for formation enthalpy prediction. In particular, a dramatic improvement of 7 times reduction in error is obtained when progressing from feature-based methods that use conventional ML (random forest, support vector regression, etc.) to the use of graph neural network techniques. Finally, I provide views on future challenges and opportunities, focusing on data size and complexity, extrapolation, interpretation, access, and relevance.

36 MATERIALS SCIENCE↗
DOE OSTI2021

A comparative study of Bi, Sb, and BiSb for electrochemical nitrogen reduction leading to a new catalyst design strategy

Recent studies identified Bi as one of the most promising non-noble metal elements that can promote the electrochemical N 2 reduction reaction (ENRR) to produce NH 3 . The electronic features that make Bi a promising ENRR catalyst may also be owned by Sb that belongs to the same group as Bi. Thus, the ENRR properties of Bi, Sb, and a BiSb alloy were investigated comparatively to identify common characteristics that facilitate the ENRR. These catalysts were prepared as uniform coating layers on high surface area carbon felt electrodes, which could serve as both regular electrodes and pseudo-gas diffusion electrodes. The experimental results demonstrated that while Bi and Sb show comparable ENRR performances, the formation of a BiSb alloy distinctively increases the faradic efficiency for NH 3 production. Additionally, the X-ray photoelectron spectroscopy results revealed that Bi in BiSb possesses a partial positive charge while Sb in BiSb possesses a partial negative charge, which can impact the way the catalyst surface interacts with the reactants and reaction intermediates of the ENRR and hydrogen evolution reaction (HER), the major competing reaction with the ENRR. Computational investigations including the Bader charge analysis and Gibbs free energy calculations for the elemental steps of the ENRR and HER provided an explanation of how the formation of a BiSb alloy can change the selectivity for the ENRR. The combined experimental and theoretical results and discussion contained in this study lead to a new strategy for designing efficient metal catalysts for the ENRR. Additionally, this study investigated how the use of gas phase and dissolved N 2 affected the ENRR performances of the Bi, Sb, and BiSb catalysts.

36 MATERIALS SCIENCE↗
DOE OSTI2023

Data-driven electrolyte design for lithium metal anodes

Improving Coulombic efficiency (CE) is key to the adoption of high energy density lithium metal batteries. Liquid electrolyte engineering has emerged as a promising strategy for improving the CE of lithium metal batteries, but its complexity renders the performance prediction and design of electrolytes challenging. Here, we develop machine learning (ML) models that assist and accelerate the design of high-performance electrolytes. Using the elemental composition of electrolytes as the features of our models, we apply linear regression, random forest, and bagging models to identify the critical features for predicting CE. Our models reveal that a reduction in the solvent oxygen content is critical for superior CE. We use the ML models to design electrolyte formulations with fluorine-free solvents that achieve a high CE of 99.70%. This work highlights the promise of data-driven approaches that can accelerate the design of high-performance electrolytes for lithium metal batteries.

25 ENERGY STORAGE↗
DOE OSTI2023

Automated Design-for-Reliability of a Power Electronics Module

The design of a power electronics module is a multiphysics problem and involves electrical, thermal, and reliability targets and constraints. Although the electrical and thermal design consider multiple aspects such as materials, geometry, and energy losses, reliability design is often limited to the selection of the attachment materials and electrical interconnect types. In this paper, we incorporate reliability metrics in the design phase by investigating the impact of package geometry on the thermomechanical behavior of the die-attachment material. To this end, we conducted thermal and power cycling simulations of a commercial six-pack power module with silicon carbide devices to compute junction temperature and strain energy density per cycle, respectively. We performed multiple simulations with different geometric dimensions and established a correlation between input features and output variables using subspace-based dimension reduction. The machine learning-based dimension reduction method serves as a surrogate model, which can be employed to identify the optimal module design from a thermal and reliability standpoint.

design-for-reliability↗
DOE OSTI2025

Hydrodynamic characterization of the redox chemistry of crown-encapsulated uranyl complexes

The redox properties of actinide-containing species strongly influence their reactivity, speciation, and interfacial behavior, but the experimental quantification of the electrochemical characteristics of molecular actinide complexes in nonaqueous media has not received the attention it deserves. Here, results from hydrodynamic methods and electrochemical simulations of U(VI)/U(V) redox are reported, including quantification of heterogeneous electron-transfer kinetics and estimation of chemical reversibility of U(VI)/U(V) interconversion at electrodes in acetonitrile-based electrolyte. The complexes investigated are recently reported U(VI) and U(V) complexes in which the uranyl ion (UO 2 n+ ) is encapsulated in a macrocyclic 18-crown-6-like moiety templated by a Pt(II) center. These complexes feature the most positive value U VI /U V reduction potential yet reported and are thus particularly relevant to study of facile U(V) generation from U(VI) precursors as well as uranium electroanalysis. Rotating disk electrode (RDE) studies have been used to quantify the diffusion coefficients of the U(VI) and U(V) complexes, and standard heterogeneous electron transfer rate constants ( k 0 ) for the redox have been determined using a conventional Koutecký-Levich analysis. Rotating ring-disk electrode (RRDE) studies have been used to directly interrogate the chemical reversibility of U(VI)-U(V) interconversion, confirming that reduction of the U(VI) complex at an Au disk is associated with formation of the U(V) analogue that can be readily re-oxidized at a Pt ring under hydrodynamic (rotating) conditions. Because measurements of the type reported here are generally associated with current flows that are larger than those found in corresponding quiescent (unstirred) conditions, our findings suggest that hydrodynamic methods could be advantageous for design of electroanalytical approaches to detection of actinide species and study of their redox properties.

actinides↗
DOE OSTI2022

Deep Convolutional Neural Networks Exploit High-Spatial- and -Temporal-Resolution Aerial Imagery to Phenotype Key Traits in Miscanthus

Miscanthus is one of the most promising perennial crops for bioenergy production, with high yield potential and a low environmental footprint. The increasing interest in this crop requires accelerated selection and the development of new screening techniques. New analytical methods that are more accurate and less labor-intensive are needed to better characterize the effects of genetics and the environment on key traits under field conditions. We used persistent multispectral and photogrammetric UAV time-series imagery collected 10 times over the season, together with ground-truth data for thousands of Miscanthus genotypes, to determine the flowering time, culm length, and biomass yield traits. We compared the performance of convolutional neural network (CNN) architectures that used image data from single dates (2D-spatial) versus the integration of multiple dates by 3D-spatiotemporal architectures. The ability of UAV-based remote sensing to rapidly and non-destructively assess large-scale genetic variation in flowering time, height, and biomass production was improved through the use of 3D-spatiotemporal CNN architectures versus 2D-spatial CNN architectures. The performance gains of the best 3D-spatiotemporal analyses compared to the best 2D-spatial architectures manifested in up to 23% improvements in R2, 17% reductions in RMSE, and 20% reductions in MAE. The integration of photogrammetric and spectral features with 3D architectures was crucial to the improved assessment of all traits. In conclusion, our findings demonstrate that the integration of high-spatiotemporal-resolution UAV imagery with 3D-CNNs enables more accurate monitoring of the dynamics of key phenological and yield-related crop traits. This is especially valuable in highly productive, perennial grass crops such as Miscanthus, where in-field phenotyping is especially challenging and traditionally limits the rate of crop improvement through breeding.

47 OTHER INSTRUMENTATION↗