Search NASA⌕ Search

SEARCH · Search NASA

Results for “predict”

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 1,315 records · Page 73

Multiscale Prediction of Yarn Pullout Failure Mode in Unreinforced Textile Fabrics

Unreinforced woven fabrics have been implemented in a variety of high performance applications, including body armor, deployable structures, and as the reinforcement material in composites. Multiscale modeling techniques have significantly improved the capabilities of simulation-based tools to capture fabric mechanics efficiently and accurately, but often lack in their prediction of failure and require pairing with finite element analysis (FEA) software, limiting their application to the design of ‘fit-for-purpose’ materials. NASA’s Multiscale Analysis Tool (NASMAT) is a standalone multiscale program that has been traditionally used in the analysis of reinforced composites materials. More recently, it has been amended to simulate unreinforced fabric behavior by allowing the geometric state of the tows to change with applied loading due to the lack of a reinforcement material, such as the matrix seen in composites. Previous work has shown the ability of NASMAT to capture nonlinear macroscale behavior by predicting geometric changes in the state of each subcell as a function of the applied loading and allowing each subcell in the analysis to rotate according to these predicted changes, as well as predict nonlinear behavior due to the fiber breakage failure mode. In this work, the capability of predicting the onset and propagation of failure in plain woven fabrics in NASMAT is presented for the yarn pullout failure mode, which occurs when a fabric is loaded at an off-axis angle relative to the warp of weft tow direction. Yarn pullout behavior is initiated by determining the applied load in which the shear resistance of the contact area between yarn families is overcome. When failure is initiated, yarn pullout is determined to have occurred when the applied displacement, calculated from global strain, exceeds the deformed position of a given contact points between yarn families, determined from pin-joint kinematics. Contact points where pullout has occurred contribute to a global damage parameter used to modify the homogenized stiffness of the fabric, resulting in nonlinear behavior observed at the macroscale. The off-axis loading behavior and yarn pullout failure theory have been developed and implemented into NASMAT such that users can simulate off-axis tensile behavior of fabrics in a single, standalone multiscale tool. Simulations are compared to uniaxial tensile tests at various off-axis angles to demonstrate the capability of the tool in its prediction of both the onset of failure at each off-axis angles and the stress-strain behavior as failure progresses.

Materials↗

CFD Predictions of Fluctuating Pressure Environments on NASA’s SLS Ascent Unsteady Aerodynamics Wind Tunnel Test

This paper details work that is being conducted on predicting fluctuating pressure environments on NASA launch vehicles using computational fluid dynamics (CFD). The accurate characterization of these aeroacoustics environments is necessary in the analysis of vehicle structural health and aerodynamic performance and in the vehicle design process. High-fidelity simulations of the unsteady flow over NASA’s SLS Ascent Unsteady Aerodynamics wind tunnel Test (AUAT) using the hybrid Reynolds-averaged Navier-Stokes (RANS)-large eddy simulations (LES) methodologies in Mississippi States’ Loci/CHEM solver are presented. The two available methods, Nelson-Nichols and dynamic hybrid RANS-LES (DHRL), are compared head-to-head on identical grids at two Mach numbers. At the high-subsonic Mach number, locally supersonic flow expanding over a shoulder induces a separation-reattachment system that is predicted poorly by the Nelson-Nichols method due to an under-prediction in unsteady content. The DHRL solution predicts the spatial extent and frequency domain response of this system well but under-predicts the peak noise levels due to a delay in the RANS-to-LES transition. At the low-supersonic Mach number, compression corner dynamics dominate the unsteady flow. While the Nelson Nichols solution captures this phenomenon well, another solution on a refined grid indicates that some model stress depletion is occurring. The DHRL solution again performs well at this Mach number and matches the wind tunnel data fairly closely, but perhaps most impressively, is fairly agnostic to increasing grid resolution. The findings in this paper are expected to be generally applicable and will be used to guide the prediction of aeroacoustics environments of other NASA launch vehicles.

CFD↗

Broadband Noise Prediction of Two Small Hovering Rotors using FUN3D-ANOPP2

A computational study was performed on two small hovering rotors, the ideally twisted rotor and the optimum hovering 2-bladed rotor, using FUN3D-ANOPP2. The effects of different inviscid upwinding techniques and temporal resolutions on aerodynamic performance and acoustic prediction were compared against experimental data obtained in the Small Hover Anechoic Chamber facility at the NASA Langley Research Center. Decreasing the numerical dissipation with 𝜅 tuning or by using a higher-order inviscid flux reconstruction scheme was shown to improve aerodynamic performance predictions when compared to measured data for most cases. The effect of dissipation on the vorticity field was also seen to influence the amount of turbulence surrounding the vortex core of the first blade-vortex interaction (BVI), the coherence of the second and higher BVI vortex cores, and the downstream breakdown of the rotor wake for both rotor geometries. Tonal and broadband noise predictions from simulation cases with lower dissipation were seen to agree better with the experiment for the ideally twisted rotor, whereas temporal resolution differences had negligible impact on the broadband noise results. Similar fundamental BPF directivity patterns were obtained by all simulation cases for the optimum hovering 2-bladed rotor, while the most accurate 2*BPF directivity trend, compared to the experiment, was predicted by the lower-dissipation case at the coarse temporal resolution, albeit with an amplitude underprediction. Broadband noise spectral roll-off behavior was seen to occur at a higher frequency when using a finer temporal resolution for the optimum hovering 2-bladed rotor, unlike for the ideally twisted rotor geometry. The broadband noise prediction for the optimum hovering 2-bladed rotor from the higher-dissipation case at the coarse temporal resolution was also seen to predict a tone centered around approximately 8.5 kHz, which was similar in amplitude to the measured tone around 16.5 kHz.

Christopher S. Thurman↗

Traffic Prediction for Uncommunicative Aircraft in Terminal Airspace: Development Framework and Performance Evaluations

This paper presents an air traffic prediction algorithm that takes observations of an aircraft and classifies aircraft type, estimates the aircraft's intent to and method of joining an airport traffic pattern, and predicts the aircaft's future trajectory. To develop algorithms that enable autonomous aircraft to safely insert into un-towered traffic patterns, several challenges need to be addressed. These challenges range from traffic detection to sensor fusion to own-ship trajectory replanning. Critical to a trajectory replanning algorithm is information regarding the future behavior of all traffic aircraft in the operational environment. The presented traffic prediction algorithm generates this information using regular measurements of traffic aircraft position and velocity to classify the aircraft by speed-class, estimate how the aircraft will approach the runway, and construct a predicted trajectory to the runway including future positions and velocities at specific times. The predictions of the presented algorithm are the necessary inputs for any downstream traffic pattern sequencing and own-ship trajectory planning routines. The presented algorithm is benchmarked using approximately 300 randomized traffic trajectories, spanning four vehicle weight classes and eight traffic entry types. While the algorithm can process multiple traffic vehicles in the terminal area, there is no prediction of traffic-on-traffic interaction. Each traffic vehicle is processed separately.

John D McMinn↗

Quantifying effectiveness of failure prediction and response in HPC systems : methodology and example.

Effective failure prediction and mitigation strategies in high-performance computing systems could provide huge gains in resilience of tightly coupled large-scale scientific codes. These gains would come from prediction-directed process migration and resource servicing, intelligent resource allocation, and checkpointing driven by failure predictors rather than at regular intervals based on nominal mean time to failure. Given probabilistic associations of outlier behavior in hardware-related metrics with eventual failure in hardware, system software, and/or applications, this paper explores approaches for quantifying the effects of prediction and mitigation strategies and demonstrates these using actual production system data. We describe context-relevant methodologies for determining the accuracy and cost-benefit of predictors. While many research studies have quantified the expected impact of growing system size, and the associated shortened mean time to failure (MTTF), on application performance in large-scale high-performance computing (HPC) platforms, there has been little if any work to quantify the possible gains from predicting system resource failures with significant but imperfect accuracy. This possibly stems from HPC system complexity and the fact that, to date, no one has established any good predictors of failure in these systems. Our work in the OVIS project aims to discover these predictors via a variety of data collection techniques and statistical analysis methods that yield probabilistic predictions. The question then is, 'How good or useful are these predictions?' We investigate methods for answering this question in a general setting, and illustrate them using a specific failure predictor discovered on a production system at Sandia.

99 GENERAL AND MISCELLANEOUS↗

Damage predictions of aluminum thin-walled structures subjected to explosive loads.

Predicting failure of thin-walled structures from explosive loading is a very complex task. The problem can be divided into two parts; the detonation of the explosive to produce the loading on the structure, and secondly the structural response. First, the factors that affect the explosive loading include: size, shape, stand-off, confinement, and chemistry of the explosive. The goal of the first part of the analysis is predicting the pressure on the structure based on these factors. The hydrodynamic code CTH is used to conduct these calculations. Secondly, the response of a structure from the explosive loading is predicted using a detailed finite element model within the explicit analysis code Presto. Material response, to failure, must be established in the analysis to model the failure of this class of structures; validation of this behavior is also required to allow these analyses to be predictive for their intended use. The presentation will detail the validation tests used to support this program. Validation tests using explosively loaded aluminum thin flat plates were used to study all the aspects mentioned above. Experimental measurements of the pressures generated by the explosive and the resulting plate deformations provided data for comparison against analytical predictions. These included pressure-time histories and digital image correlation of the full field plate deflections. The issues studied in the structural analysis were mesh sensitivity, strain based failure metrics, and the coupling methodologies between the blast and structural models. These models have been successfully validated using these tests, thereby increasing confidence of the results obtained in the prediction of failure thresholds of complex structures, including aircraft.

36 MATERIALS SCIENCE↗

PTRANSP Tests of TGLF and Predictions for ITER

A new numerical solver for stiff transport predictions has been developed and implemented in the PTRANSP predictive transport code. The TGLF and GLF23 predictive codes have been incorporated in the solver, verified by comparisons with predictions from the XPTOR code, and validated by comparing predicted and measured profiles. Predictions for ITER baseline plasmas are presented.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Parameter and Prediction uncertainty in an optimized terrestrial carbon cycle model: Effects of constraining variables and data record length

Many parameters in terrestrial biogeochemical models are inherently uncertain, leading to uncertainty in predictions of key carbon cycle variables. At observation sites, this uncertainty can be quantified by applying model-data fusion techniques to estimate model parameters using eddy covariance observations and associated biometric data sets as constraints. Uncertainty is reduced as data records become longer and different types of observations are added. We estimate parametric and associated predictive uncertainty at the Morgan Monroe State Forest in Indiana, USA. Parameters in the Local Terrestrial Ecosystem Carbon (LoTEC) are estimated using both synthetic and actual constraints. These model parameters and uncertainties are then used to make predictions of carbon flux for up to 20 years. We find a strong dependence of both parametric and prediction uncertainty on the length of the data record used in the model-data fusion. In this model framework, this dependence is strongly reduced as the data record length increases beyond 5 years. If synthetic initial biomass pool constraints with realistic uncertainties are included in the model-data fusion, prediction uncertainty is reduced by more than 25% when constraining flux records are less than 3 years. If synthetic annual aboveground woody biomass increment constraints are also included, uncertainty is similarly reduced by an additional 25%. When actual observed eddy covariance data are used as constraints, there is still a strong dependence of parameter and prediction uncertainty on data record length, but the results are harder to interpret because of the inability of LoTEC to reproduce observed interannual variations and the confounding effects of model structural error.

Post, Wilfred M↗

Towards Adaptive Educational Assessments: Predicting Student Performance using Temporal Stability and Data Analytics in Learning Management Systems

Data-driven assessments and adaptive feedback are becoming a cornerstone research in educational data analytics and involve developing methods for exploring the unique types of data that come from the educational context. For example, predicting college student performance is crucial for both the students and educational institutions. It can support timely intervention to prevent students from failing a course, increasing efficacy of advising functions, and improving course completion rate. In this paper, we present our efforts in using data analytics that enable educationists to design novel data-driven assessment and feedback mechanisms. In order to achieve this objective, we investigate temporal stability of students grades and perform predictive analytics on academic data collected from 2009 through 2013 in one of the most commonly used learning management systems, called Moodle. First, we have identified the data features useful for assessments and predicting student outcomes such as students scores in homework assignments, quizzes, exams, in addition to their activities in discussion forums and their total Grade Point Average(GPA) at the same term they enrolled in the course. Second, time series models in both frequency and time domains are applied to characterize the progression as well as overall projections of the grades. In particular, the model analyzed the stability as well as fluctuation of grades among students during the collegiate years (from freshman to senior) and disciplines. Third, Logistic Regression and Neural Network predictive models are used to identify students as early as possible who are in danger of failing the course they are currently enrolled in. These models compute the likelihood of any given student failing (or passing) the current course. The time series analysis indicates that assessments and continuous feedback are critical for freshman and sophomores (even with easy courses) than for seniors, and those assessments may be provided using the predictive models. Numerical results are presented to evaluate and compare the performance of the developed models and their predictive accuracy. Our results show that there are strong ties associated with the first few weeks for coursework and they have an impact on the design and distribution of individual modules.

adaptive learning↗

Vehicle trajectory prediction near or at traffic signal

A system and method for determining a predicted trajectory of a human-driven host vehicle as the human-driven host vehicle approaches a traffic signal. The method includes: obtaining a host vehicle-traffic light distance d x and a longitudinal host vehicle speed v x that are each taken when the human-driven host vehicle approaches the traffic signal; obtaining a traffic light signal phase P t and an traffic light signal timing T t ; obtaining a time of day TOD; providing the host vehicle-traffic light distance d x , the longitudinal host vehicle speed v x , the traffic light signal phase P t , the traffic light signal timing T t , and the time of day TOD as input into an artificial intelligence (AI) vehicle trajectory prediction application, wherein the AI vehicle trajectory prediction application implements an AI vehicle trajectory prediction model; and determining the predicted trajectory of the human-driven host vehicle using the AI vehicle trajectory prediction application.

97 MATHEMATICS AND COMPUTING↗

Predicting Correlation Coefficients for Monte Carlo Eigenvalue Simulations

Monte Carlo methods are most often considered as a reference for neutron transport simulations since very limited approximations are made about nuclear data and system geometry. To report uncertainty of any tally evaluated as generation averages, the sample variance is divided by the number of active generations, which is based on the assumption that the neutron generations are independent. Correlation effects between neutrons in multiplying systems, particularly when performing power iteration to evaluate eigenvalues have been observed in previous work. Neglecting the correlation effect results in an underestimate of uncertainty reported by Monte Carlo calculations. Previous work has also proposed methods to predict the underestimation ratio. Yamamoto et al expanded the fission source distribution with diffusion equation modes, performed numerical simulation of the AR (autoregressive) process of the expansion coefficients and used the correlation of the AR process to predict that of the Monte Carlo eigenvalue simulation. Sutton applied the discretized phase space (DPS) approach to predict the underestimation ratio but the method cannot predict the ratio when one neutron generates offspring in different phase space regions or generates a random number of offspring. This paper presents a method to predict the correlation effect with the model of multi-type branching processes (MBP). The method requires simulations for one generation of neutrons without knowing the source distribution and can predict the underestimation ratio for the cases where the traditional DPS approach does not work. The generation-to-generation correlation determines the convergence rate of active generations, the bias of variance estimator for each generation and the underestimation ratio of variance estimator for tallies averaged over active generations. The generation-to-generation correlation is characterized by the Auto-Correlation Coefficients (ACC) between tallies from different generations. (authors)

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Extending Conformal Prediction to Hidden Markov Models with Exact Validity via de Finetti’s Theorem for Markov Chains

Conformal prediction is a widely used method to quantify uncertainty in settings where the data is independent and identically distributed (IID), or more generally, exchangeable. Conformal prediction takes in a pre-trained classifier and a calibration dataset as inputs, and returns a function which maps feature vectors to subsets of classes. The output of the returned function for a new feature vector is guaranteed to contain the true class with a pre-specified confidence. Despite its success and usefulness in IID settings, extending conformal prediction to non-exchangeable (e.g., Markovian) data in a manner that provably preserves all desirable theoretical properties has largely remained an open problem. As a solution, we extend conformal prediction to the setting of a Hidden Markov Model (HMM) with unknown parameters. The key idea behind the proposed method is to partition the non-exchangeable Markovian data from the HMM into exchangeable blocks by exploiting the de Finetti’s Theorem for Markov Chains discovered by Diaconis and Freedman (1980). The permutations of the exchangeable blocks are then viewed as randomizations of the observed Markovian data from the HMM. The proposed method provably retains all desirable theoretical guarantees offered by the classical conformal prediction framework and is general enough to be useful in many sequential prediction problems.

Nettasinghe, Don Buddhika Wijayantha↗

BatteryPro: A Python Toolkit for Battery Data Analysis and Machine Learning Predictions

Analyzing battery test data for research & development can be time-consuming since battery tests often run on the order of months to years, generating large volumes of data. BatteryPro is a comprehensive Python package and software designed to facilitate advanced analysis and performance predictions for battery test data. Developed for battery researchers, it supports data types from widely used battery testing instruments, including MACCOR and Biologic cycling systems. The software provides a variety of tools for extracting and plotting key battery parameters such as time, voltage, capacity, current, and pressure. In addition to its extensive data analysis capabilities, BatteryPro features a dedicated machine learning module that employs a Bayesian Gaussian Mixture Model (GMM) to predict battery performance and degradation. Users can generate synthetic capacity fade data, calculate fade metrics, and leverage predictive models to forecast long-term battery behavior. The software's graphical user interface (GUI) enhances usability, allowing researchers to upload, merge, and analyze multiple data files with full customizability. The GUI also supports machine learning predictions, enabling users to fit models and make predictions based on selected data and parameters. BatteryPro is built using QtDesigner, scikit-learn, matplotlib, and pandas, ensuring a high level of customization, flexibility, and accuracy in battery data analysis. This tool aims to empower researchers with the ability to perform detailed battery analysis and make informed predictions, ultimately advancing the field of battery research.

25 - ENERGY STORAGE↗

A Data-Driven Framework for Predicting the Sorting and Screening Performance of an Integrated Biomass Feedstock Preprocessing System

The characteristics of mechanically sorted and screened lignocellulosic biomass, such as the mass contents of corn stover anatomical fractions (leaves, husks, stalks, cobs, etc.), can be used to calculate the intermediate feedstock quality attributes “yield” and “purity” that indicate the conversion efficiency of biocrude. No prior study has investigated the correlations from the characteristics of raw biomass and preprocessing unit operation parameters to those intermediate feedstock quality attributes. This work presents a data-driven framework for assessing and predicting the intermediate feedstock quality attributes in an integrated biomass feedstock preprocessing system. Our study used corn stover as a typical type of herbaceous biomass because of its abundance in the U.S. It began with data acquisition of moisture content, particle size distribution, and anatomical fractions of the materials after each unit operation in the system. The objective of this preprocessing system is to minimize husks and leaves and maximizing cobs and stalks by mechanically separating the materials into three streams via disc screen and air separator. Prototype neural network models were then developed to evaluate the feasibility of predicting process outcomes based on measurable parameters. It is found that incorporating physical constraints into these prediction models significantly enhances the accuracy of the predicted yield and purity against the ground truth data. The experimental data and model predictions indicate that decreasing throughput increases purity, while higher throughput results in lower purity. Finally, an optimization problem was introduced to search optimal combinations of feed material properties and preprocessing unit operation parameters, as the intermediate feedstock quality attributes – yield and purity, appeared to be competing factors. The study also suggests the continual need to improve the data-driven framework’s predictability by incorporating more accurate physical models to describe the dynamics in the preprocessing units such as the air separator.

09 - BIOMASS FUELS↗

Improved Prediction of the Doppler Effect in TRISO Fuel

The Doppler feedback mechanism is a major contributor to the passive safety of gas-cooled, graphite-moderated High Temperature Reactors that use fuel based on TRISO particles. It follows that the correct prediction of the magnitude and time-dependence of this feedback effect is essential to the conduct of safety analyses for these reactors. Since the effect is directly dependent on the actual temperature reached by the fuel during transients, the underlying phenomena of heat transfer and temperature rise must be correctly predicted. This paper presents an improved model for the TRISO particle and its thermal behavior during transients. The improved approach incorporates an explicit TRISO heat conduction model to better quantify the time dependence of the temperature in the various layers of the TRISO particle, including its fuel central zone. There follows a better treatment of the Doppler Effect within said fuel zone. The new model is based on a 1-D analytic solution for composite media using the Green’s function technique. The modeling improvement takes advantage of some of the physical behavior of TRISO fuel under irradiation and includes a distinctive look at the physics of the neutronic Doppler Effect. The new methodology has been implemented within the coupled R-Z nodal diffusion code CYNOD-THERMIX. The new model has been applied to the analysis of earthquakes (presented in a companion paper). In this paper, the model is applied to the control rod ejection event, as specified in the OECD PBMR-400 benchmark, but with temperature dependent thermal properties. The results obtained for this transient using the enhanced code are a considerable improvement over the predictions of the original code. The incorporation of the enhanced model shows that the Doppler Effect plays a more significant role than predicted by the original unenhanced model based on the THERMIX homogenized fuel region model. The new model shows that the overall energy generation during the rod ejection transient is significantly lower than predicted by the unenhanced model. The fuel temperature reaches a slightly higher maximum, but at no time does it approach the nominal allowable TRISO fuel temperature. The analyses with the enhanced model also show that the reactor period during the cool down is larger than previously predicted with the homogenous fuel region model.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Bayesian learning for rapid prediction of lithium-ion battery-cycling protocols

Advancing lithium-ion battery technology requires the optimization of cycling protocols. A new data-driven methodology is demonstrated for rapid, accurate prediction of the cycle life obtained by new cycling protocols using a single test lasting only 3 cycles, enabling rapid exploration of cycling protocol design spaces with orders of magnitude reduction in testing time. We achieve this by combining lifetime early prediction with a hierarchical Bayesian model (HBM) to rapidly predict performance distributions without the need for extensive repetitive testing. The methodology is applied to a comprehensive dataset of lithium-iron-phosphate/graphite comprising 29 different fast-charging protocols. HBM alone provides high protocol-lifetime prediction performance, with 6.5% of overall test average percent error, after cycling only one battery to failure. Here, by combining HBM with a battery lifetime prediction model, we achieve a test error of 8.8% using a single 3-cycle test. In addition, the generalizability of the HBM approach is demonstrated for lithium-manganese-cobalt-oxide/graphite cells.

25 ENERGY STORAGE↗

A comparison of computational models for predicting yield sooting index

Sooting propensity, a measurement of how much particulate matter is produced when a fuel is burned, is a property of significant interest among researchers who are striving to discover the next generation of cleaner, more efficient fuels and fuel additives. Many compounds are not viable as fuels and/or fuel additives, and as a result, designing cleaner-burning biofuels using only experimental techniques is inefficient. Predictive models have been instrumental in reducing this inherent difficulty, providing researchers with a tool to preemptively screen compounds before production and testing. The present work compares the accuracies and interpretabilities of existing models used to predict a particular measure of sooting propensity, Yield Sooting Index (YSI). These models include artificial neural networks, graph neural networks, and multivariate equations. A novel equation for predicting YSI based on atom path count and bond order is proposed, which can highlight key structural components that contribute to YSI. It was found that artificial neural networks slightly outperform graph neural networks and greatly outperform multivariate equations in blind (test set) prediction accuracy; however, graph neural networks and multivariate equations provide significantly more interpretability as to how compound structure relates to YSI. Predictions of YSI are compared to experimental measurements for previously un-tested compounds with cetane numbers comparable to diesel fuel (50-60) (butyl decanoate, ethyl decanoate, 1,4-bis(ethenoxymethyl)cyclohexane, and 5-heptyloxolan-2-one), and it was found that these compounds produce significantly less soot compared to diesel fuel.

09 BIOMASS FUELS↗

QC-GN 2 oMS 2 : a Graph Neural Net for High Resolution Mass Spectra Prediction

Predicting the mass spectrum of a molecular ion is often accomplished via three generalized approaches: rules-based methods for bond breaking, deep learning, or quantum chemical (QC) modeling. Rules-based approaches are often limited by the conditions for different chemical subspaces and perform poorly under chemical regimes with few defined rules. QC modeling is theoretically robust but requires significant amounts of computational time to produce a spectrum for a given target. Among deep learning techniques, graph neural networks (GNNs) have performed better than previous work with fingerprint-based neural networks in mass spectra prediction. To explore this technique further, we investigate the effects of including quantum chemically derived information as edge features in the GNN to increase predictive accuracy. The models we investigated include categorical bond order, bond force constants derived from extended tight-binding (xTB) quantum chemistry, and acyclic bond dissociation energies. Throughout this work, we evaluated these models against a control GNN with no edge features in the input graphs. Bond dissociation enthalpies yielded the best improvement with a cosine similarity score of 0.462 relative to the baseline model (0.437). In this work we also apply dynamic graph attention which improves performance on benchmark problems and supports the inclusion of edge features. Between implementations, we investigate the nature of the molecular embedding for spectra prediction and discuss the recognition of fragment topographies in distinct chemistries for further development in tandem mass spectrometry prediction.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗