Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Bayesian Robustness”

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 397 records · Page 22

Evaluation of Deep Learning Model Architectures for Point-of-Care Ultrasound Diagnostics

Point-of-care ultrasound imaging is a critical tool for patient triage during trauma for diagnosing injuries and prioritizing limited medical evacuation resources. Specifically, an eFAST exam evaluates if there are free fluids in the chest or abdomen but this is only possible if ultrasound scans can be accurately interpreted, a challenge in the pre-hospital setting. In this effort, we evaluated the use of artificial intelligent eFAST image interpretation models. Widely used deep learning model architectures were evaluated as well as Bayesian models optimized for six different diagnostic models: pneumothorax (i) B- or (ii) M-mode, hemothorax (iii) B- or (iv) M-mode, (v) pelvic or bladder abdominal hemorrhage and (vi) right upper quadrant abdominal hemorrhage. Models were trained using images captured in 27 swine. Using a leave-one-subject-out training approach, the MobileNetV2 and DarkNet53 models surpassed 85% accuracy for each M-mode scan site. The different B-mode models performed worse with accuracies between 68% and 74% except for the pelvic hemorrhage model, which only reached 62% accuracy for all model architectures. These results highlight which eFAST scan sites can be easily automated with image interpretation models, while other scan sites, such as the bladder hemorrhage model, will require more robust model development or data augmentation to improve performance. With these additional improvements, the skill threshold for ultrasound-based triage can be reduced, thus expanding its utility in the pre-hospital setting.

47 OTHER INSTRUMENTATION↗

Domain Knowledge Guided Bayesian Optimization For Autonomous Alignment Of Complex Scientific Instruments

Bayesian Optimization (BO) is a powerful tool for optimizing complex non-linear systems. However, its performance degrades in high-dimensional problems with tightly coupled parameters and highly asymmetric objective landscapes, where rewards are sparse. In such needle-in-a-haystack scenarios, even advanced methods like trust-region BO (TurBO) often lead to unsatisfactory results. We propose a domain knowledge guided Bayesian Optimization approach, which leverages physical insight to fundamentally simplify the search problem by transforming coordinates to decouple input features and align the active subspaces with the primary search axes. We demonstrate this approach's efficacy on a challenging 12-dimensional, 6-crystal Split-and-Delay optical system, where conventional approaches, including standard BO, TuRBO and multi-objective BO, consistently led to unsatisfactory results. When combined with an reverse annealing exploration strategy, this approach reliably converges to the global optimum. The coordinate transformation itself is the key to this success, significantly accelerating the search by aligning input co-ordinate axes with the problem's active subspaces. As increasingly complex scientific instruments, from large telescopes to new spectrometers at X-ray Free Electron Lasers are deployed, the demand for robust high-dimensional optimization grows. Our results demonstrate a generalizable paradigm: leveraging physical insight to transform high-dimensional, coupled optimization problems into simpler representations can enable rapid and robust automated tuning for consistent high performance while still retaining current optimization algorithms.

FOS: Computer and information sciences↗

A Scalable Gradient Free Method for Bayesian Experimental Design with Implicit Models

Bayesian experimental design (BED) is to answer the question that how to choose designs that maximize the information gathering. For implicit models, where the likelihood is intractable but sampling is possible, conventional BED methods have difficulties in efficiently estimating the posterior distribution and maximizing the mutual information (MI) between data and parameters. Recent work proposed the use of gradient ascent to maximize a lower bound on MI to deal with these issues. However, the approach requires a sampling path to compute the pathwise gradient of the MI lower bound with respect to the design variables, and such a pathwise gradient is usually inaccessible for implicit models. In this paper, we propose a novel approach that leverages recent advances in stochastic approximate gradient ascent incorporated with a smoothed variational MI estimator for efficient and robust BED. Without the necessity of pathwise gradients, our approach allows the design process to be achieved through a unified procedure with an approximate gradient for implicit models. Several experiments show that our approach outperforms baseline methods, and significantly improves the scalability of BED in high-dimensional problems.

Zhang, Jiaxin↗

EFIT-Prime: Probabilistic and physics-constrained reduced-order neural network model for equilibrium reconstruction in DIII-D

We introduce EFIT-Prime, a novel machine learning surrogate model for EFIT (Equilibrium FIT) that integrates probabilistic and physics-informed methodologies to overcome typical limitations associated with deterministic and ad hoc neural network architectures. EFIT-Prime utilizes a neural architecture search-based deep ensemble for robust uncertainty quantification, providing scalable and efficient neural architectures that comprehensively quantify both data and model uncertainties. Physically informed by the Grad–Shafranov equation, EFIT-Prime applies a constraint on the current density J tor and a smoothness constraint on the first derivative of the poloidal flux, ensuring physically plausible solutions. Furthermore, the spatial location of the diagnostics is explicitly incorporated in the inputs to account for their spatial correlation. Extensive evaluations demonstrate EFIT-Prime's accuracy and robustness across diverse scenarios, most notably showing good generalization on negative-triangularity discharges that were excluded from training. Timing studies indicate an ensemble inference time of 15 ms for predicting a new equilibrium, offering the possibility of plasma control in real-time, if the model is optimized for speed.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

A Markov framework for generalized post-event systems recovery modeling: From single to multihazards

State-dependent models can be used to represent the system recovery process as a series of stochastic transitions from lower to higher functional states. However, the applications of these models have been limited in scope and there is a lack of a generalized recovery modeling framework. A generalized framework would permit a robust forecasting of systems and system-of-systems recovery under multiple hazards, and more broadly, would contribute to community disaster preparedness. This paper develops a generalized post hazard-event recovery modeling framework based on state-dependent Markov-type processes. We then apply the proposed framework to solve a spectrum of problems that range from hind-casting single-system recovery following a single hazard event to forecasting post-event trajectories under multiple hazards and modeling the recovery of a system-of-systems. First, Markov chains are used to hind-cast the observed recovery for a portfolio of buildings affected by the 2014 South Napa, California, earthquake. Next, Markov processes are used to formulate a parametric post hazard-event recovery model, which can be updated using Bayesian statistics when relevant datasets become available. Semi-Markov processes are then used to develop a more general model of single hazard recovery, which accounts for the intensity of the loading and level of damage caused by the event. Semi-Markov processes with non-renewal features are then used to account for multihazard interactions in a post-event recovery model, and applied to a case study that involves a community in Charleston, South Carolina. Lastly, Markov-type processes are combined with Bayesian networks to model the recovery of residential, commercial, educational, and industrial buildings (system-of-systems) following a hazard event. Overall, these applications demonstrate the versatility of the Markov framework towards handling recovery problems with varying levels of complexity.

42 ENGINEERING↗

Decision Analyses Related to Groundwater Remediation - 17051

Traditional decision analyses of groundwater remediation scenarios frequently fail because the probability of adverse, unanticipated events occurring is often high. This is driven by many factors including that (1) the models developed to represent the flow and transport in contaminated aquifers are typically simpler than reality and do not account for all the important governing processes, (2) probability distributions are assigned to critical performance parameters (such as model inputs or characteristics of the engineered remediation systems) even though some of these parameters might be unknown or not be very well constrained. However, there is a novel decision analysis methodology and a computational tool based on Bayesian Information-Gap Decision Theory (BIG-DT) that are designed to mitigate the shortcomings of the models and probabilistic decision analyses by leveraging a non-probabilistic decision theory -- information-gap decision theory. The new decision analysis methodology considers possible models that have not been explicitly enumerated and does not require us to commit to a particular probability distribution for model and remediation-design parameters. Both the set of possible models and the set of possible probability distributions grow as the degree of uncertainty increases. The fundamental question that BIG-DT asks is 'How large can these sets be before a particular decision results in an undesirable outcome?'. The decision that allows these sets to be the largest is considered to be the best option. In this way, BIG-DT enables robust decision-support for groundwater remediation problems. Here, we apply BIG-DT to a representative groundwater remediation scenario where different options for hydraulic containment and pump and treat are being considered. BIG-DT requires many model runs and high-performance computing resources are needed when complex models are used. We demonstrate the BIG-DT analyses on a series of synthetic problems that are designed to be consistent with real-world problems such as Los Alamos National Laboratory (LANL) contamination sites. We also discuss a general framework for how the BIG-DT analyses will be carried out at the chromium groundwater contamination site at LANL. BIG-DT is implemented in Julia (a high-level, high-performance dynamic programming language for technical computing) and is part of the MADS framework (http://mads.lanl.gov and http://madsjulia.github.io/Mads.jl). (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

A Behavioral Robotics Approach to Radiation Mapping Using Adaptive Sampling

Radiation mapping is a desirable task to automate because of the inherent risks involved and its tedious nature. A novel system was designed to address this by combining various existing technologies, utilizing behavior-based robotics and Bayesian optimization. The system uses a quadruped robot equipped with a manipulator and gamma detector to take measurements at locations that are selected based on the uncertainty of a surrogate model used to estimate the true radiation field. The robot uses input from the world with depth cameras to avoid collisions with the robot’s body, and unreachable points for the end effector are addressed by both allowing for a soft collision with the environment to occur, prompting the system to abandon that point, and varying the exploration tendency of the optimization based on consecutive collisions. This approach provides unique traversability and adaptability over other strategies in the literature. Experiments were performed by placing a Cesium-137 source on the ground and varying geometric setups and an optimization parameter demonstrating the adaptability to diverse environments and the increased robustness resulting from the designed behavior. The results additionally demonstrate that dynamically adjusting the optimization algorithm’s exploration tendency based on the arm’s collision history improves the system’s ability to navigate cluttered environments and construct accurate radiation maps without getting stuck in unreachable areas.

Adams, Joel↗

Assimilating irregularly spaced sparsely observed turbulent signals with hierarchical Bayesian reduced stochastic filters

In this paper, we consider a practical filtering approach for assimilating irregularly spaced, sparsely observed turbulent signals through a hierarchical Bayesian reduced stochastic filtering framework. The proposed hierarchical Bayesian approach consists of two steps, blending a data-driven interpolation scheme and the Mean Stochastic Model (MSM) filter. We examine the potential of using the deterministic piecewise linear interpolation scheme and the ordinary kriging scheme in interpolating irregularly spaced raw data to regularly spaced processed data and the importance of dynamical constraint (through MSM) in filtering the processed data on a numerically stiff state estimation problem. In particular, we test this approach on a two-layer quasi-geostrophic model in a two-dimensional domain with a small radius of deformation to mimic ocean turbulence. Our numerical results suggest that the dynamical constraint becomes important when the observation noise variance is large. Second, we find that the filtered estimates with ordinary kriging are superior to those with linear interpolation when observation networks are not too sparse; such robust results are found from numerical simulations with many randomly simulated irregularly spaced observation networks, various observation time intervals, and observation error variances. Third, when the observation network is very sparse, we find that both the kriging and linear interpolations are comparable.

97 MATHEMATICS AND COMPUTING↗

Tidal Deformabilities and Radii of Neutron Stars from the Observation of GW170817

In this study, we use gravitational-wave observations of the binary neutron star merger GW170817 to explore the tidal deformabilities and radii of neutron stars. We perform a Bayesian parameter estimation with the source location and distance informed by electromagnetic observations. We also assume that the two stars have the same equation of state; we demonstrate that, for stars with masses comparable to the component masses of GW170817, this is effectively implemented by assuming that the stars’ dimensionless tidal deformabilities are determined by the binary’s mass ratio q by Λ 1 /Λ 2 = q 6 . We investigate different choices of prior on the component masses of the neutron stars. We find that the tidal deformability and 90% credible interval is ˜Λ = 222$+420\atop{-138}$ for a uniform component mass prior, ˜Λ = 245$+453\atop{-151}$ for a component mass prior informed by radio observations of Galactic double neutron stars, and ˜Λ = 233$+448\atop{-144}$ for a component mass prior informed by radio pulsars. We find a robust measurement of the common areal radius of the neutron stars across all mass priors of 8.9 ≤ $\hat{R}$ ≤ 13.2 km , with a mean value of <$\hat{R}$> = 10.8 km . Finally, our results are the first measurement of tidal deformability with a physical constraint on the star’s equation of state and place the first lower bounds on the deformability and areal radii of neutron stars using gravitational waves.

79 ASTRONOMY AND ASTROPHYSICS↗

Constraints on neutrino physics from DESI DR2 BAO and DR1 full shape

The Dark Energy Spectroscopic Instrument (DESI) Collaboration has obtained robust measurements of baryon acoustic oscillations in the redshift range 0.1 < 𝑧 < 4.2, based on the Lyman-𝛼 forest and galaxies from data release 2. We combine these measurements with cosmic microwave background (CMB) data from Planck and the Atacama Cosmology Telescope to place our tightest constraints yet on the sum of neutrino masses. Assuming the cosmological Λ⁢ CDM model and three degenerate neutrino states, we find ∑𝑚 𝜈 < 0.0642 eV (95%) with a marginalized error of 𝜎⁡(∑𝑚 𝜈 ) = 0.020 eV. We also constrain the effective number of neutrino species, finding 𝑁 eff = 3.2⁢3$^{+0.35}_{−0.34}$ (95%), in line with the Standard Model prediction. When accounting for neutrino oscillation constraints, we find a preference for the normal mass ordering and an upper limit on the lightest neutrino mass of 𝑚 𝑙 < 0.023 eV (95%). However, we determine using frequentist and Bayesian methods that our constraints are in tension with the lower limits derived from neutrino oscillations. Correcting for the physical boundary at zero mass, we report a 95% Feldman-Cousins upper limit of ∑𝑚 𝜈 < 0.053 eV, breaching the lower limit from neutrino oscillations. Considering a more general Bayesian analysis with an effective cosmological neutrino mass parameter, ∑𝑚 𝜈,eff , that allows for negative energy densities and removes unsatisfactory prior weight effects, we derive constraints that are in 3⁢𝜎 tension with the same oscillation limit, while the error rises to 𝜎⁡(∑𝑚 𝜈,eff ) = 0.053 eV. In the absence of unknown systematics, this finding could be interpreted as a hint of new physics not necessarily related to neutrinos. The preference of DESI and CMB data for an evolving dark energy model offers one possible solution. In the 𝑤 0 ⁢𝑤 𝑎 ⁢CDM model, we find ∑𝑚 𝜈 < 0.163 eV (95%), relaxing the neutrino tension. These constraints all rely on the effects of neutrinos on the cosmic expansion history. Using full-shape power spectrum measurements of data release 1 galaxies, we place complementary constraints that rely on neutrino free streaming. Our strongest such limit in Λ ⁢CDM, using selected CMB priors, is ∑𝑚 𝜈 < 0.193 eV (95%).

79 ASTRONOMY AND ASTROPHYSICS↗

Joint Bayesian analysis of large angular scale CMB temperature anomalies

Cosmic microwave background measurements show an agreement with the concordance cosmology model except for a few notable anomalies: Power Suppression , the lack of large scale power in the temperature data compared to what is expected in the concordance model, and Cosmic Hemispherical Asymmetry , a dipolar breakdown of statistical isotropy. An expansion of the CMB covariance in Bipolar Spherical Harmonics naturally parametrizes both these large-scale anomalies, allowing us to perform an exhaustive, fully Bayesian joint analysis of the power spectrum and violations of statistical isotropy up to the dipole level. Our analysis sheds light on the scale dependence of the Cosmic Hemispherical Asymmetry. Assuming a scale-dependent dipole modulation model with a two-parameter power law form, we explore the posterior pdf of amplitude A($l$ = 16) and the power law index α and find the maximum a posteriori values A * (l = 16) = 0.064 ± 0.022 and $α_*$ = -0.92 ± 0.22. The maximum a posteriori direction associated with the Cosmic Hemispherical Asymmetry is ($l,b$) = (247.8 o , -19.6 o ) in Galactic coordinates, consistent with previous analyses. We evaluate the Bayes factor BSI-DM to compare the Cosmic Hemispherical Asymmetry model with the isotropic model. The data prefer but do not substantially favor the anisotropic model (BSI-DM=0.4). We consider several priors and find that this evidence ratio is robust to prior choice. The large-scale power suppression does not soften when jointly inferring both the isotropic power spectrum and the parameters of the asymmetric model, indicating no evidence that these anomalies are coupled.

79 ASTRONOMY AND ASTROPHYSICS↗

Using Gaia DR2 to constrain local dark matter density and thin dark disk

We use stellar kinematics from the latest Gaia data release (DR2) to measure the local dark matter (DM) density ρ DM in a heliocentric cylinder of radius R= 150 pc and half-height z= 200 pc. We also explore the prospect of using our analysis to estimate the DM density in local substructure by setting constraints on the surface density and scale height of a thin dark disk aligned with the baryonic disk and formed due to dissipative dark matter self-interactions. Performing the statistical analysis within a Bayesian framework for three types of tracers, we obtain ρDM = 0.016 ± 0.010 M ⊙ /pc 3 for A stars; early G stars give a similar result, while F stars yield a significantly higher value. For a thin dark disk, A stars set the strongest constraint: excluding surface densities (5–12) M ⊙ /pc 2 for scale heights below 100 pc with 95% confidence. The upper bound of this constraint implies ≲ 1% of the Milky Way DM mass is present in a dissipative dark sector. Comparing our results with those derived using Tycho-Gaia Astrometric Solution (TGAS) data, we find that the uncertainty in our measurements of the local DM content is dominated by systematic errors that arise from assumptions of our dynamical analysis in the low z region. Furthermore, there will only be a marginal reduction in these uncertainties with more data in the Gaia era. We comment on the robustness of our method and discuss potential improvements for future work.

79 ASTRONOMY AND ASTROPHYSICS↗

TOWARD CHARACTERIZATION OF THE TYPE IIP SUPERNOVA PROGENITOR POPULATION: A STATISTICAL SAMPLE OF LIGHT CURVES FROM Pan-STARRS1

In recent years, wide-field sky surveys providing deep multiband imaging have presented a new path for indirectly characterizing the progenitor populations of core-collapse supernovae (SNe): systematic light-curve studies. We assemble a set of 76 grizy-band Type IIP SN light curves from Pan-STARRS1, obtained over a constant survey program of 4 yr and classified using both spectroscopy and machine-learning-based photometric techniques. We develop and apply a new Bayesian model for the full multiband evolution of each light curve in the sample. We find no evidence of a subpopulation of fast-declining explosions (historically referred to as ''Type IIL'' SNe). However, we identify a highly significant relation between the plateau phase decay rate and peak luminosity among our SNe IIP. These results argue in favor of a single parameter, likely determined by initial stellar mass, predominantly controlling the explosions of red supergiants. This relation could also be applied for SN cosmology, offering a standardizable candle good to an intrinsic scatter of ≲ 0.2 mag. We compare each light curve to physical models from hydrodynamic simulations to estimate progenitor initial masses and other properties of the Pan-STARRS1 Type IIP SN sample. We show that correction of systematic discrepancies between modeled and observed SN IIP light-curve properties and an expanded grid of progenitor properties are needed to enable robust progenitor inferences from multiband light-curve samples of this kind. This work will serve as a pathfinder for photometric studies of core-collapse SNe to be conducted through future wide-field transient searches.

79 ASTRONOMY AND ASTROPHYSICS↗

BeyondPlanck: XV. Limits on large-scale polarized anomalous microwave emission from Planck LFI and WMAP

We constrained the level of polarized anomalous microwave emission (AME) on large angular scales using Planck Low-Frequency Instrument (LFI) and WMAP polarization data within a Bayesian cosmic microwave background (CMB) analysis framework. We modeled synchrotron emission with a power-law spectral energy distribution, as well as the sum of AME and thermal dust emission through linear regression with the Planck High-Frequency Instrument (HFI) 353 GHz data. This template-based dust emission model allowed us to constrain the level of polarized AME while making minimal assumptions on its frequency dependence. We neglected CMB fluctuations, but show through simulations that these fluctuations have a minor impact on the results. We find that the resulting AME polarization fraction confidence limit is sensitive to the polarized synchrotron spectral index prior. In addition, for prior means β s < -3.1 we find an upper limit of p AME max ≲ 0.6% (95% confidence). In contrast, for means β s = -3.0, we find a nominal detection of p AME = 2.5 ± 1.0% (95% confidence). These data are thus not strong enough to simultaneously and robustly constrain both polarized synchrotron emission and AME, and our main result is therefore a constraint on the AME polarization fraction explicitly as a function of β s . Combining the current Planck and WMAP observations with measurements from high-sensitivity low-frequency experiments such as C-BASS and QUIJOTE will be critical to improve these limits further.

79 ASTRONOMY AND ASTROPHYSICS↗

Seismic moment tensor classification using elliptical distribution functions on the hypersphere

Discrimination of underground explosions from naturally occurring earthquakes and other anthropogenic sources is one of the fundamental challenges of nuclear explosion monitoring. In an operational setting, the number of events that can be thoroughly investigated by analysts is limited by available resources. The capability to rapidly screen out events that can be robustly identified as not being explosions is, therefore, of great potential benefit. Nevertheless, possible mis-classification of explosions as earthquakes currently limits the use of screening methods for verification of test-ban treaties. Moment tensors provide a physics-based classification tool for the characterization of different seismic sources and have enabled the advent of new techniques for discriminating between earthquakes and explosions. Following normalization and projection of their six-degree vectors onto the hypersphere, existing screening approaches use spherically symmetric metrics to determine whether any new moment tensor may have been an explosion. Here, we show that populations of moment tensors for both earthquakes and explosions are anisotropically distributed on the hypersphere. Distributions possessing elliptical symmetry, such as the scaled von Mises–Fisher distribution, therefore provide a better description of these populations than the existing spherically symmetric models. We describe a method that uses these elliptical distributions in combination with a Bayesian classifier to achieve successful classification rates of 99 per cent for explosions and 98 per cent for earthquakes using existing catalogues of events from the western United States. The 1983 May 5 Crowdie underground nuclear test and 2018 July 20 DAG-1 deep-borehole chemical explosion are the only two explosions out of 140 that are incorrectly classified. Application of the method to the 2006–2017 nuclear tests in the Democratic People’s Republic of Korea yields 100 per cent identification rates and we provide a simple routine MTid for general usage. The approach provides a means to rapidly assess the likelihood of an event being an explosion and can be built into monitoring workflows that rely on simultaneously assessing multiple different discrimination metrics.

58 GEOSCIENCES↗

Gaussian process-based surrogate modeling framework for process planning in laser powder-bed fusion additive manufacturing of 316L stainless steel

Here, Laser Powder-Bed Fusion (L-PBF) metal-based additive manufacturing (AM) is complex and not fully understood. Successful processing for one material, might not necessarily apply to a different material. This paper describes a workflow process that aims at creating a material data sheet standard that describes regimes where the process can be expected to be robust. The procedure consists of building a Gaussian process-based surrogate model of the L-PBF process that predicts melt pool depth in single-track experiments given a laser power, scan speed, and laser beam size combination. The predictions are then mapped onto a power versus scan speed diagram delimiting the conduction from the keyhole melting controlled regimes. This statistical framework is shown to be robust even for cases where experimental training data might be suboptimal in quality, if appropriate physics-based filters are applied. Additionally, it is demonstrated that a high-fidelity simulation model of L-PBF can equally be successfully used for building a surrogate model, which is beneficial since simulations are getting more efficient and are more practical to study the response of different materials, than to re-tool an AM machine for new material powder.

316L stainless steel↗

A FRAMEWORK FOR QUANTIFYING THE DEGENERACIES OF EXOPLANET INTERIOR COMPOSITIONS

Several transiting super-Earths are expected to be discovered in the coming few years. While tools to model the interior structure of transiting planets exist, inferences about the composition are fraught with ambiguities. We present a framework to quantify how much we can robustly infer about super-Earth and Neptune-size exoplanet interiors from radius and mass measurements. We introduce quaternary diagrams to illustrate the range of possible interior compositions for planets with four layers (iron core, silicate mantles, water layers, and H/He envelopes). We apply our model to CoRoT-7b, GJ 436b, and HAT-P-11b. Interpretation of planets with H/He envelopes is limited by the model uncertainty in the interior temperature, while for CoRoT-7b observational uncertainties dominate. We further find that our planet interior model sharpens the observational constraints on CoRoT-7b's mass and radius, assuming the planet does not contain significant amounts of water or gas. We show that the strength of the limits that can be placed on a super-Earth's composition depends on the planet's density; for similar observational uncertainties, high-density super-Mercuries allow the tightest composition constraints. Finally, we describe how techniques from Bayesian statistics can be used to take into account in a formal way the combined contributions of both theoretical and observational uncertainties to ambiguities in a planet's interior composition. On the whole, with only a mass and radius measurement an exact interior composition cannot be inferred for an exoplanet because the problem is highly underconstrained. Detailed quantitative ranges of plausible compositions, however, can be found.

79 ASTRONOMY AND ASTROPHYSICS↗

Lithium-Ion Battery Diagnostics Using Electrochemical Impedance via Machine-Learning

Diagnosing battery states such as health, state-of-charge, or temperature is crucial for ensuring the safety and reliability of electrochemical energy storage systems. While some states, such as temperature, may be measured using cheap sensors, accurate diagnosis of battery health metrics usually requires time-consuming performance measurements, making them infeasible for use in real-world operation. These health metrics can be measured during lab-testing and then estimated on-line using predictive life models or via state observer algorithms such as Kalman filters, but these predictive methods should be supplemented by actual measurement of battery health whenever possible to ensure reliability. Rapid measurement of battery health may be done by various types of fast diagnostic techniques such as electrochemical impedance spectroscopy (EIS), which can be performed in only a few minutes and require only a fraction of the energy and power needed for a full charge and discharge measurement. But there is a substantial challenge for estimating battery health using EIS data, as EIS is sensitive to cell temperature, state-of-charge, current, and resting time in addition to health. Thus, utilizing EIS data to predict battery capacity requires correcting for all these additional variables, a task that is extremely difficult to handle analytically. This talk utilizes machine-learning methods to estimate the effectiveness of battery capacity prediction from EIS data, leveraging a data set of hundreds of EIS measurements recorded at varying temperature and state-of-charge throughout a 500-day aging study of 32 commercial, large-format NMC-Graphite lithium-ion batteries. Using EIS as input to machine-learning models is complicated by the nonlinear response of impedance to battery health, temperature, and state-of-charge, as well as the collinearity between the impedance response at neighboring frequencies, which can easily lead to overfit models. To train robust models, features from EIS data need to be extracted from the data or some subset of critical frequencies selected. Many approaches for extracting and selecting features from EIS data from electrochemical analysis and machine-learning fields were identified for analysis: using the entire raw spectra; selection of one, two, or many frequencies from the entire spectra; selecting interesting points from the EIS measurement using domain knowledge; fitting EIS with an equivalent-circuit model; calculating statistics on the raw impedance values; and reducing the dimensionality of the data using unsupervised linear (principal component analysis) and non-linear (uniform manifold approximation and projection) methods. These approaches were rigorously compared using a machine-learning pipeline approach, training linear, Gaussian process, and random forest regression models and quantifying performance using cross-validation as well as a held-out test set. An artificial neural network model trained on the raw spectra was also tested. Promising pipelines were fine-tuned via Bayesian hyperparameter optimization using cross-validation loss and training with class-specific weights to counter data set imbalance. The most reliable method for utilizing impedance in this work was the selection of two optimal frequencies through an exhaustive search, resulting in about 2% mean absolute error on test data for both Gaussian process and random forest model architectures. Interrogation of a variety of models reveals critical frequencies of 100 Hz and 103 Hz for this data set, though the optimal set of frequencies is not necessarily intuitive, i.e., the best performing models are not simply those that use impedance at frequencies that have the highest correlation to the relative discharge capacity. The best performing model is an ensemble model, which is able to predict battery capacity with 1.9% mean absolute error for unseen cells using impedance recorded at a variety of temperatures and states-of-charge.

battery↗