Search NASA⌕ Search

SEARCH · Search NASA

Results for “loss function regularization”

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 73 records · Page 4

A semi-Lagrangian direct-interaction closure of the spectra of isotropic variable-density turbulence

A study of variable-density homogeneous stationary isotropic turbulence based on the sparse direct-interaction perturbation (SDIP) and supporting direct numerical simulations (DNS) is presented. The non-solenoidal flow considered here is an example of turbulent mixing of gases with different densities. The spectral statistics of this type of flow are substantially more difficult to understand theoretically than those of the similar solenoidal flows. In the approach described here, the nonlinearly coupled velocity and scalar (which determine the density of the fluid) equations are expanded in terms of a normalised density ratio parameter. A new set of coupled integro-differential SDIP equations are derived and then solved numerically for the first-order correction to the incompressible equations in the variable-density expansion parameter. By adopting a regular expansion approach, one obtains leading-order corrections that are universal and therefore interesting in their own right. The predictions are then compared with DNS of forced variable-density flow with different density contrasts. It is found that the velocity spectrum owing to variable density is indistinguishable from that of constant-density turbulence, as it is supported by a wealth of indirect experimental evidence, but the scalar spectra show significant deviations, and even loss of monotonicity, as a function of the type and strength of the large-scale source of the mixing. Furthermore, the analysis helps clarify what may be the proper approach to interpret the power spectrum of variable-density turbulence.

Mechanics↗

CRAGE-CRISPR facilitates rapid activation of secondary metabolite biosynthetic gene clusters in bacteria

With the advent of genome sequencing and mining technologies, secondary metabolite biosynthetic gene clusters (BGCs) within bacterial genomes are becoming easier to predict. For subsequent BGC characterization, clustered regularly interspaced short palindromic repeats (CRISPR) has contributed to knocking out target genes and/or modulating their expression; however, CRISPR is limited to strains for which robust genetic tools are available. Here we present a strategy that combines CRISPR with chassis-independent recombinase-assisted genome engineering (CRAGE), which enables CRISPR systems in diverse bacteria. To demonstrate CRAGE-CRISPR, we select 10 polyketide/non-ribosomal peptide BGCs in Photorhabdus luminescens as models and create their deletion and activation mutants. Subsequent loss- and gain-of-function studies confirm 22 secondary metabolites associated with the BGCs, including a metabolite from a previously uncharacterized BGC. These results demonstrate that the CRAGE-CRISPR system is a simple yet powerful approach to rapidly perturb expression of defined BGCs and to profile genotype-phenotype relationships in bacteria.

59 BASIC BIOLOGICAL SCIENCES↗

Balancing the Power-to-Load Ratio for a Novel Variable Geometry Wave Energy Converter with Nonideal Power Take-Off in Regular Waves: Preprint

This work attempts to balance power absorption against structural loading for a novel variable geometry wave energy converter. The variable geometry consists of four identical flaps that will be opened in ascending order starting with the flap closest to the seafloor and moving to the free surface. The influence of a pitch motion constraint on power absorption when utilizing a nonideal power take-off (PTO) is examined and found to reduce the losses associated with bidirectional energy flow. The power-to-load ratio is evaluated using pseudo-spectral control to determine the optimum PTO torque based on a multiterm objective function. The pseudo-spectral optimal control problem is extended to include load metrics in the objective function, which may now consist of competing terms. Separate penalty weights are attached to the surge-foundation force and PTO control torque to tune the optimizer performance to emphasize either power absorption or load shedding. PTO efficiency is not included in the objective function, but the penalty weights are utilized to limit the force and torque amplitudes, thereby reducing losses associated with bidirectional energy flow. Results from pseudo-spectral control demonstrate that shedding a portion of the available wave energy can provide greater reductions in structural loads and reactive power.

16 TIDAL AND WAVE POWER↗

High-harmonic gyrotron with sectioned cavity

High-harmonic large-orbit gyrotrons require long-length operating cavities because of both a weak electron-wave coupling and relatively low electron currents. Since diffraction Q factors of such cavities are very high, a large fraction of the radiated power is dissipated in Ohmic losses. A sectioned klystronlike cavity can be a way to combine a long electron-wave interaction region with a relatively low diffraction Q factor. In this paper, a design of a third-harmonic terahertz gyrotron is studied in detail and discussed. As compared to a regular cavity, the use of a sectioned microwave system provides an enhancement of the output rf power by several times along with the halving of the Ohmic losses.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

SIDSIM (Sub-Interval Distribution-function in Simulation) [SWR-17-22]

Time-series simulation is a technique used to estimate the energy delivery of a power generation system with output that changes under changing conditions. Such variable resources include photovoltaic energy systems that experience different levels of sunlight or wind farms with changing wind patterns. As distributed generation from these variable resources continues to flourish, having improved software tools to accurately measure, model, and estimate phenomena of interest to grid integration, in addition to energy output, will be crucial for safeguarding smooth integration and operation of these energy systems. Currently, time-series simulations assume conditions and power output are constant over the duration of a time-step (steady-state assumption). NREL researchers improve the accuracy with a new simulation software called SIDSIM (Sub-Interval Distribution Simulation) that replaces this steady value (solar or wind resource) with a distribution function that has a maximum value, a minimum value, and a shape to the curve within each time-step. The distribution function used for modeling varying values in the SIDSIM software is a novel equation and presents advantageous new benefits over the conventional methods. The first advantage comes in the ability to determine maximum instantaneous power within a time-step. This enables better protection of a system from overcurrent issues and it also allows for better forecasting and reservation of transmission capacity; such capabilities are essential for energy system integration. Second, a distribution function helps to determine the amount of energy loss due to “clipping” in cases where the photovoltaic direct current (DC) system capacity exceeds the inverter capacity, the AC interconnect capacity, or any other limiting component or threshold value. It improves estimates of battery throughput by allowing both charge and discharge of a battery within the duration of a time-step. Third, it provides a more refined estimate of utility-bill impacts by considering the delivered power may be higher than the site load or lower than the regular average power estimates that are made using a quasi-steady-state assumption. SEE ALSO OSTI ID 1,361,018

Walker, Andy↗

CaloTrilogy: Toward a Breakthrough in One-Step, End-to-End, Physics-Guided Shower Generation for Modern Calorimeters

High-precision calorimeter simulation at current and future colliders imposes rapidly growing computational demands, motivating the development of machine-learning surrogates for traditional Monte Carlo tools such as Geant4. Flow matching and diffusion-based generative models have become leading approaches for high-dimensional fast simulation because of their sample quality, but typically require ${\cal O}(100)$ function evaluations at inference and often rely on auxiliary networks to constrain global observables, compromising streamlined end-to-end generation. We introduce a unified framework that improves the balance between speed, shower quality, and physics fidelity. The method combines: (i) an average velocity field integrator that enables sampling in one or a few evaluations; (ii) a learned generative prior in shower space, constructed from data rather than random noise; and (iii) physics-guided loss terms that impose inductive biases on key observables during training. These elements are training time regularizers, preserving end-to-end inference with no additional cost. With only one or a few evaluation steps, the model achieves shower quality competitive with state-of-the-art flow and diffusion approaches, tested on several public high granularity calorimeter datasets. The results demonstrate inter-layer shower structure consistent with the underlying physics, providing a strong candidate for future fast simulation workflows.

Jiang, Cheng [Edinburgh U.]↗

A data-driven framework for sparsity-enhanced surrogates with arbitrary mutually dependent randomness

The challenge of quantifying uncertainty propagation in real-world systems is rooted in the high-dimensionality of the stochastic input and the frequent lack of explicit knowledge of its probability distribution. Traditional approaches show limitations for such problems, especially when the size of the training data is limited. To address these diculties, we have developed a general framework of constructing surrogate models on spaces of stochastic input with arbitrary probability measure irrespective of the mutual dependencies between individual components of the random inputs and the analytical form. The present Data-driven Sparsity-enhancing Rotation for Arbitrary Randomness (DSRAR) framework includes a data-driven construction of multivariate polynomial basis for arbitrary mutually dependent probability measure and a sparsity enhancement rotation procedure. This sparsity-enhancing rotation method was initially proposed in our previous work [1] for Gaussian density distributions, which may not be feasible for non-Gaussian distributions due to the loss of orthogonality after the rotation. To remedy such diculties, we developed a new data-driven approach to construct orthonormal polynomials for polynomials for arbitrary mutually dependent (amdP) randomness, ensuring the constructed basis maintains the orthogonality/near-orthogonality with respect to the density of the rotated random vector, where directly applying the regular polynomial chaos including arbitrary polynomial chaos (aPC) [2] shows limitations due to the assumption of the mutual independence between the components of the random inputs. The developed DSRAR framework leads to accurate recovery, with only limited training data, of a sparse representation of the target functions. The eectiveness of our method is demonstrated in challenging problems such as PDEs and realistic molecular systems within high-dimensional conformational space (O(10)) where the underlying density is implicitly represented by a large collection of sample data, as well as systems with explicitly given non-Gaussian probabilistic measures.

Lei, Huan↗

Physics-guided dual implicit neural representations for source separation

Significant challenges exist in efficient data analysis of most advanced experimental and observational techniques because the collected signals often include unwanted contributions, such as background and signal distortions, that can obscure the physically relevant information of interest. To address this, we have developed a self-supervised machine-learning approach for source separation using a dual implicit neural representation framework that jointly trains two neural networks: one for approximating distortions of the physical signal of interest and the other for learning the effective background contribution. Our method learns directly from the raw data by minimizing a reconstruction-based loss function without requiring labeled data or pre-defined dictionaries. We demonstrate the effectiveness of our framework by considering a challenging case study involving large-scale simulated, as well as experimental, momentum-energy-dependent inelastic neutron scattering data in a four-dimensional parameter space, characterized by heterogeneous background contributions and unknown distortions to the target signal. The method is found to successfully separate physically meaningful signals from a complex or structured background even when the signal characteristics vary across all four dimensions of the parameter space. An analytical approach that informs the choice of the regularization parameter is presented. Our method offers a versatile framework for addressing source separation problems across diverse domains, ranging from superimposed signals in astronomical measurements to structural features in biomedical image reconstructions.

47 OTHER INSTRUMENTATION↗

Enhancing Interpretability in Generative Modeling: Statistically Disentangled Latent Spaces Guided by Generative Factors in Scientific Datasets

This study addresses the challenge of statistically extracting generative factors from complex, high-dimensional datasets in unsupervised or semi-supervised settings. We investigate encoder-decoder-based generative models for nonlinear dimensionality reduction, focusing on disentangling low-dimensional latent variables corresponding to independent physical factors. Introducing Aux-VAE, a novel architecture within the classical Variational Autoencoder framework, we achieve disentanglement with minimal modifications to the standard VAE loss function by leveraging prior statistical knowledge through auxiliary variables. These variables guide the shaping of the latent space by aligning latent factors with learned auxiliary variables. We validate the efficacy of Aux-VAE through comparative assessments on multiple datasets, including astronomical simulations.

97 MATHEMATICS AND COMPUTING↗

Optimization of laser energy deposition for single-shot high aspect-ratio microstructuring of thick BK7 glass

We investigate the generation of high aspect ratio microstructures across 0.7 mm thick glass by means of single shot Bessel beam laser direct writing. We study the effect on the photoinscription of the cone angle, as well as of the energy and duration of the ultrashort laser pulse. The aim of the study is to optimize the parameters for the writing of a regular microstructure due to index modification along the whole sample thickness. By using a spectrally resolved single pulse transmission diagnostics at the output surface of the glass, we correlate the single shot material modification with observations of the absorption in different portions of the retrieved spectra, and with the absence or presence of spectral modulation. Numerical simulations of the evolution of the Bessel pulse intensity and of the energy deposition inside the sample help us interpret the experimental results that suggest to use picosecond pulses for an efficient and more regular energy deposition. Picosecond pulses take advantage of nonlinear plasma absorption and avoid temporal dynamics effects which can compromise the stationarity of the Bessel beam propagation.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Inverter Reliability Estimation for Advanced Inverter Functionality

In the near future, grid operators are expected to regularly use advanced distributed energy resource (DER) functions, defined in IEEE 1547-2018, to perform a range of grid-support operations. Many of these functions adjust the active and reactive power of the device through commanded or autonomous modes, which will produce new stresses on the grid-interfacing power electronics components, such as DC/AC inverters. In previous work, multiple DER devices were instrumented to evaluate additional component stress under multiple reactive power setpoints. We utilize quasi-static time-series simulations to determine voltage-reactive power mode (volt-var) mission profile of inverters in an active power system. Mission profiles and loss estimates are then combined to estimate the reduction of the useful life of inverters from different reactive power profiles. It was found that the average lifetime reduction was approximately 0.15% for an inverter between standard unity power factor operation and the IEEE 1547 default volt-var curve based on thermal damage due to switching in the power transistors. For an inverter with an expected 20-year lifetime, the 1547 volt-var curve would reduce the expected life of the device by 12 days. This framework for determining an inverter's useful life from experimental and modeling data can be applied to any failure mechanism and advanced inverter operation.

component degradation↗

A Physics-Driven Deep-Learning Network for Solving Nonlinear Inverse Problems

Geosteering inversion, which can be viewed as a nonlinear inverse problem, is an important technique used by directional drilling. Traditional methods that rely on iterative procedures and regularization are sensitive to the selection of initial values and can be slow due to convergence issues. In industrial applications, a lookup table is used to produce fast predictions. However, performance is not guaranteed by this approach due to the limitation of the hardware. In this paper, we propose a novel physics-driven deep-learning framework for providing a fast, accurate surrogate to solve the inverse problem. Particularly, leveraged by the forward physical model and 1D convolutional neural network (1D-CNN), the proposed method provides more reliable solutions to the inverse problem with improved performance. In addition, a new physics-driven loss function is introduced to accommodate both the model misfit and the data misfit. In conclusion, our experiments demonstrate the effectiveness of our method.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Automating Component-Level Stress Measurements for Inverter Reliability Estimation

In the near future, grid operators are expected to regularly use advanced distributed energy resource (DER) functions, defined in IEEE 1547-2018, to perform a range of grid-support operations. Many of these functions adjust the active and reactive power of the device through commanded or autonomous operating modes which induce new stresses on the power electronics components. In this work, an experimental and theoretical framework is introduced which couples laboratory-measured component stress with advanced inverter functionality and derives a reduction in useful lifetime based on an applicable reliability model. Multiple DER devices were instrumented to calculate the additional component stress under multiple reactive power setpoints to estimate associated DER lifetime reductions. A clear increase in switch loss was demonstrated as a function of irradiance level and power factor. This is replicated in the system-level efficiency measurements, although magnitudes were different—suggesting other loss mechanisms exist. Using an approximate Arrhenius thermal model for the switches, the experimental data indicate a lifetime reduction of 1.5% when operating the inverter at 0.85 PF—compared to unity PF—assuming the DER failure mechanism thermally driven within the H-bridge. If other failure mechanisms are discovered for a set of power electronics devices, this testing and calculation framework can easily be tailored to those failure mechanisms.

Flicker, Jack (ORCID:0000000212232154)↗

Microbial community shifts correspond with suppression of decomposition 25 years after liming of acidic forest soils

Microbial community structure and function regularly covary with soil pH, yet effects of these interactions on soil carbon are rarely tested experimentally within natural ecosystems. We investigated the enduring (25 year) impacts of liming on microbial community structure and decomposition at an acidic northern hardwood forest, where experimental liming increased pH one unit and surprisingly doubled the organic carbon stocks of the forest floor. Here we show that this increase in carbon storage corresponded with restructuring of the bacterial and fungal communities that drive decomposition. In the Oe horizon, liming reduced the activities of five extracellular enzymes that mediate decomposition, while the Oa horizon showed an especially large (64%) reduction in the activity of a sixth, peroxidase, which is an oxidative enzyme central to lignocellulose degradation. Decreased enzyme activities corresponded with loss of microbial taxa important for lignocellulose decay, including large reductions in the dominant ectomycorrhizal genera Russula and Cenococcum, saprotrophic and wood decaying fungi, and Actinobacteria (Thermomonosporaceae). These results demonstrate the importance of pH as a dominant regulator of microbial community structure and illustrate how changes to this structure can produce large, otherwise unexpected increases in carbon storage in forest soils.

59 BASIC BIOLOGICAL SCIENCES↗

SU-E-T-398: Evaluation of Radiobiological Parameters Using Serial Tumor Imaging During Radiotherapy as An Inverse Ill-Posed Problem

Purpose: Combination of serial tumor imaging with radiobiological modeling can provide more accurate information on the nature of treatment response and what underlies resistance. The purpose of this article is to improve the algorithms related to imaging-based radiobilogical modeling of tumor response. Methods: Serial imaging of tumor response to radiation therapy represents a sum of tumor cell sensitivity, tumor growth rates, and the rate of cell loss which are not separated explicitly. Accurate treatment response assessment would require separation of these radiobiological determinants of treatment response because they define tumor control probability. We show that the problem of reconstruction of radiobiological parameters from serial imaging data can be considered as inverse ill-posed problem described by the Fredholm integral equation of the first kind because it is governed by a sum of several exponential processes. Therefore, the parameter reconstruction can be solved using regularization methods. Results: To study the reconstruction problem, we used a set of serial CT imaging data for the head and neck cancer and a two-level cell population model of tumor response which separates the entire tumor cell population in two subpopulations of viable and lethally damage cells. The reconstruction was done using a least squared objective function and a simulated annealing algorithm. Using in vitro data for radiobiological parameters as reference data, we shown that the reconstructed values of cell surviving fractions and potential doubling time exhibit non-physical fluctuations if no stabilization algorithms are applied. The variational regularization allowed us to obtain statistical distribution for cell surviving fractions and cell number doubling times comparable to in vitro data. Conclusion: Our results indicate that using variational regularization can increase the number of free parameters in the model and open the way to development of more advanced algorithms which take into account tumor heterogeneity, for example, related to hypoxia.

60 APPLIED LIFE SCIENCES↗

Full Waveform Inversion for Seismic Velocity And Anelastic Losses in Heterogeneous Structures

We present a least-squares optimization method for solving the nonlinear full waveform inverse problem of determining the crustal velocity and intrinsic attenuation properties of sedimentary valleys in earthquake-prone regions. Given a known earthquake source and a set of seismograms generated by the source, the inverse problem is to reconstruct the anelastic properties of a heterogeneous medium with possibly discontinuous wave velocities. The inverse problem is formulated as a constrained optimization problem, where the constraints are the partial and ordinary differential equations governing the anelastic wave propagation from the source to the receivers in the time domain. This leads to a variational formulation in terms of the material model plus the state variables and their adjoints. We employ a wave propagation model in which the intrinsic energy-dissipating nature of the soil medium is modeled by a set of standard linear solids. The least-squares optimization approach to inverse wave propagation presents the well-known difficulties of ill posedness and multiple minima. To overcome ill posedness, we include a total variation regularization functional in the objective function, which annihilates highly oscillatory material property components while preserving discontinuities in the medium. To treat multiple minima, we use a multilevel algorithm that solves a sequence of subproblems on increasingly finer grids with increasingly higher frequency source components to remain within the basin of attraction of the global minimum. We illustrate the methodology with high-resolution inversions for two-dimensional sedimentary models of the San Fernando Valley, under SH-wave excitation. We perform inversions for both the seismic velocity and the intrinsic attenuation using synthetic waveforms at the observer locations as pseudoobserved data.

54 ENVIRONMENTAL SCIENCES↗

Base pairing, structural and functional insights into N 4 -methylcytidine (m 4 C) and N 4 ,N 4 -dimethylcytidine (m 4 2 C) modified RNA

The N 4 -methylation of cytidine (m 4 C and m 4 2 C) in RNA plays important roles in both bacterial and eukaryotic cells. In this work, we synthesized a series of m 4 C and m 4 2 C modified RNA oligonucleotides, conducted their base pairing and bioactivity studies, and solved three new crystal structures of the RNA duplexes containing these two modifications. Our thermostability and X-ray crystallography studies, together with the molecular dynamic simulation studies, demonstrated that m 4 C retains a regular C:G base pairing pattern in RNA duplex and has a relatively small effect on its base pairing stability and specificity. By contrast, the m 4 2 C modification disrupts the C:G pair and significantly decreases the duplex stability through a conformational shift of native Watson-Crick pair to a wobble-like pattern with the formation of two hydrogen bonds. This double-methylated m 4 2 C also results in the loss of base pairing discrimination between C:G and other mismatched pairs like C:A, C:T and C:C. The biochemical investigation of these two modified residues in the reverse transcription model shows that both mono- or di-methylated cytosine bases could specify the C:T pair and induce the G to T mutation using HIV-1 RT. In the presence of other reverse transcriptases with higher fidelity like AMV-RT, the methylation could either retain the normal nucleotide incorporation or completely inhibit the DNA synthesis. These results indicate the methylation at N 4 -position of cytidine is a molecular mechanism to fine tune base pairing specificity and affect the coding efficiency and fidelity during gene replication.

59 BASIC BIOLOGICAL SCIENCES↗

Self-consistent field theory simulations of polymers on arbitrary domains

We introduce a framework for simulating the mesoscale self-assembly of block copolymers in arbitrary confined geometries subject to Neumann boundary conditions. We employ a hybrid finite difference/volume approach to discretize the mean-field equations on an irregular domain represented implicitly by a level-set function. The numerical treatment of the Neumann boundary conditions is sharp, i.e. it avoids an artificial smearing in the irregular domain boundary. This strategy enables the study of self-assembly in confined domains and enables the computation of physically meaningful quantities at the domain interface. In addition, we employ adaptive grids encoded with Quad-/Oc-trees in parallel to automatically refine the grid where the statistical fields vary rapidly as well as at the boundary of the confined domain. This approach results in a significant reduction in the number of degrees of freedom and makes the simulations in arbitrary domains using effective boundary conditions computationally efficient in terms of both speed and memory requirement. Finally, in the case of regular periodic domains, where pseudo-spectral approaches are superior to finite differences in terms of CPU time and accuracy, we use the adaptive strategy to store chain propagators, reducing the memory footprint without loss of accuracy in computed physical observables.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗