Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Activation function”

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
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

Chemical Imaging for In Situ Detection and Discrimination of Aquatic Toxins Targeting Voltage Gated Sodium Channels

Biologically derived neurotoxins from cyanobacteria and algae impact environmental resources in addition to being considered a potential biological threat to human and animal health. Activity based assays are essential to detecting and responding to toxic neurotoxin events either naturally occurring or deliberate. Two toxins of interest include saxitoxin and brevetoxin. These toxins bind to and alter the function of voltage-gated sodium channels (NaV channels) which are essential for generating cell membrane action potential. We report the development and refinement of a System for Analysis at Liquid Vacuum Interface (SALVI) to assess the functional activity of saxitoxin and brevetoxin. This approach utilizes a vacuum-compatible microfluidic reactor that permits analysis at the liquid vacuum interface of human derived cells with a neurotoxin of interest in a biologically relevant environment.

54 ENVIRONMENTAL SCIENCES↗
NASA NTRS1992

A summary of the Active Flexible Wing program

This paper presents a summary of the NASA/Rockwell Active Flexible Wing program. Major elements of the program are presented. Key program accomplishments included single- and multiple-mode flutter suppression, load alleviation and load control during rapid roll maneuvers, and multi-input/multi-ouput multiple-function active controls tests above the open-loop flutter boundary.

Perry, Boyd, III↗
NASA NTRS1992

A summary of the active flexible wing program

A summary of the NASA/Rockwell Active Flexible Wing Program is presented. Major elements of the program are presented. Key program accomplishments included single- and multiple-mode flutter suppression, load alleviation and load control during rapid roll maneuvers, and multi-input/multi-output multiple-function active controls tests above the open-loop flutter boundary.

Perry, Boyd, III↗
NASA NTRS2011

From Paper to Production to Test: An Update on NASA's J-2X Engine for Exploration

The NASA/industry team responsible for developing the J-2X upper stage engine for the Space Launch System (SLS) Program has made significant progress toward moving beyond the design phase and into production, assembly, and test of development hardware. The J-2X engine exemplifies the SLS Program goal of using proven technology and experience from more than 50 years of United States spaceflight experience combined with modern manufacturing processes and approaches. It will power the second stage of the fully evolved SLS Program launch vehicle that will enable a return to human exploration of space beyond low earth orbit. Pratt & Whitney Rocketdyne (PWR) is under contract to develop and produce the engine, leveraging its flight-proven LH2/LOX, gas generator cycle J-2 and RS-68 engine capabilities, recent experience with the X-33 aerospike XRS-2200 engine, and development knowledge of the J-2S tap-off cycle engine. The J- 2X employs a gas generator operating cycle designed to produce 294,000 pounds of vacuum thrust in primary operating mode with its full nozzle extension. With a truncated nozzle extension suitable to support engine clustering on the stage, the nominal vacuum thrust level in primary mode is 285,000 pounds. It also has a secondary mode, during which it operates at 80 percent thrust by altering its mixture ratio. The J-2X development philosophy is based on proven hardware, an aggressive development schedule, and early risk reduction. NASA Marshall Space Flight Center (MSFC) and PWR began development of the J-2X in June 2006. The government/industry team of more than 600 people within NASA and PWR successfully completed the Critical Design Review (CDR) in November 2008, following extensive risk mitigation testing. Assembly of the first development engine was completed in May 2011 and the first engine test was conducted at the NASA Stennis Space Center (SSC), test stand A2, on 14 July 2011. Testing of the first development engine will continue through the autumn of 2011, be paused for test stand modifications to the passive diffuser, and then restart in the spring of 2012. This testing will be followed by specialized powerpack testing intended to examine the design and operating margins of the engine turbomachinery. The development plan beyond this point leads through more system-level, engine testing of several samples, analytical model validation activities, functional and performance verification, and then ultimate certification to support human spaceflight. This paper will discuss the J-2X development background, provide top-level information on design and development planning, and will explore some of the development challenges and mitigation activities pursued to date.

Kynard, Michael↗
NASA NTRS1978

Conditioned reflex activity of rats at later periods after the end of flight aboard the Kosmos-605 biosatellite

The aftereffects of spaceflight on the higher nervous activity of rats were studied. A five lane maze with a feeding terminal was used to check such factors as transfer of experience, the habit and speed of reaching the goal in the maze, long term memory, and the dynamics of errors. During the 3rd-7th postflight week, functional disturbances in the rat HNA were manifested in the deterioration of the capacity for the transfer of experience and for locating the feeding compartment in the maze, thus indicating a general decrease of work capacity. The increased number of errors and failures pointed to exhaustion of higher nervous processes and to the weakened functional activity of the brain.

Livshits, N. N.↗
DOE OSTI2025

Backpropagation-based learning with local derivative approximation and memory replay in biologically plausible neural systems

When learning, the brain modifies individual synaptic connections to reach a desired behavior. Animal and human brains have been shown to be incredibly capable of learning complex and varied functions across a wide variety of tasks. In recent years, artificial neural networks, inspired by human and animal brains, have shown great capabilities in learning a wide variety of difficult tasks. However, artificial neural networks primarily teach themselves through the use of backpropagation, a learning method which has no clear analogue within the brain. Additionally, Artificial Neural Networks primarily use continuous activation functions, which differ significantly from the spiking neuronal behavior present in the brain. In this paper, we discuss and demonstrate a biologically plausible learning method that approximates backpropagation through two techniques on Spiking Neural Networks. First, we show that the local temporal derivatives that are necessary for backpropagation can be approximately recovered through reconstruction using spike timings. Second, we show that through learning during a sleep phase, inspired by neuroscience research into memory replay, the localized parallel feedback path can learn to approximate the derivative through the forward path weight matrix, thus solving the weight transport problem. Lastly, we demonstrate that the combination of these two methods can approach or exceed the accuracy of backpropagation-based methods for a variety of neuromorphic vision tasks while maintaining biological plausibility.

42 ENGINEERING↗
DOE OSTI2025

Triangle Method for Dense ReLU Layers [SWR-25-72]

This software is an implementation of the methods for initializing and training neural networks to be more efficient per parameter, described more fully below and in the related publication: In theory, depth should make a ReLU network EXPONENTIALLY more efficient by enabling it to produce an exponential number of piecewise linear sections in its output. This reasoning is largely based on the work of mathematicians that have hand-constructed networks that make good use of depth. In practice however, even very deep ReLU networks that have been randomly initialized will behave identically to their shallow counterparts - missing an entire exponential dimension of efficiency. The triangle method is a first attempt at realizing the exponential potential of deep networks. Instead of randomly setting weights, we force pairs of neurons in each layer learn to build triangles (i.e. functions from [0,1] -> [0,1] that look like triangles). This is a very efficient pattern for generating lots of linear pieces because composing two triangular functions doubles the number of pieces with each composition. The triangle method is more than just a different initialization, it is a new paradigm of training. Instead of making direct updates to the matrix weights, we do an extra step of backpropagation to collect the derivatives of the loss function with respect to the shapes of the triangles, training them to tilt left or right. This process essentially holds the networks hand throughout the loss landscape and forces it to always use depth effectively by producing triangular shapes internally. This can produce several orders of magnitude of improvement on convex one-dimensional regression problems. Much more theoretical work is needed to realize its full potential beyond this context, but the implementation in this repository will still work in arbitrary numbers of dimensions. The file Triangle_Method.py is a generalized form of the method that will build each neuron its own custom 1-d convex activation function (with exponential efficiency). Example usage on one dimensional problems can be found in Example_Usage.ipynb and an example of using this in a real neural network can be found in Example_VGG16_CIFAR10.ipynb.

Milkert, Max [National Renewable Energy Laboratory↗
NASA NTRS1994

Binding and activation of major histocompatibility complex class II-deficient macrophages by staphylococcal exotoxins

Macrophages from C2D transgenic mice deficient in the expression of major histocompatibility complex (MHC) class II proteins were used to identify binding sites for superantigens distinct from the MHC class II molecule. Iodinated staphylococcal enterotoxins A and B (SEA and SEB) and exfoliative toxins A and B (ETA and ETB) bound to C2D macrophages in a concentration-dependent and competitive manner. All four toxins increased F-actin concentration within 30 s of their addition to C2D macrophages, indicating that signal transduction occurred in response to toxin in the absence of class II MHC. Furthermore, ETA, ETB, SEA, and, to a lesser extent, SEB induced C2D macrophages to produce interleukin 6. Several molecular species on C2D macrophages with molecular masses of 140, 97, 61, 52, 43, and 37 kDa bound SEA in immunoprecipitation experiments. These data indicate the presence of novel, functionally active toxin binding sites on murine macrophages distinct from MHC class II molecules.

Non-NASA Center↗
NASA NTRS1986

Nitrogen and dry-matter partitioning in soybean plants during onset of and recovery from nitrogen stress

The study tested the hypothesis that resupplying nitrogen after a period of nitrogen stress leads to restoration of the balance between root and shoot growth and normal functional activity. Nonnodulated soybean plants were grown hydroponically for 14 days with 1.0 mM NO3- in a complete nutrient solution. One set of plants was continued on the complete nutrient solution for 25 days; a second set was given 0.0 mM NO3- for 25 days; and the third set was given 0.0 mM NO3- for 10 days followed by transfer to the complete solution with 1.0 mM NO3- for 15 days. In continuously nitrogen-stressed plants, emergence and expansion of main-stem and branch leaves were severely inhibited as low nitrogen content limited further growth. This was followed by a shift in partitioning of dry matter from the leaves to the roots, resulting in an initial stimulation of root growth and a decreased shoot:root ratio. Reduced nitrogen also was redistributed from the leaves into the stem and roots. When nitrogen stress was relieved, leaf initiation and expansion were renewed. With the restoration of the balance between root and shoot function, the shoot:root ratio and distribution of reduced nitrogen within the plant organs returned to levels similar to those of nonstressed plants.

NASA Discipline Number 61-10↗
DOE OSTI2025

Shape Anisotropy-Dependent Leaking in Magnetic Neurons for Bio-Mimetic Neuromorphic Computing

Spiking neural networks seek to emulate biological computation through interconnected artificial neuron and synapse devices. Spintronic neurons can leverage magnetization physics to mimic biological neuron functions, such as integration tied to magnetic domain wall (DW) propagation in a patterned nanotrack and firing tied to the resistance change of a magnetic tunnel junction (MTJ), captured in the domain wall-magnetic tunnel junction (DW-MTJ) device. Leaking, relaxation of a neuron when it is not under stimulation, is also predicted to be implemented based on DW drift as a DW relaxes to a low energy position, but it has not been well explored or demonstrated in device prototypes. Here, in this work, we study DW-MTJ artificial neurons capable of leaky integrate-and-fire (LIF) behavior and demonstrate geometry-dependent leaking dynamics that results in repeatable, tunable LIF operation. Studying the behavior of five different device designs, we show tuning the geometry, stimulating fields and currents, and location of electrical contacts results in a wide range of neuron behavior. Additionally, implementation of an asymmetric notch allows for nonlinear pinning which increased expressivity without sacrificing leaking. The measured behavior is implemented in a simulated spiking neural network that outperforms a 1D model of continuous DW motion and approaches the performance of an ideal LIF activation function. The results show that the analog LIF capability of DW-MTJ neurons combines many desirable neuron functions into a single device, which can result in varied forms of multifunctional neuromorphic computing.

42 ENGINEERING↗
DOE OSTI2025

Human IgE monoclonal antibodies define two unusual epitopes trapping dog allergen Can f 1 in different conformations

Abstract Molecular analysis of interactions between IgE antibody and allergen allows the structural basis of IgE recognition to be defined. Human IgE (hIgE) epitopes of respiratory lipocalin allergens, including Can f 1, remain elusive due to a lack of IgE‐allergen complexes. This study aims to map the structure of allergenic epitopes on Can f 1. The fragment antigen‐binding (Fab) regions of Can f 1 specific human IgE monoclonal antibodies (hIgE mAb) were used to determine the structures of IgE epitopes. Epitope mutants were designed to target Can f 1 epitopes. Immunoassays and a human FcεRIαtransgenic mouse model of passive anaphylaxis in vivo were used to assess the functional activity of epitope mutants. Crystal structures of natural or recombinant Can f 1 complexed with two hIgE mAb 1J11 and 12F3 Fabs, respectively, were determined. The hIgE mAb bound to two partially overlapping epitopes and recognized two different Can f 1 conformations. The hIgE mAb 12F3 showed an unusual mode of binding by protruding its heavy chain CDR3 inside the Can f 1 calyx. Epitope mutants generated based on the structural analyses displayed a 64%–89% reduction in IgE antibody binding and failed to induce passive anaphylaxis in a human FcεRIαtransgenic mouse model. In summary, the structures of Can f 1‐hIgE Fab complexes revealed two unique and partially overlapping epitopes on Can f 1. The modification of the identified IgE epitopes provides a pathway for the design of hypoallergens to treat dog allergies.

Biochemistry & Molecular Biology↗
DOE OSTI2025

The Psyche Gamma-Ray and Neutron Spectrometer

A Gamma-Ray and Neutron Spectrometer (GRNS) instrument has been developed as part of the science payload for NASA’s Discovery Program Psyche mission to the M-class asteroid (16) Psyche. The GRNS instrument is designed to measure the elemental composition of Psyche with the goal to understand the origin of this mysterious, potentially metal-rich planetary body. The GRNS will measure the near-surface abundances for the elements Ni, Fe, Si, K, S, Al, and Ca, as well as the spatial distribution of Psyche’s metal-to-silicate fraction (or metal fraction). These measurements address three of the five Psyche mission science objectives: determine if Psyche is a core; determine whether small metal bodies incorporate light elements into the metal phase; and determine whether Psyche was formed under reducing conditions. The Gamma-Ray Spectrometer (GRS) uses a cryocooled, high-purity Ge (HPGe) sensor to detect cosmic-ray generated gamma rays in the 60 to 9000-keV energy range. The HPGe sensor is surrounded by a borated plastic anticoincidence shield that provides three functions: active background rejection from charged particle interactions in the HPGe sensor; fast neutron measurements; and direct measurements of the incident galactic cosmic ray flux. The Neutron Spectrometer (NS) uses three 3 He gas proportional sensors, each with different material wraps to measure thermal (<0.4 eV), low-energy epithermal (0.4 eV to 1 keV), and high-energy epithermal (up to 100 keV) neutrons. This paper provides an overview of the Psyche GRNS, including: its science and measurement objectives; the design of the instrument hardware, software, and operation; pre-launch performance measurements and its initial performance in space; and an overview of its data products and expected operation for different Psyche mission phases.

Engineering - Instrumentation related to nuclear s↗
DOE OSTI2025

From PINNs to PIKANs: recent advances in physics-informed machine learning

Physics-Informed Neural Networks (PINNs) have emerged as a key tool in Scientific Machine Learning since their introduction in 2017, enabling the efficient solution of ordinary and partial differential equations using sparse measurements. Over the past few years, significant advancements have been made in the training and optimization of PINNs, covering aspects such as network architectures, adaptive refinement, domain decomposition, and the use of adaptive weights and activation functions. A notable recent development is the Physics-Informed Kolmogorov-Arnold Networks (PIKANS), which leverage a representation model originally proposed by Kolmogorov in 1957, offering a promising alternative to traditional PINNs. In this review, we provide a comprehensive overview of the latest advancements in PINNs, focusing on improvements in network design, feature expansion, optimization techniques, uncertainty quantification, and theoretical insights. We also survey key applications across a range of fields, including biomedicine, fluid and solid mechanics, geophysics, dynamical systems, heat transfer, chemical engineering, and beyond. Lastly, we review computational frameworks and software tools developed by both academia and industry to support PINN research and applications.

Kolmogorov-Arnold networks↗
DOE OSTI2024

Interface PINNs (I-PINNs): A physics-informed neural networks framework for interface problems

Here, we present a novel physics-informed neural networks (PINNs) framework for modeling interface problems, termed Interface PINNs (I-PINNs). I-PINNs uses different neural networks for any two subdomains separated by a sharp interface such that the neural networks differ only through their activation functions while the other parameters remain identical. The performance of I-PINNs, conventional PINNs, and other existing domain-decomposition PINNs methods such as extended PINNs (XPINNs) and multi-domain PINN (M-PINN) is compared through several one-dimensional, two-dimensional, and three-dimensional benchmark elliptic interface problems. The results demonstrate that I-PINNs provides a root-mean-square-error accuracy, at least two orders of magnitude better than conventional PINNs and XPINNs at approximately one-tenth of the computational cost of conventional PINNs and half the cost of XPINNs. Additionally, while I-PINNs and M-PINN provide comparable accuracies, M-PINN is found to be approximately 50% more expensive.

42 ENGINEERING↗
DOE OSTI2025

Adaptive Interface-PINNs (AdaI-PINNs) for transient diffusion: Applications to forward and inverse problems in heterogeneous media

We model transient diffusion in heterogeneous materials using a novel physics-informed neural networks framework (PINNs) termed Adaptive interface physics-informed neural networks or AdaI-PINNs (Roy et al. arXiv preprint arXiv:2406.04626, 2024). AdaI-PINNs utilize different activation functions with trainable slopes tailored to each material region within the computational domain, allowing for a fully automated and adaptive PINNs approach to model interface problems with strongly and weakly discontinuous solutions. To enhance its performance in highly heterogeneous transient diffusion systems, we prescribe a suite of robust practices, including appropriate non-dimensionalization of equations, a biased sampling method, Glorot initialization, and the hard enforcement of boundary and initial conditions. Here we evaluate the efficacy of the proposed method on several benchmark forward and inverse problems. Comparative studies on one-dimensional and two-dimensional benchmark problems reveal that the modified AdaI-PINNs outperform its unmodified counterpart, achieving root-mean-square errors that are at least two orders of magnitude better in forward problems. For inverse problems, the maximum errors in the approximated diffusion coefficients by modified AdaI-PINNs are four orders of magnitude better than those of the unmodified version. Additionally, modified AdaI-PINNs demonstrate improved stability in problems with large material mismatches.

42 ENGINEERING↗
DOE OSTI2026

Learning interpretable surface elasticity properties from bulk properties via neural network equation learners

Surface elasticity is central to understanding the mechanics and stability of surfaces and interfaces. It is characterized by quantities such as surface tension, residual surface stress, and surface stiffness. However their analytical expressions are typically difficult to derive from atomistic data, and depend strongly on modeling choices. This work presents a neural network-based equation learner which combines customized activation functions and connection-based pruning to discover parsimonious, closed-form equations for surface elasticity from atomistic simulations. Applying the method to seven face-centered cubic (FCC) metals, our equation learner uncovers interpretable equations that describe both low-Miller index and high-Miller index surface properties, capturing long-tail property distributions accurately. The discovered expressions are decoupled into two components: a universal, geometry-driven orientation function, and material-specific baseline coefficients. We find that lower-order properties such as surface tension are fundamentally geometry dependent, while higher-order properties such as surface stress and elasticity show more complex geometry and material dependence. We also relate material dependent coefficients to bulk properties, forming a clear map from bulk material properties to surface elasticity. Overall, this approach demonstrates that interpretable neurosymbolic machine learning can bridge the gap between atomistic simulations and physical laws, enabling the discovery of generalizable structure–property relationships for materials science phenomena such as surface elasticity.

Equation learning↗
DOE OSTI2025

Annihilation-limited long-range exciton transport in high-mobility conjugated copolymer films

A combination of ultrafast, long-range, and low-loss excitation energy transfer from the photoreceptor location to a functionally active site is essential for cost-effective polymeric semiconductors. Delocalized electronic wavefunctions along π-conjugated polymer (CP) backbone can enable efficient intrachain transport, while interchain transport is generally thought slow and lossy due to weak chain–chain interactions. In contrast to the conventional strategy of mitigating structural disorder, amorphous layers of rigid CPs, exemplified by highly planar poly(indacenodithiophene-co-benzothiadiazole) (IDT-BT) donor-accepter copolymer, exhibit trap-free transistor performance and charge-carrier mobilities similar to amorphous silicon. Here, we report long-range exciton transport in HJ -aggregated IDTBT thin-film, in which the competing exciton transport and exciton–exciton annihilation (EEA) dynamics are spectroscopically separated using a phase-cycling-based scheme and shown to depart from the classical diffusion-limited and strong-coupling regime. In the thin film, we find an annihilation-limited mechanism with ≪100% per-encounter annihilation probability, facilitating the minimization of EEA-induced excitation losses. In contrast, excitons on isolated IDTBT chains diffuse over 350 nm with 0.56 cm 2 s −1 diffusivity, before eventually annihilating with unit probability on first contact. We complement the pump–probe studies with temperature-dependent photocurrent and EEA measurements from 295 K to 77 K and find a remarkable correspondence of annihilation rate and photocurrent activation energies in the 140 K to 295 K temperature range.

Shi, Yuping (ORCID:0000000318289415)↗
DOE OSTI2025

Artificial to Spiking Neural Networks Conversion with Calibration in Scientific Machine Learning

Here, we introduce a method to convert physics-informed neural networks (PINNs), commonly used in scientific machine learning, to spiking neural networks (SNNs), which are expected to have higher energy efficiency compared to traditional artificial neural networks (ANNs). We first extend the calibration technique of SNNs to arbitrary activation functions beyond ReLU, making it more versatile, and we prove a theorem that ensures the effectiveness of the calibration. We successfully convert PINNs to SNNs, enabling computational efficiency for diverse regression tasks in solving multiple differential equations, including the unsteady Navier–Stokes equations. We demonstrate great gains in terms of overall efficiency, including separable PINNs (SPINNs), which accelerate the training process. Overall, this is the first work of this kind and the proposed method achieves relatively good accuracy with low spike rates.

PINN↗