Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “generative AI”

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 685 records · Page 38
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

Enabling HPC Scientific Workflows for Serverless

The convergence of edge computing, big data analytics, and AI with traditional scientific calculations is increasingly being adopted in HPC workflows. Workflow management systems are crucial for managing and orchestrating these complex computational tasks. However, it is difficult to identify patterns within the growing population of HPC workflows. Serverless has emerged as a novel computing paradigm, offering dynamic resource allocation, quick response time, fine-grained resource management and auto-scaling. In this paper, we propose a framework to enable HPC scientific workflows on serverless. Our approach integrates a widely used traditional HPC workflow generator with an HPC serverless workflow management system to create benchmark suites of scientific workflows with diverse characteristics. These workflows can be executed on different serverless platforms. We comprehensively compare executing workflows on traditional local containers and serverless computing platforms. Our results show that serverless can reduce CPU and memory usage respectively by 78.11% and 73.92% without compromising performance.

Andrei da silva, Anderson↗
DOE OSTI2025

An Advanced Machine Learning and Artificial Intelligence System for Demonstrating Radiation Regulatory Compliance in DOE Accelerator Facilities

In this Phase II proposal, Applied Research LLC (ARLLC), Thomas Jefferson National Accelerator Facility (Jefferson Lab), and Old Dominion University (ODU) propose the combination of domain knowledge (beam characteristics, fixed structural shielding, earthen burden (the soil and foliage added to the dome of the experimental halls as additional shielding), etc.), machine learning (ML) and/or artificial intelligence (AI) to correlate a variety of multi-modal onsite signals and the radiation fields seen in accessible areas of the accelerator site and the site boundary. The ML/AI will consider the complex influence of environmental parameters affecting the radon contribution of the measurements, focusing on actual data obtained from Jefferson Lab. In Phase I, the coded beam and location data were fed into a deep learning model to predict doses at several designated locations in Jefferson Lab’s facility. Moreover, a dense radiation map was generated using only a sparse collection of the samples in a facility. In Phase II, we will develop a software prototype containing a radiation prediction algorithm, dense radiation map algorithms, and background noise prediction algorithms, with actual data used to evaluate the prototype. This work will provide a framework for evaluation of radiation measurement results around the site based on learned responses. In addition, the proposed approach allows more granular mapping of radiation levels. Better understanding and communication of these levels is related to the overall approach in keeping doses to personnel ALARA.

43 PARTICLE ACCELERATORS↗
DOE OSTI2024

Complexes of tubulin oligomers and tau form a viscoelastic intervening network cross-bridging microtubules into bundles

Abstract The axon-initial-segment (AIS) of mature neurons contains microtubule (MT) fascicles (linear bundles) implicated as retrograde diffusion barriers in the retention of MT-associated protein (MAP) tau inside axons. Tau dysfunction and leakage outside of the axon is associated with neurodegeneration. We report on the structure of steady-state MT bundles in varying concentrations of Mg 2+ or Ca 2+ divalent cations in mixtures containing αβ-tubulin, full-length tau, and GTP at 37 °C in a physiological buffer. A concentration-time kinetic phase diagram generated by synchrotron SAXS reveals a wide-spacing MT bundle phase (B ws ), a transient intermediate MT bundle phase (B int ), and a tubulin ring phase. SAXS with TEM of plastic-embedded samples provides evidence of a viscoelastic intervening network (IN) of complexes of tubulin oligomers and tau stabilizing MT bundles. In this model, αβ-tubulin oligomers in the IN are crosslinked by tau’s MT binding repeats, which also link αβ-tubulin oligomers to αβ-tubulin within the MT lattice. The model challenges whether the cross-bridging of MTs is attributed entirely to MAPs. Tubulin-tau complexes in the IN or bound to isolated MTs are potential sites for enzymatic modification of tau, promoting nucleation and growth of tau fibrils in tauopathies.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2025

Data Centers and Digital Assurance Workshop 3 – Mitigations for Digital Assurance Risks

The third session of the TADA (Technical Assistance for Digital Assurance) Data Centers Cohort, held on November 18, 2025, focused on developing mitigation strategies for digital assurance risks identified in previous workshops. Hosted by Idaho National Laboratory (INL) and ScottMadden, the session emphasized the application of Cyber-Informed Engineering (CIE) to data center infrastructure, particularly at the utility–data center interface. Participants revisited and ranked key digital assurance risks, including architecture and interface weaknesses, governance gaps, and AI-enabled threats. The workshop introduced the 12 principles of CIE, advocating for consequence-focused design, engineered controls, and secure information architecture to proactively reduce cyber-physical vulnerabilities. These principles were applied to critical data center systems such as power distribution, UPS, cooling, SCADA/BMS, and grid-forming batteries. The session also addressed governance challenges at the interconnection boundary, highlighting the need for clear roles in telemetry sharing, firmware management, and trip settings. Special attention was given to emerging risks from behind-the-meter (BTM) generation, including reverse-power flow and the integration of small modular reactors (SMRs), which shift data centers from large loads to complex generation nodes. Participants explored how interconnection agreements can serve as enforceable instruments for digital assurance, and reviewed gaps in current standards such as NERC CIP, IEC 62443, and IEEE 1547. The workshop concluded with pathways to standardization, including model agreement language, state-level programs, and expanded NERC guidance. INL also presented tools and frameworks for secure procurement and supplier risk management, reinforcing the need for integrated engineering and policy solutions to secure the evolving data center–grid ecosystem. Session 3 of 3.

24 - POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2026

Are the U.S. Biorefineries Over the Hurdle of 2000 Ton Daily Throughput Yet?

The efficient utilization of lignocellulosic biomass for biofuel and biochemical production is hindered by material handling issues such as clogging and segregation among other challenges. Preprocessing methods such as drying, screening, and milling have improved conversion yield but have not sufficiently enhanced flowability, especially herbaceous biomass. The poor flowability of herbaceous biomass is rooted in some particle attributes that remain less altered by those methods, e.g., irregular particle shape, high roughness, and high compressibility, making it hard to scale up throughput to a key benchmark for a biorefinery – 2000 ton per day. Applying additional preprocessing methods like pelletization and torrefaction to drastically change those particle attributes can improve flow and handling but has not been comprehensively verified through test. The flowability of herbaceous biomass feedstock formats generated by three different preprocessing methods was recently assessed at Idaho National Laboratory’s Biomass Feedstock National User Facility: first, loose particles size reduced from as-received materials; second, pellets produced from an efficient densification process; and third, powders milled from torrefied pellets. Benchmarking tests including static angle of repose, basic flow energy measured in a powder rheometer, and discharge flow in an adjustable hopper, were conducted to evaluate those feedstock formats. Beyond the capacity of existing experimental apparatuses, a digital engineering approach involving flow simulations and AI models were used to identity the material attributes and processing parameters that have dominant influences on flow throughput. Techno-economic analysis focusing on hopper flow as a typical material handling operation was conducted for those feedstock formats. Perspectives will be discussed on whether the 2000-ton daily throughput for a biorefinery is achievable at an acceptable cost by using any of the tested preprocessing methods.

09 - BIOMASS FUELS↗
NASA NTRS1994

AI and workflow automation: The prototype electronic purchase request system

Automating 'paper' workflow processes with electronic forms and email can dramatically improve the efficiency of those processes. However, applications that involve complex forms that are used for a variety of purposes or that require numerous and varied approvals often require additional software tools to ensure that (1) the electronic form is correctly and completely filled out, and (2) the form is routed to the proper individuals and organizations for approval. The prototype electronic purchase request (PEPR) system, which has been in pilot use at NASA Ames Research Center since December 1993, seamlessly links a commercial electronics forms package and a CLIPS-based knowledge system that first ensures that electronic forms are correct and complete, and then generates an 'electronic routing slip' that is used to route the form to the people who must sign it. The PEPR validation module is context-sensitive, and can apply different validation rules at each step in the approval process. The PEPR system is form-independent, and has been applied to several different types of forms. The system employs a version of CLIPS that has been extended to support AppleScript, a recently-released scripting language for the Macintosh. This 'scriptability' provides both a transparent, flexible interface between the two programs and a means by which a single copy of the knowledge base can be utilized by numerous remote users.

Compton, Michael M.↗
NASA NTRS1985

Satellite operations support expert system

The Satellite Operations Support Expert System is an effort to identify aspects of satellite ground support activity which could profitably be automated with artificial intelligence (AI) and to develop a feasibility demonstration for the automation of one such area. The hydrazine propulsion subsystems (HPS) of the International Sun Earth Explorer (ISEE) and the International Ultraviolet Explorer (IUS) were used as applications domains. A demonstration fault handling system was built. The system was written in Franz Lisp and is currently hosted on a VAX 11/750-11/780 family machine. The system allows the user to select which HPS (either from ISEE or IUE) is used. Then the user chooses the fault desired for the run. The demonstration system generates telemetry corresponding to the particular fault. The completely separate fault handling module then uses this telemetry to determine what and where the fault is and how to work around it. Graphics are used to depict the structure of the HPS, and the telemetry values displayed on the screen are continually updated. The capabilities of this system and its development cycle are described.

Source record↗
DOE OSTI2026

Augmenting LLM-Based Agents for Improved Performance in Pentesting and Commissioning Operational Technology in Critical Infrastructure

Artificial intelligence (AI), and more specifically large language models (LLMs) have the potential for use in penetration testing (“pentesting”) against devices, networks, and computer systems in information technology (IT). We explore the possibility of extending pentesting from IT systems to operational technology (OT) systems, which are more obscure than IT systems in their protocols and design. A challenge therefore exists when applying pretrained LLMs to OT systems as corpora are likely to underrepresent OT systems in comparison to other more prevalent systems. We evaluate augmentations of LLMs with various methods, especially retrieval augmented generation (RAG), to improve performance of the LLMs in the OT domain. In addition to pentesting, some of the testing of these OT devices may include commissioning to ensure that the newly installed devices work correctly. Our framework may also be applied in such cases.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

From Electronic Structure to Ion Transport: Photoelectron Spectroscopy and Molecular Dynamics Simulations Reveal the Role of Anions in Lithium Battery Electrolytes

Electrolyte anions are pivotal for lithium battery performance, yet their fundamental electronic structural properties are not well understood. In this work, we employ a combination of negative-ion photoelectron spectroscopy (NIPES), ab initio calculations, and molecular dynamics (MD) simulations to investigate the electronic structures of three representative electrolyte anions. This multiscale approach enables us to elucidate how their intrinsic electronic properties govern anion–solvent interactions in gas-phase clusters, as well as lithium-ion (Li + ) solvation structures and ion transport behavior in the condensed phase. NIPES reveals that difluoro(oxalato)borate (DFOB – ), bis(fluorosulfonyl)imide (FSI – ), and bis(oxalato)borate (BOB – ) all exhibit high electron binding energies, with vertical/adiabatic detachment energies increasing from DFOB – (6.09/5.70 eV) to FSI – (6.80/6.10 eV) to BOB – (6.82/6.40 eV), correlating with enhanced oxidation stability. Ab initio calculations reveal that DFOB – /FSI – –solvent complexes bind Li + ∼ 10 kcal/mol stronger than BOB – series, aligning with the strength of a Li + –anion model. DFOB – exhibits pronounced charge localization on both oxygen and fluorine atoms, enabling their involvement in Li + coordination. In contrast, fluorine atoms in FSI – are largely electron-depleted and remain excluded from direct Li + binding. MD simulations further demonstrate that LiDFOB and LiFSI systems exhibit Li + diffusion coefficients three and five times higher than those of LiBOB across four common solvents. Notably, LiFSI salt in acetonitrile (AN) exhibits the fastest Li + diffusion among 12 electrolyte systems, highlighting the synergistic effect of FSI – and AN in promoting ion mobility. In conclusion, these findings provide a molecular-level understanding of the critical roles of anion and its microsolvation in optimizing Li + diffusion dynamics, once again emphasizing the positioning of FSI – and DFOB – as prime candidates for next-generation electrolytes.

25 ENERGY STORAGE↗
DOE OSTI2024

Neural network methods for radiation detectors and imaging

Recent advances in image data proccesing through deep learning allow for new optimization and performance-enhancement schemes for radiation detectors and imaging hardware. This enables radiation experiments, which includes photon sciences in synchrotron and X-ray free electron lasers as a subclass, through data-endowed artificial intelligence. We give an overview of data generation at photon sources, deep learning-based methods for image processing tasks, and hardware solutions for deep learning acceleration. Most existing deep learning approaches are trained offline, typically using large amounts of computational resources. However, once trained, DNNs can achieve fast inference speeds and can be deployed to edge devices. A new trend is edge computing with less energy consumption (hundreds of watts or less) and real-time analysis potential. While popularly used for edge computing, electronic-based hardware accelerators ranging from general purpose processors such as central processing units (CPUs) to application-specific integrated circuits (ASICs) are constantly reaching performance limits in latency, energy consumption, and other physical constraints. These limits give rise to next-generation analog neuromorhpic hardware platforms, such as optical neural networks (ONNs), for high parallel, low latency, and low energy computing to boost deep learning acceleration (LA-UR-23-32395).

edge computing↗
DOE OSTI2016

Development of Real-Time Quantitative Safety Evaluation Method for Nuclear Power Plants

Human operators involved with large process control systems such as nuclear power plants (NPPs) perform significant roles in the operation of systems. In this reason, deficiencies in human factors such as poor HSI (Human System Interface) design, procedure, and training, are significant contributing factors to NPPs incidents and accidents. In this regard, many kinds of studies including human reliability analysis (HRA), operation support system, automation, have been conducted in order to reduce human errors. Among these fields of studies, automation has its own specialty since it mainly focuses on eliminating the human error-inducing tasks to reduce human error, while other topics are primarily focusing on reducing human error by aiding human operators during operation. It is commonly believed that replacing human operators to the automated system would guarantee greater efficiency, lower workloads, and fewer human errors. However, these beliefs are not always true in large and complex systems such as NPP, since excessive automation can generate new roles for human operators, change human operators' activities in unexpected ways, and inducing 'out-of-the-loop' (OOTL) problem. Especially, although human error probability drastically increases during abnormal situations in NPP due to overload of information, high workload, and short time available for diagnosis, automation is not actively adopted since conventional machine learning techniques are neither capable nor reliable to handle overwhelmingly complex abnormal situations in NPP. Recently, newly developed 'deep learning' techniques have been actively applied to many fields, and some of these studies proved that the artificial intelligences (AI) based on deep learning techniques show better performance than that of human being. In 2015, deep Q-network (DQN) which is one of the deep learning techniques was developed and applied to train AI that automatically plays various Atari 2800 games, and this AI surpassed the human-level playing in many kind of games. Also in 2016, 'AlphaGo', which was developed by 'Google Deepmind' based on deep learning technique to play the game of Go, was defeated Se-dol Lee who is the World Go champion with score of 4:1. In the point of view that the game of Go is considered as one of the challenging area in machine learning field, it is verified that these new machine learning techniques can perform well, even better than the human being in high-complexity tasks. Previously, most of AI applications in nuclear field were focusing on supporting human operators' tasks such as diagnosis and decision making by rapid classification and prediction, based on fixed training data sets. However, it is expected that AI can be applied to solve more complicated problems (i.e. automation during abnormal states in NPP) by adopting advanced machine learning methodologies. In order to apply the automation system in NPP, firstly it is needed to evaluate the status of the plant quantitatively in real-time, since the information about which state is better (i.e. safe) or worse (i.e. unsafe) is essential for training properly working automation AI. Currently, there are several methodologies for NPP safety evaluation such as probabilistic safety assessment (PSA), safety performance indicator (SPI). However, existing methodologies are not feasible for development of automated system since they are not capable for the cases which are not included in the underlying model, not able to consider delicate changes in the NPP, and hard to evaluate in real-time. Therefore, development of new NPP safety evaluation methodology which can cover previously mentioned requirements is needed to conduct the study on NPP automation further. In this study, the concept of early warning score (EWS) in medical field was adopted to develop the methodology which can evaluate the safety level of the NPP quantitatively and in real-time. (authors)

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2026

Optimizing enzymes for plastic upcycling using machine learning design and high throughput experiments

Plastic use is ubiquitous in the modern world, and polyethylene terephthalate (PET) is one of the most abundantly produced plastics (and the most highly produced polyester), with ~65 million metric tons manufactured annually. To the consumer, PET is likely most recognizable as the plastic used to make beverage bottles. Like many plastics, traditional mechanical or chemical means of PET deconstruction and upcycling are costly and inefficient. Because of these challenges, recycled plastic is generally of lower quality and is more expensive to produce than virgin plastic derived from petroleum. Ultimately, this results in most plastic ending up as waste. We view plastic waste as an underutilized resource which, with the development of more efficient and high-quality recycling processes, could (1) generate significant economic value while (2) decreasing petroleum usage and greenhouse gas emissions, as well as (3) minimizing its negative environmental and health impacts. Biocatalytic recycling, or biomanufacturing the basic building blocks of new plastic from plastic waste, is a promising approach to plastic reuse that complements existing recycling technologies. Recently, biological enzymes capable of breaking down PET have garnered significant attention as an attractive means of dealing with the plastic problem. These enzymes are currently undergoing pilot studies for implementation in industrial-scale enzyme-based recycling. However, there are significant limitations to current enzymes, including the need to perform costly pre-processing of the plastic waste before the enzymes are able to work. Further optimization of these enzymes is necessary to make these technologies competitive, and ultimately incentivise industry-wide adoption of this biology-based green recycling technology. n this work we demonstrate a means to design and generate performant biological enzymes, capable of efficiently deconstructing plastic waste. Specifically, we applied recent advances in artificial intelligence, machine learning, and statistical analysis to design new versions and discover natural enzymes capable of breaking down PET. We focused on optimizing key properties that are important for industrial-scale enzymatic recycling such as pH and thermotolerance. Normal testing of enzymatic plastic-deconstruction is extremely labor intensive and so through this work we also developed a robotic-assisted experimental pipeline capable of characterizing thousands of candidate enzymes. The results of this iterative, AI-guided, multi-discipline approach have led to increases in enzymatic breakdown of over 150X over starting enzymes. This work supports the rapidly developing and transformative field of biocatalytic solutions to environmental problems beyond the discovery and predictive understanding of enzymes for polymer recycling, and has wide implications for tackling numerous energy problems such as carbon capture and fixation (e.g., engineering carbon monoxide dehydrogenase and the rubisco-pathway), biomining (e.g., design of lanthanide-binding proteins) and biomanufacturing (e.g., lignin-deconstruction enzymes).

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2023

A transfer learning approach for acoustic emission zonal localization on steel plate-like structure using numerical simulation and unsupervised domain adaptation

The detection and localization of damage in metallic structures using acoustic emission (AE) monitoring and artificial intelligence technology such as deep learning has been widely studied. However, a current challenge of this approach is the difficulty of obtaining sufficient labeled historical AE signals for the training process of deep learning models. This problem can be approached through the implementation of transfer learning. The innovation of this paper lies in the development of a transfer learning approach for AE source localization on a stainless-steel structure when no historical labeled AE signals are available for training. A finite element model is developed to generate numerical AE signals for the training. Unsupervised domain adaptation (UDA) technology is utilized to reduce the distribution difference between the numerical and the realistic AE signals and to derive the localization results of the unlabeled realistic AE signals. Finally, the results suggest that the proposed approach is capable of localizing AE signals with high accuracy in the absence of labeled training data.

42 ENGINEERING↗
DOE OSTI2024

Autonomous Inverter Controls for Resilient and Secure Grid Operation: Vector Control Design for Grid Forming

The project addresses both fundamental and practical challenges of GFM/GFL inverter control for the power grids with high inverter based resources (IBRs) penetration. A data- driven modeling technique is applied to accurately model dynamics of PWM inverters, including electromagnetic-transient (EMT). Systematic and integrative designs of grid- forming (GFM) and grid-following (GFL) primary controls are developed to guarantee system performance under either normal or abnormal operating conditions without violating constraints. This modeling and control framework provides black-start capability in case of an outage without relying on rotating generators, and its secondary control is also shown to enhance resilience against cyber-physical attacks.

14 SOLAR ENERGY↗
DOE OSTI2014

In-Situ MVA of CO 2 Sequestration Using Smart Field Technology

Capability of underground carbon dioxide storage to confine and sustain injected CO 2 for a long period of time is the main concern for geologic CO 2 sequestration. If a leakage from a geological CO 2 sequestration site occurs, it is crucial to find the approximate amount and the location of the leak, in a timely manner, in order to implement proper remediation activities. An overwhelming majority of research and development for storage site monitoring has been concentrated on atmospheric, surface or near surface monitoring of the sequestered CO 2 . This study aims to monitor the integrity of CO 2 storage at the reservoir level. This work proposes developing in-situ CO 2 Monitoring and Verification technology based on the implementation of Permanent Down-hole Gauges (PDG) or “Smart Wells” along with Artificial Intelligence and Data Mining (AI&DM). The technology attempts to identify the characteristics of the CO 2 leakage by de-convolving the pressure signals collected from Permanent Down-hole Gauges (PDG). Citronelle field, a saline aquifer reservoir, located in the U.S. was considered as the basis for this study. A reservoir simulation model for CO 2 sequestration in the Citronelle field was developed and history matched. PDGs were installed, and therefore were considered in the numerical model, at the injection well and an observation well. Upon completion of the history matching process, high frequency pressure data from PDGs were generated using the history matched numerical model using different CO 2 leakage scenarios. Since pressure signal behaviors were too complicated to de-convolute using any existing mathematical formulations, a Machine Learning-based technology was introduced for this purpose. An Intelligent Leakage Detection System (ILDS) was developed as the result of this effort using the machine learning and pattern recognition technologies. The ILDS is able to detect leakage characteristics in a short period of time (less than a day from its occurrence) demonstrating the capability of the system in quantifying leakage characteristics subject to complex rate behaviors. The performance of ILDS is examined under different conditions such as multiple well leakages, cap rock leakage, availability of an additional monitoring well, presence of pressure drift and noise in the pressure sensor and uncertainty in the reservoir model.

54 ENVIRONMENTAL SCIENCES↗
NASA NTRS2010

MCNP Simulations of Measurement of Insulation Compaction in the Cryogenic Rocket Fuel Tanks at Kennedy Space Center by Fast/Thermal Neutron Techniques

MCNP simulations have been run to evaluate the feasibility of using a combination of fast and thermal neutrons as a nondestructive method to measure of the compaction of the perlite insulation in the liquid hydrogen and oxygen cryogenic storage tanks at John F. Kennedy Space Center (KSC). Perlite is a feldspathic volcanic rock made up of the major elements Si, AI, Na, K and 0 along with some water. When heated it expands from four to twenty times its original volume which makes it very useful for thermal insulation. The cryogenic tanks at Kennedy Space Center are spherical with outer diameters of 69-70 feet and lined with a layer of expanded perlite with thicknesses on the order of 120 cm. There is evidence that some of the perlite has compacted over time since the tanks were built 1965, affecting the thermal properties and possibly also the structural integrity of the tanks. With commercially available portable neutron generators it is possible to produce simultaneously fluxes of neutrons in two energy ranges: fast (14 Me V) and thermal (25 me V). The two energy ranges produce complementary information. Fast neutrons produce gamma rays by inelastic scattering, which is sensitive to Fe and O. Thermal neutrons produce gamma rays by prompt gamma neutron activation (PGNA) and this is sensitive to Si, Al, Na, K and H. The compaction of the perlite can be measured by the change in gamma ray signal strength which is proportional to the atomic number densities of the constituent elements. The MCNP simulations were made to determine the magnitude of this change. The tank wall was approximated by a I-dimensional slab geometry with an 11/16" outer carbon steel wall, an inner stainless wall and 120 cm thick perlite zone. Runs were made for cases with expanded perlite, compacted perlite or with various void fractions. Runs were also made to simulate the effect of adding a moderator. Tallies were made for decay-time analysis from t=0 to 10 ms; total detected gamma-rays; detected gamma-rays from thermal neutron reactions d. detected gamma-rays from non-thermal neutron reactions and total detected gamma-rays as a function of depth into the annulus volume. These indicated a number of possible independent metrics of perlite compaction. For example the count rate for perlite elements increased from 3600 to 8500 cps for an increase in perlite density from 6 lbs/lcf to 16.5 lbs/cf. Thus the MCNP simulations have confirmed the feasibility of using neutron methods to map the compaction of perlite in the walls of the cryogenic tanks.

Livingston, R. A.↗
NASA NTRS2015

Certification Considerations for Adaptive Systems

Advanced capabilities planned for the next generation of aircraft, including those that will operate within the Next Generation Air Transportation System (NextGen), will necessarily include complex new algorithms and non-traditional software elements. These aircraft will likely incorporate adaptive control algorithms that will provide enhanced safety, autonomy, and robustness during adverse conditions. Unmanned aircraft will operate alongside manned aircraft in the National Airspace (NAS), with intelligent software performing the high-level decision-making functions normally performed by human pilots. Even human-piloted aircraft will necessarily include more autonomy. However, there are serious barriers to the deployment of new capabilities, especially for those based upon software including adaptive control (AC) and artificial intelligence (AI) algorithms. Current civil aviation certification processes are based on the idea that the correct behavior of a system must be completely specified and verified prior to operation. This report by Rockwell Collins and SIFT documents our comprehensive study of the state of the art in intelligent and adaptive algorithms for the civil aviation domain, categorizing the approaches used and identifying gaps and challenges associated with certification of each approach.

Bhattacharyya, Siddhartha↗
DOE OSTI2024

Analyzing inference workloads for spatiotemporal modeling

Ensuring power grid resiliency, forecasting climate conditions, and optimization of transportation infrastructure are some of the many application areas where data is collected in both space and time. Spatiotemporal modeling is about modeling those patterns for forecasting future trends and carrying out critical decision-making by leveraging machine learning/deep learning. Once trained offline, field deployment of trained models for near real-time inference could be challenging because performance can vary significantly depending on the environment, available compute resources and tolerance to ambiguity in results. Users deploying spatiotemporal models for solving complex problems can benefit from analytical studies considering a plethora of system adaptations to understand the associated performance-quality trade-offs. To facilitate the co-design of next-generation hardware architectures for field deployment of trained models, it is critical to characterize the workloads of these deep learning (DL) applications during inference and assess their computational patterns at different levels of the execution stack. In this paper, we develop several variants of deep learning applications that use spatiotemporal data from dynamical systems. We study the associated computational patterns for inference workloads at different levels, considering relevant models (Long short-term Memory, Convolutional Neural Network and Spatio-Temporal Graph Convolution Network), DL frameworks (Tensorflow and PyTorch), precision (FP16, FP32, AMP, INT16 and INT8), inference runtime (ONNX and AI Template), post-training quantization (TensorRT) and platforms (Nvidia DGX A100 and Sambanova SN10 RDU). Overall, our findings indicate that although there is potential in mixed-precision models and post-training quantization for spatiotemporal modeling, extracting efficiency from contemporary GPU systems might be challenging. Instead, co-designing custom accelerators by leveraging optimized High Level Synthesis frameworks (such as SODA High-Level Synthesizer for customized FPGA/ASIC targets) can make workload-specific adjustments to enhance the efficiency.

97 MATHEMATICS AND COMPUTING↗