Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “identification problem”

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 163 records · Page 9
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

Reductive Analysis with Compiler-Guided Large Language Models for Input-Centric Code Optimizations

Input-centric program optimization aims to optimize code by considering the relations between program inputs and program behaviors. Despite its promise, a long-standing barrier for its adoption is the difficulty of automatically identifying critical features of complex inputs. This paper introduces a novel technique, reductive analysis through compiler-guided Large Language Models (LLMs), to solve the problem through a synergy between compilers and LLMs. It uses a reductive approach to overcome the scalability and other limitations of LLMs in program code analysis. The solution, for the first time, automates the identification of critical input features without heavy instrumentation or profiling, cutting the time needed for input identification by 44× (or 450× for local LLMs), reduced from 9.6 hours to 13 minutes (with remote LLMs) or 77 seconds (with local LLMs) on average, making input characterization possible to be integrated into the workflow of program compilations. Optimizations on those identified input features show similar or even better results than those identified by previous profiling-based methods, leading to optimizations that yield 92.6% accuracy in selecting the appropriate adaptive OpenMP parallelization decisions, and 20-30% performance improvement of serverless computing while reducing resource usage by 50-60%.

Input-Centric Optimization↗
DOE OSTI2021

Signal Preconditioning to Minimize Impulse Response Contribution

A study was performed to identify a method to minimize the effect of a linear time-invariant (LTI) system impulse response on an input. Three methods were studied: Wiener filter, the N4SID algorithm and transfer function estimation, the latter two using functions from MATLAB’s System Identification Toolbox. Although all three methods were able estimate an unknown forward impulse response given an input/output time series pair, only the Wiener filter was able to estimate a system inverse which satisfactorily solved the problem using a cosine similarity measure.

42 ENGINEERING↗
DOE OSTI2024

LENS: Learning Enabled Network Synthesis

RTRC and UMD have developed novel machine learning based methods under the ARPA-E DIFFERENTIATE program for rapid acceleration of hypothesis generation in complex architecture design spaces involving both discrete choices of component inclusion and interconnection and continuous parametric decisions. The project named Learning Enabled Network Synthesis (LENS) further demonstrated the developed methods on challenging electrical power converter design problems by identifying the most suitable circuit topologies and simultaneously selecting the most appropriate components to achieve optimized design of power converter with improved performances. We demonstrated that LENS could enable exploration of very large design space of circuit topologies and components by addressing the limitations of conventional design process in non-linear, high switching speed, multi-dimensional power converter design and optimization. The key innovation developed in LENS is the seamless integration of statistical learning and logical reasoning techniques and building on the individual strengths of these techniques for rapid hypothesis discovery. The main component of LENS comprises of: 1) Graph Reasoning Engine (GRE) to enforce composition rules that rapidly reject all discrete architectures that are composed incorrectly and generates an adaptive database of feasible designs which can be used by ML modules, 2) Graph Generative Learning module which is a deep neural network based generative model for graph architectures which can enable design space exploration beyond the dataset generated by the GRE, 3) Graph Reduced Order Model (ROM) for graph domains for accelerating computation of output metrics, and 4) Active learning and Rule Discovery module for sample efficient learning and extracting logical rules from the learned ML models which will be integrated in the GRE to enhance the filtering effectiveness. LENS approach can be applied to any design domains where designs can be represented as multi-attribute graphs. The LENS team integrated the various technical innovations listed above into an optimization pipeline and exercised the optimization pipeline on the converter design problem. The LENS project demonstrated that the developed AI/ML technologies can be used to generate novel converter circuits >45x faster than experts on chosen use-cases. This can enable faster design space exploration and identification of new designs which are not considered by experts due to the increasing design space complexity. This has significant potential impact on the public and energy needs of the country. It is currently estimated that 30% of all electrical powers generated passes through power converters. The future estimate is that 80% of all power generated would be passing through converters. LENS fills a critical gap in this space since by accelerating the design process the designers would be able to generate more efficient converters which can lead to significant energy savings for the country.

42 ENGINEERING↗
DOE OSTI2026

An uncertainty visualization framework for large-scale cardiovascular flow simulations: A case study on aortic stenosis

We present a generalizable uncertainty quantification (UQ) and visualization framework for lattice Boltzmann method simulations of high Reynolds number vascular flows, demonstrated on a patient-specific stenosed aorta. The framework combines EasyVVUQ for parameter sampling with large-eddy simulation turbulence modeling in HemeLB, and executes ensembles on the Frontier exascale supercomputer. Spatially resolved metrics, including entropy and isosurface-crossing probability, are used to map uncertainty in pressure and wall shear stress fields directly onto vascular geometries. Two sources of model variability are examined: inlet peak velocity and the Smagorinsky constant. Inlet velocity variation produces high uncertainty downstream of the stenosis where turbulence develops, while upstream regions remain stable. Smagorinsky constant variation has little effect on the large-scale pressure field but increases WSS uncertainty in localized high-shear regions. In both cases, the stenotic throat manifests low entropy, indicative of robust identification of elevated WSS. By linking quantitative UQ measures to three-dimensional anatomy, the framework improves interpretability over conventional 1D UQ plots and supports clinically relevant decision-making, with broad applicability to vascular flow problems requiring both accuracy and spatial insight.

Hemodynamics↗
DOE OSTI2022

Waveform Data Quality Assessment

Healthy seismoacoustic sensors and accurate metadata are critical to all the science objectives of both the Source Physics Experiment (SPE) Rock Valley Direct Comparison (RV/DC) and LowYield Physics Experiment-1 (LYNM PE-1). Both projects share similar seismoacoustic stations. We provide a plan on how to move forward with waveform data quality assessments (QA) to the working groups. The identification of stations with instrument response errors will be useful to researchers using amplitudes to study source effects and Earth attenuation. The data QA can also be useful for field technicians to help identify and correct problem sites.

47 OTHER INSTRUMENTATION↗
DOE OSTI2023

SHAPER: can you hear the shape of a jet?

The identification of interesting substructures within jets is an important tool for searching for new physics and probing the Standard Model at colliders. Many of these substructure tools have previously been shown to take the form of optimal transport problems, in particular the Energy Mover’s Distance (EMD). In this work, we show that the EMD is in fact the natural structure for comparing collider events, which accounts for its recent success in understanding event and jet substructure. We then present a Shape Hunting Algorithm using Parameterized Energy Reconstruction (SHAPER), which is a general framework for defining and computing shape-based observables. SHAPER generalizes N-jettiness from point clusters to any extended, parametrizable shape. This is accomplished by efficiently minimizing the EMD between events and parameterized manifolds of energy flows representing idealized shapes, implemented using the dual-potential Sinkhorn approximation of the Wasserstein metric. We show how the geometric language of observables as manifolds can be used to define novel observables with built-in infrared-and-collinear safety. We demonstrate the efficacy of the SHAPER framework by performing empirical jet substructure studies using several examples of new shape-based observables.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2024

Multilabel proportion prediction and out-of-distribution detection on gamma spectra of short-lived fission products

In the machine learning problem of multilabel classification, the objective is to determine for each test instance which classes the instance belongs to. In this work, we consider an extension of multilabel classification, called multilabel proportion prediction, in the context of radioisotope identification (RIID) using gamma spectra data. We aim to not only predict radioisotope proportions, but also identify out-of-distribution (OOD) spectra. We achieve this goal by viewing gamma spectra as discrete probability distributions, and based on this perspective, we develop a custom semi-supervised loss function that combines a traditional supervised loss with an unsupervised reconstruction error function. Our approach was motivated by its application to the analysis of short-lived fission products from spent nuclear fuel. In particular, we demonstrate that a neural network model trained with our loss function can successfully predict the relative proportions of 37 radioisotopes simultaneously. The model trained with synthetic data was then applied to measurements taken by Pacific Northwest National Laboratory (PNNL) to conduct analysis typically done by subject-matter experts. Here, we also extend our approach to successfully identify when measurements are OOD, and thus should not be trusted, whether due to the presence of a novel source or novel proportions.

Anomaly detection↗
DOE OSTI2024

Efficient species identification for Pacific salmon genetic monitoring programs

Abstract Genetic monitoring of Pacific salmon in the Columbia River basin provides crucial information to fisheries managers that is otherwise challenging to obtain using traditional methods. Monitoring programs such as genetic stock identification (GSI) and parentage‐based tagging (PBT) involve genotyping tens of thousands of individuals annually. Although rare, these large sample collections inevitably include misidentified species, which exhibit low genotyping success on species‐specific Genotyping‐in‐Thousands by sequencing (GT‐seq) panels. For laboratories involved in large‐scale genotyping efforts, diagnosing non‐target species and reassigning them to the appropriate monitoring program can be costly and time‐consuming. To address this problem, we identified 19 primer pairs that exhibit consistent cross‐species amplification among salmonids and contain 51 species informative variants. These genetic markers reliably discriminate among 11 salmonid species and two subspecies of Cutthroat Trout and have been included in species‐specific GT‐seq panels for Chinook Salmon, Coho Salmon, Sockeye Salmon, and Rainbow Trout commonly used for Pacific salmon genetic monitoring. The majority of species‐informative amplicons (16) were newly identified from the four existing GT‐seq panels, thus demonstrating a low‐cost approach to species identification when using targeted sequencing methods. A species‐calling script was developed that is tailored for routine GT‐seq genotyping pipelines and automates the identification of non‐target species. Following extensive testing with empirical and simulated data, we demonstrated that the genetic markers and accompanying script accurately identified species and are robust to missing genotypic data and low‐frequency, shared polymorphisms among species. Finally, we used these tools to identify Coho Salmon incidentally caught in the Columbia River Chinook Salmon sport fishery and used PBT to determine their hatchery of origin. These molecular and computing resources provide a valuable tool for Pacific salmon conservation in the Columbia River basin and demonstrate a cost‐effective approach to species identification for genetic monitoring programs.

Robinson, Zachary L.↗
DOE OSTI2024

Peptide-mimetic treatment of Pseudomonas aeruginosa in a mouse model of respiratory infection

The rise of drug resistance has become a global crisis, with >1 million deaths due to resistant bacterial infections each year. Pseudomonas aeruginosa, in particular, remains a serious problem with limited solutions due to complex resistance mechanisms that now lead to more than 32,000 multidrug-resistant (MDR) infections and over 2000 deaths in the U.S. annually. While the emergence of resistant bacteria has become ominously common, identification of useful new drug classes has been limited over the past over 40 years. We found that a potential novel therapeutic, the peptide-mimetic TM5, is effective at killing P. aeruginosa and displays sufficiently low toxicity in mammalian cells to allow for use in treatment of infections. Interestingly, TM5 kills P. aeruginosa more rapidly than traditional antibiotics, within 30–60 min in vitro, and is effective against a range of clinical isolates, including extensively drug resistant strains. In vivo, TM5 significantly reduced bacterial load in the lungs within 24 h compared to untreated mice and demonstrated few adverse effects. Taken together, these observations suggest that TM5 shows promise as an alternative therapy for MDR P. aeruginosa respiratory infections.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2022

Koopman Model Predictive Control for Eco-Driving of Automated Vehicles

In this paper, we develop a data-driven process for building a model predictive control (MPC) for eco-driving of automated vehicles. The process involves performing system identification in which the non-linear vehicle dynamics model is approximated by the Koopman operator, a linear predictor of higher state-dimension, in a data-driven framework. This approach allows us to formulate the eco-driving problem in a constrained quadratic program that leads to a computationally fast MPC. The MPC is then implemented as a closed-loop control of an electric vehicle in numerical simulations for demonstration.

autonomous vehicle↗
DOE OSTI2022

Deep Learning for Simultaneous Inference of Hydraulic and Transport Properties

Abstract Identification of a heterogeneous conductivity field and reconstruction of a contaminant release history are key aspects of subsurface remediation. These two goals are achieved by combining model predictions with sparse and noisy hydraulic head and concentration measurements. Solution of this inverse problem is notoriously difficult due to, in part, high dimensionality of the parameter space and high computational cost of repeated forward solves. We use a convolutional adversarial autoencoder (CAAE) to parameterize a heterogeneous non‐Gaussian conductivity field via a low‐dimensional latent representation. A three‐dimensional dense convolutional encoder‐decoder (DenseED) network serves as a forward surrogate of the flow and transport model. The CAAE‐DenseED surrogate is fed into the ensemble smoother with multiple data assimilation (ESMDA) algorithm to sample from the Bayesian posterior distribution of the unknown parameters, forming a CAAE‐DenseED‐ESMDA inversion framework. The resulting CAAE‐DenseED‐ESMDA inversion strategy is used to identify a three‐dimensional contaminant source and conductivity field. A comparison of the inversion results from CAAE‐ESMDA with physical flow and transport simulator and from CAAE‐DenseED‐ESMDA shows that the latter yields accurate reconstruction results at the fraction of the computational cost of the former.

Zhou, Zitong↗
DOE OSTI2023

$\text{GPLaSDI}$: Gaussian Process-based interpretable Latent Space Dynamics Identification through deep autoencoder

Numerically solving partial differential equations (PDEs) can be challenging and computationally expensive. This has led to the development of reduced-order models (ROMs) that are accurate but faster than full order models (FOMs). Recently, machine learning advances have enabled the creation of non-linear projection methods, such as Latent Space Dynamics Identification (LaSDI). LaSDI maps full-order PDE solutions to a latent space using autoencoders and learns the system of ODEs governing the latent space dynamics. By interpolating and solving the ODE system in the reduced latent space, fast and accurate ROM predictions can be made by feeding the predicted latent space dynamics into the decoder. In this paper, we introduce GPLaSDI, a novel LaSDI-based framework that relies on Gaussian process (GP) for latent space ODE interpolations. Using GPs offers two significant advantages. First, it enables the quantification of uncertainty over the ROM predictions. Second, leveraging this prediction uncertainty allows for efficient adaptive training through a greedy selection of additional training data points. This approach does not require prior knowledge of the underlying PDEs. Consequently, GPLaSDI is inherently non-intrusive and can be applied to problems without a known PDE or its residual. Here we demonstrate the effectiveness of our approach on the Burgers equation, Vlasov equation for plasma physics, and a rising thermal bubble problem. Our proposed method achieves between 200 and 100,000 times speed-up, with up to 7% relative error.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

Identification of a pathway for electron uptake in Shewanella oneidensis

Extracellular electron transfer (EET) could enable electron uptake into microbial metabolism for the synthesis of complex, energy dense organic molecules from CO 2 and renewable electricity. Theoretically EET could do this with an efficiency comparable to H 2 -oxidation but without the need for a volatile intermediate and the problems it causes for scale up. However, significant gaps remain in understanding the mechanism and genetics of electron uptake. For example, studies of electron uptake in electroactive microbes have shown a role for the Mtr EET complex in the electroactive microbe Shewanella oneidensis MR-1, though there is substantial variation in the magnitude of effect deletion of these genes has depending on the terminal electron acceptor used. This speaks to the potential for previously uncharacterized and/or differentially utilized genes involved in electron uptake. To address this, we screened gene disruption mutants for 3667 genes, representing ≈99% of all nonessential genes, from the S. oneidensis whole genome knockout collection using a redox dye oxidation assay. Confirmation of electron uptake using electrochemical testing allowed us to identify five genes from S. oneidensis that are indispensable for electron uptake from a cathode. Knockout of each gene eliminates extracellular electron uptake, yet in four of the five cases produces no significant defect in electron donation to an anode. This result highlights both distinct electron uptake components and an electronic connection between aerobic and anaerobic electron transport chains that allow electrons from the reversible EET machinery to be coupled to different respiratory processes in S. oneidensis. Homologs to these genes across many different genera suggesting that electron uptake by EET coupled to respiration could be widespread. These gene discoveries provide a foundation for: studying this phenotype in exotic metal-oxidizing microbes, genetic optimization of electron uptake in S. oneidensis; and genetically engineering electron uptake into a highly tractable host like E. coli to complement recent advances in synthetic CO2 fixation.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2024

Towards optimal sensor placement for inverse problems in spaces of measures

The objective of this work is to quantify the reconstruction error in sparse inverse problems with measures and stochastic noise, motivated by optimal sensor placement. To be useful in this context, the error quantities must be explicit in the sensor configuration and robust with respect to the source, yet relatively easy to compute in practice, compared to a direct evaluation of the error by a large number of samples. In particular, we consider the identification of a measure consisting of an unknown linear combination of point sources from a finite number of measurements contaminated by Gaussian noise. The statistical framework for recovery relies on two main ingredients: first, a convex but non-smooth variational Tikhonov point estimator over the space of Radon measures and, second, a suitable mean-squared error based on its Hellinger–Kantorovich distance to the ground truth. To quantify the error, we employ a non-degenerate source condition as well as careful linearization arguments to derive a computable upper bound. This leads to asymptotically sharp error estimates in expectation that are explicit in the sensor configuration. Thus they can be used to estimate the expected reconstruction error for a given sensor configuration and guide the placement of sensors in sparse inverse problems.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Time dependent supervisory control update with FARM using rolling window

This report describes improvements to the Feasible Actuator Range Modifier (FARM) component of the RAVEN-based HYBRID framework for analysis of Integrated Energy Systems (IES). FARM supports the HERON plug-in that solves the power dispatch problem. The solution to the dispatch problem involves economically optimal dispatches that satisfy limits on production variables and their rates of variation (explicit constraints) as well as process variables tied to the service life of equipment (implicit constraints). FARM serves to validate or confirm that a HERON solution for explicit constraints also satisfies the implicit constraints. FARM-alpha was released by Argonne National Laboratory in January 2021 followed by FARM-Beta in January 2022 with the latter providing increased flexibility for the user. In this report, FARM-Gamma, the latest version of the code, is described. The major improvement is the implementation of a system identification algorithm based on the Dynamic Mode Decomposition with Control (DMDc) coupled with a “Rolling Window” scheme that allows obtaining linear time-varying state-space models. This feature equips FARM with the most accurate approximation of system dynamics, and it relieves the user from the burden of performing an exhaustive off-line characterization of the dynamics. FARM-Gamma capabilities are assessed by solving the power dispatch problem for a representative IES unit. The simulation times corresponding to the different releases are estimated and compared. These values capture the increasing computational burden of the successively higher fidelity state-space models adopted by FARM-Alpha, FARM-Beta and FARM-Gamma. The code implementation provides significant flexibility, i.e., the user can always select the most suitable version of FARM according to the problem to be solved and the available computational resources. It is anticipated that FARM will play a role in addressing several future IES applications. We outline how it can support the coordinated management and safe operation of a nuclear plant coupled to industrial processes to produce hydrogen and synfuels.

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗
DOE OSTI2020

A Critical Review of High Burnup Fuel Fragmentation, Relocation, and Dispersal under Loss-Of-Coolant Accident Conditions

High burnup fuel in excess of ~6268 GWd/MTU has been observed to severely fragment when subjected to temperature transient conditions as in a loss-of-coolant accident. In the event of cladding failure, the severely fragmented fuel can relocate into the balloon region and potentially disperse through the burst opening into the reactor’s primary system. This observation has led to implementation of several independent test programs designed to focus on various aspects of fuel fragmentation, relocation, and dispersal. The purpose of this review is to perform a critical, holistic assessment of fuel fragmentation, relocation, and dispersal under loss-of-coolant conditions to identify data gaps in the experimental data base. Phenomena have been identified and well defined through the various testing programs, and general agreement regarding the governing parameters (temperature, burnup, heating rate, cladding deformation, etc.) has been established. However, there is a significant data gap connecting research to commercial application. Furthermore, previous research attempted to evaluate the problem holistically, but has yet to provide a comprehensive understanding of the problem. Major identified data gaps consist of (1) a comprehensive understanding of pretransient fuel rod conditions (fuel temperature, fuel stress, rod internal pressures, microstructure, etc.), (2) definition of prototypic loss-of-coolant accident conditions (heating rate, fuel temperatures, fuel stress), (3) identification of the differences between nuclear and electrical integral loss-of-coolant accident tests, and connection of fuel rod performance (cladding ballooning, burst opening, potential relocation, fragmentation susceptibility, etc.) to fuel fragmentation, relocation, and dispersal severity. Finally, the content of this manuscript provides a clear path for connecting testing and data acquisition to commercial application by providing concrete means for filling data gaps and prioritized testing regimes.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2020

Stakeholder-Engaged Structured Decision Making for the Los Alamos Legacy Cleanup Mission - 20501

The Los Alamos National Laboratory (LANL) environmental legacy cleanup program requires decisions to be made for environmental remediation, decommissioning and disposal or management of radioactive waste. This legacy cleanup program was established to address groundwater contamination, material disposal areas (MDAs) that have been used to dispose of radioactive and other waste material, and 'aggregate areas' that might produce radioactive or other chemical waste as a result of remediation activities. The LANL site is regulated for environmental concerns under the Resource Conservation and Recovery Act (RCRA). However, some parts of LANL, such as material disposal area G (MDA G), have disposed of radioactive waste under DOE Order 435.1, and are subject to other regulations. For example, decommissioning the remote-handled TRU material stored in 33 shafts at MDA G falls under EPA's 40 CFR 191. Collectively, the regulations are all aimed in the same direction of finding the best solution, either through constructs such as 'as low as reasonably achievable' (ALARA), considering balancing factors as opposed to only cost and human health risk, and, under EPA regulations such as RCRA and NEPA, evaluating impact from all chemicals and both human health and ecological endpoints. Despite the basic goals and objectives of the regulations or their guidance, the main challenge is in their implementation. Arguably perhaps, but really in principle, all of these (and similar) regulations are asking for a decision analysis to be performed. Implementation challenges encountered have included lack of understanding of decision analysis in the industry, lack of effective stakeholder engagement in the decision analysis process, and lack of appreciation of the need to separate value judgments from science, the latter leading to developing conservative, or protective, science-based models. Conservative models lead to poor solutions, lack of effective stakeholder engagement leads to long drawn out protracted approaches to finding a solution (which also might never be found with this approach), and lack of understanding of decision analysis and Bayesian statistics causes poor models to be built, which creates unfortunate situations of 'garbage in, garbage out' becoming the basis for decision making. Stakeholder-engaged structured decision making (SDM) is an approach to solving problems that relies on the theory of decision science to involve stakeholders in the decision-making process. This approach incorporates stakeholder values using a scientifically rigorous methodology that separates value judgments from science in a way that helps avoid the pitfalls of biased, protective, or conservative modeling. This approach has its foundation in Keeney's 1992 treatise on value-focused thinking [1]. Keeney advocated a paradigm shift in decision making based on the idea that the standard way of thinking about decisions is backwards. The standard approach of focusing first on identifying alternatives rather than on articulating values results in a reactive approach with the emphasis on mechanics and fixed choices instead of the core values that have meaning to stakeholders. This paradigm shift effectively engages all stakeholders in the decision-making process while using a values focused thinking approach that can lead to the identification of decision opportunities and the creation of better alternatives. The intent is to be proactive and generate solutions that are related directly to values and objectives as identified by stakeholders. There are, perhaps, two overarching reasons why SDM can be used to benefit LANL's environmental legacy cleanup. Some of LANL's remaining waste management and environmental management problems are challenging and complex (for example, the Cr and RDX plumes, and MDAs) and while the traditional approach has, arguably, worked well for relatively simple risk-based problems, it cannot, or should not, be applied to more complex problems if the most effective and efficient solutions are desired. The second reason is cost. This has perhaps become more critical since publication of the Government Accountability Office (GAO) reports that DoE's environmental liability is considered a high-risk concern for the nation [2]. The focus of SDM is on structuring solutions to decision risk problems by first addressing stakeholder and decision maker values and subsequently developing decision objectives and ways to measure those objectives, preference weighting across objectives, identifying decision alternatives that best achieve those values, and characterizing uncertainty in predictions of the measures. Because a complete decision model is created using SDM, it can be evaluated to find the main elements of the model that drive, or predict, the best solution. This approach creates complete decision models that are transparent, traceable, reproducible and technically defensible. The science behind SDM, or decision analysis, is well founded, yet it is not unusual to see ad hoc approaches to decision making implemented under various environmental regulations that are pertinent to the LANL site, including NEPA, RCRA and DOE Order 435.1. Such ad hoc approaches are often not transparent or traceable, and lack reproducibility and technical defensibility. The LANL legacy cleanup program has embarked on using SDM to address the complex problems that remain. Stakeholder meetings have been held, and a prototype version of the stakeholder value system has been developed. Further meetings are expected in the future to address specific project needs. This is a long-term endeavor considering the complex environmental problems faced by DOE EM in Los Alamos (EM-LA), and careful planning, consideration of stakeholder value systems, and engagement with stakeholders throughout the SDM process is expected to lead to a successful endpoint. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗
DOE OSTI2023

Performance Comparison of Object Detection Networks for Shrapnel Identification in Ultrasound Images

Ultrasound imaging is a critical tool for triaging and diagnosing subjects but only if images can be properly interpreted. Unfortunately, in remote or military medicine situations, the expertise to interpret images can be lacking. Machine-learning image interpretation models that are explainable to the end user and deployable in real time with ultrasound equipment have the potential to solve this problem. We have previously shown how a YOLOv3 (You Only Look Once) object detection algorithm can be used for tracking shrapnel, artery, vein, and nerve fiber bundle features in a tissue phantom. However, real-time implementation of an object detection model requires optimizing model inference time. Here, we compare the performance of five different object detection deep-learning models with varying architectures and trainable parameters to determine which model is most suitable for this shrapnel-tracking ultrasound image application. We used a dataset of more than 16,000 ultrasound images from gelatin tissue phantoms containing artery, vein, nerve fiber, and shrapnel features for training and evaluating each model. Every object detection model surpassed 0.85 mean average precision except for the detection transformer model. Overall, the YOLOv7tiny model had the higher mean average precision and quickest inference time, making it the obvious model choice for this ultrasound imaging application. Other object detection models were overfitting the data as was determined by lower testing performance compared with higher training performance. In summary, the YOLOv7tiny object detection model had the best mean average precision and inference time and was selected as optimal for this application. Next steps will implement this object detection algorithm for real-time applications, an important next step in translating AI models for emergency and military medicine.

60 APPLIED LIFE SCIENCES↗