Microstructure reconstruction is a long-standing problem in experimental and computational materials science, for which numerous attempts have been made to solve. However, the majority of approaches often treats microstructure as discrete phases, which, in turn, reduces the quality of the resulting microstructures and limits its usage to the computational level of fidelity, but not the experimental level of fidelity. In this work, we applied our previously proposed approach [41] to generate synthetic microstructure images at the experimental level of fidelity for the UltraHigh Carbon Steel DataBase (UHCSDB) [13].
Interlaminar fracture is a failure mode that fiber-reinforced polymer composites (FRPC) are commonly susceptible to during loading. The strain energy release rate associated with delaminations can be the limiting factor in a laminate’s design. Standard test methods have been developed to measure the critical strain energy release rates using precracked coupons, such as the double cantilever beam (DCB). However, since the adherends in these coupons are laminates themselves, often the crack can initiate a secondary crack within one of the adherends and propagate along a secondary interface as well as the primary, precracked, interface. Deconvoluting the effects of the two cracks, a bridged ply, and multiple crack tips can turn a relatively simple test in determining a material property into a very complicated structural problem. In most cases, it is best to scrap the data collected after the crack has jumped interfaces and start with a fresh specimen. For fabric composites, this phenomenon can be quite common due to the variation in bond line thickness between plies resulting from the architecture of the fabric itself (tow size, weave architecture) as well as manufacturing flaws (voids, foreign object debris). This study aims to use the crack jumping phenomenon to learn more about the characteristics of the process zone through the insertion of designed flaws as well as design a method for evaluating the fracture properties of a toughened film adhesive in a co-cured context.
Determining process–structure–property linkages is one of the key objectives in material science, and uncertainty quantification plays a critical role in understanding both process–structure and structure–property linkages. In this work, we seek to learn a distribution of microstructure parameters that are consistent in the sense that the forward propagation of this distribution through a crystal plasticity finite element model matches a target distribution on materials properties. This stochastic inversion formulation infers a distribution of acceptable/consistent microstructures, as opposed to a deterministic solution, which expands the range of feasible designs in a probabilistic manner. To solve this stochastic inverse problem, we employ a recently developed uncertainty quantification framework based on push-forward probability measures, which combines techniques from measure theory and Bayes’ rule to define a unique and numerically stable solution. This approach requires making an initial prediction using an initial guess for the distribution on model inputs and solving a stochastic forward problem. To reduce the computational burden in solving both stochastic forward and stochastic inverse problems, we combine this approach with a machine learning Bayesian regression model based on Gaussian processes and demonstrate the proposed methodology on two representative case studies in structure–property linkages.
We describe efforts in generating synthetic malware samples that have specified behaviors that can then be used to train a machine learning (ML) algorithm to detect behaviors in malware. The idea behind detecting behaviors is that a set of core behaviors exists that are often shared in many malware variants and that being able to detect behaviors will improve the detection of novel malware. However, empirically the multi-label task of detecting behaviors is significantly more difficult than malware classification, only achieving on average 84% accuracy across all behaviors as opposed to the greater than 95% multi-class or binary accuracy reported in many malware detection studies. One of the difficulties in identifying behaviors is that while there are ample malware samples, most data sources do not include behavioral labels, which means that generally there is insufficient training data for behavior identification. Inspired by the success of generative models in improving image processing techniques, we examine and extend a 1) conditional variational auto-encoder and 2) a flow-based generative model for malware generation with behavior labels. Initial experiments indicate that synthetic data is able to capture behavioral information and increase the recall of behaviors in novel malware from 32% to 45% without increasing false positives and to 52% with increased false positives.
This report describes the high-level accomplishments from the Plasma Science and Engineering Grand Challenge LDRD at Sandia National Laboratories. The Laboratory has a need to demonstrate predictive capabilities to model plasma phenomena in order to rapidly accelerate engineering development in several mission areas. The purpose of this Grand Challenge LDRD was to advance the fundamental models, methods, and algorithms along with supporting electrode science foundation to enable a revolutionary shift towards predictive plasma engineering design principles. This project integrated the SNL knowledge base in computer science, plasma physics, materials science, applied mathematics, and relevant application engineering to establish new cross-laboratory collaborations on these topics. As an initial exemplar, this project focused efforts on improving multi-scale modeling capabilities that are utilized to predict the electrical power delivery on large-scale pulsed power accelerators. Specifically, this LDRD was structured into three primary research thrusts that, when integrated, enable complex simulations of these devices: (1) the exploration of multi-scale models describing the desorption of contaminants from pulsed power electrodes, (2) the development of improved algorithms and code technologies to treat the multi-physics phenomena required to predict device performance, and (3) the creation of a rigorous verification and validation infrastructure to evaluate the codes and models across a range of challenge problems. These components were integrated into initial demonstrations of the largest simulations of multi-level vacuum power flow completed to-date, executed on the leading HPC computing machines available in the NNSA complex today. These preliminary studies indicate relevant pulsed power engineering design simulations can now be completed in (of order) several days, a significant improvement over pre-LDRD levels of performance.
Multi-shaker testing is used to represent the response of a structure to a complex operational load in a laboratory setting. One promising method of multi-shaker testing is Impedance Matched Multi-Axis Testing (IMMAT). IMMAT targets responses at discrete measurement points to control the multiple shaker input excitations, resulting in a laboratory response representative of the expected operational response at the controlled measurement points. However, the relationship between full-field operational responses and the full-field IMMAT response has not been thoroughly explored. Poorly chosen excitation positions may match operational responses at the control points, but over or under excite uncontrolled regions of the structure. Additionally, the effectiveness of the IMMAT method on the whole test structure could depend on the type of operational excitation. Spatially distributed excitations, such as acoustic loading, may be difficult to reproduce over the whole test structure in a lab setting using the point force IMMAT excitations. This work will simulate operational and IMMAT responses of a lab-scale structure to analyze the accuracy of IMMAT at uncontrolled regions of the structure. Determination of the effect of control locations and operational locations on the IMMAT method will lead to better test design and improved predictive capabilities.
Expansion techniques have been used for many years to predict the response of un-instrumented locations on structures. These methods use a projection or transformation matrix to estimate the response at un-instrumented locations based on a sparse set of measurements. The transformation to un-instrumented locations can be done using modal projections or transmissibilities. Here, both expansion methods are implemented to demonstrate that expansion can be used for acoustic problems, where a sparse set of pressure measurements, say from a set of microphones in a cavity or room, are used to expand and predict the response at any location in the domain. The modal projection method is applied to a small acoustic cavity, where the number of active modes is small, and the transmissibility method is applied to a large acoustic domain, where the number of active modes is very large. In each case, expansion is shown to work well, though each case has its benefits and drawbacks. The numerical studies shown here indicate that expansion could be accurate and therefore useful for a wide range of interior acoustic problems where only sparse measurements are available, but full-field information is desired, such as field reconstruction problems, or model validation problems.
The Sodium Chemistry (NAC) package in MELCOR has been developed to enhance application to sodium cooled fast reactor. The models in the NAC package have been assessed through benchmark analyses. The F7-1 sodium pool fire experimental analysis is conducted within the framework of the U.S.-Japan collaboration under the Civil Nuclear Energy Research and Development Working Group. This study assesses the capability of the improved models proposed for the sodium pool fire in MELCOR through comparison with the F7-1 experimental data and the results of the SPHINCS code developed in Japan. Pool heat transfer, pool oxide layer, and pool spreading models are improved in this study to mitigate the deviations exhibited in the previous study where the original CONTAIN-LMR models are used: the overestimation of combustion rate and associated temperature during the initial phase of the sodium fire relative to the experimental data and SPHINCS results, and the underestimation of them during the later phase. The analytical result of the improved sodium pool fire model agrees well with the experimental data and SPHINCS results over the entire course of the sodium fire. This study illustrates these enhanced capabilities for MELCOR to reliably simulate sodium pool fire events.
Industrial and security communities leverage x-ray computed tomography for several applications in non-destructive evaluation such as material detection and metrology. Many of these applications ultimately reach a limit as most x-ray systems have a nonlinear mathematical operator due to the Bremsstrahlung radiation emitted from the x-ray source. This work proposes a design of a multi-metal pattered anode coupled with a hyperspectral X-ray detector to improve spatial resolution, absorption signal, and overall data quality for various quantitative. The union of a multi-metal pattered anode x-ray source with an energy-resolved photon counting detector permits the generation and detection of a preferential set of X-ray energy peaks. When photons about the peaks are detected, while rejecting photons outside this neighborhood, the overall quality of the image is improved by linearizing the operator that defines the image formation. Additionally, the effective X-ray focal spot size allows for further improvement of the image quality by increasing resolution. Previous works use machine learning techniques to analyze the hyperspectral computed tomography signal and reliably identify and discriminate a wide range of materials based on a material's composition, improving data quality through a multi-material pattern anode will further enhance these identification and classification methods. This work presents initial investigations of a multi-metal patterned anode along with a hyperspectral detector using a general-purpose Monte Carlo particle transport code known as PHITS version 3.24. If successful, these results will have tremendous impact on several nondestructive evaluation applications in industry, security, and medicine.
Six degree of freedom shaker tests are becoming more popular as they save testing time because they test a component in multiple directions in one test rather than executing multiple tests in one direction at a time. However, there are several difficulties in conducting a component six degree of freedom shaker test in a way that adequately replicates the component field stress. One difficulty is knowing if a classical rigid test fixture will produce component modes that span the displacement space of the component in the field environment. If the modes of the component while attached to a rigid fixture do not span the space of the component in the field environment, then the test will be unable to replicate that motion and corresponding stresses. This paper will examine the motion of the Removable Component of the BARC hardware in an field assembly and calculate the modal projection error expected by executing a six degree of freedom shaker test on a rigid fixture. The paper will conclude by examining the data and comparing it to the pre-test predictions of error calculated by the modal projection error.
This report discusses the progress on the collaboration between Sandia National Laboratories (Sandia) and Japan Atomic Energy Agency (JAEA) on the sodium fire research in fiscal year 2020. First, the current sodium pool fire model in MELCOR, which is adapted from CONTAIN-LMR code, is discussed. The associated sodium fire input requirements are also presented. These input requirements are flexible enough to permit further model development via control functions to enhance the current model without modifying the source code. The theoretical pool fire model improvement developed at Sandia is discussed. A control function model has been developed from this improvement. Then, the validation study of the sodium pool fire model in MELCOR carried out by both Sandia and JAEA’s staff is described. To validate this pool fire model with the enhancement, a JAEA sodium pool fire experiment (F7-1 test) is used. The results of the calculation are discussed as well as suggestions for further model improvement. Finally, recommendations are made for new MELCOR simulations for next fiscal year, 2021.
The Sandia Critical Experiments (SCX) Program provides a specialized facility for performing water moderated and reflected critical experiments with UO2 fuel rod arrays. A history of safe reactor operations and flexibility in reactor core configuration has resulted in the completion of several benchmark critical experiment evaluations that are published in the International Criticality Safety Benchmark Evaluation Project (ICSBEP) Handbook. The LEUCOMP-THERM-078 and LEU-COMP-THERM-080 evaluations from the handbook provide similar cases for reference. The set of experiments described here were performed using the Seven Percent Critical Experiment (7uPCX) fuel to measure the effects of decreasing the fuel-to-water volume ratio on the critical array size. This was accomplished by using fuel loading patterns to effectively increase the pitch of the fuel arrays in the assembly. The fuel rod pitch variations provided assembly configurations that ranged from strongly undermoderated to slightly overmoderated.
Degraded visual environments like fog pose a major challenge to safety and security because light is scattered by tiny particles. We show that by interpreting the scattered light it is possible to detect, localize, and characterize objects normally hidden in fog. First, a computationally efficient light transport model is presented that accounts for the light reflected and blocked by an opaque object. Then, statistical detection is demonstrated for a specified false alarm rate using the Neyman-Pearson lemma. Finally, object localization and characterization are implemented using the maximum likelihood estimate. These capabilities are being tested at the Sandia National Laboratory Fog Chamber Facility.
The Space Nuclear Thermal Propulsion (SNTP) project was an attempt to create a more powerful and more efficient rocket engine utilizing nuclear technologies. As part of this project a zero-power critical assembly referred to as SNTPCX was designed and installed at Sandia National Laboratories. The SNTP-CX was a light water moderated particle bed reactor utilizing highly enriched uranium fuel in the form of UC particles. The SNTP-CX performed 142 runs covering numerous experiments from the year 1989 to 1992. The program was canceled in 1994 as the nation’s priorities shifted. Now these experiments are being evaluated for use as criticality safety benchmarks. Nineteen of the 142 reactor runs were dedicated to a series of experiments to calculate the worth of the boron used in the light water moderator. This series of experiments has been selected for further evaluation as a critical benchmark for the International Criticality Safety Benchmark Evaluation Project (ICSBEP).
Phenomenology-Informed (PI) Machine Learning is introduced to address the unique challenges faced when applying modern machine-learning object recognition techniques to the SAR domain. PI-ML includes a collection of data normalization and augmentation techniques inspired by successful SAR ATR algorithms designed to bridge the gap between simulated and real-world SAR data for use in training Convolutional Neural Networks (CNNs) that perform well in the low-noise, feature-dense space of camera-based imagery. The efficacy of PI-ML will be evaluated using ResNet, EfficientNet, and other networks, using both traditional training techniques and all-SAR transfer learning.
Bayesian optimization (BO) is a flexible and powerful framework that is suitable for computationally expensive simulation-based applications and guarantees statistical convergence to the global optimum. While remaining as one of the most popular optimization methods, its capability is hindered by the size of data, the dimensionality of the considered problem, and the nature of sequential optimization. These scalability issues are intertwined with each other and must be tackled simultaneously. In this work, we propose the Scalable3-BO framework, which employs sparse GP as the underlying surrogate model to scope with Big Data and is equipped with a random embedding to efficiently optimize high-dimensional problems with low effective dimensionality. The Scalable3-BO framework is further leveraged with asynchronous parallelization feature, which fully exploits the computational resource on HPC within a computational budget. As a result, the proposed Scalable3-BO framework is scalable in three independent perspectives: with respect to data size, dimensionality, and computational resource on HPC. The goal of this work is to push the frontiers of BO beyond its well-known scalability issues and minimize the wall-clock waiting time for optimizing high-dimensional computationally expensive applications. We demonstrate the capability of Scalable3-BO with 1 million data points, 10,000-dimensional problems, with 20 concurrent workers in an HPC environment.
Phenomenology-Informed (PI) Machine Learning is introduced to address the unique challenges faced when applying modern machine-learning object recognition techniques to the SAR domain. PI-ML includes a collection of data normalization and augmentation techniques inspired by successful SAR ATR algorithms designed to bridge the gap between simulated and real-world SAR data for use in training Convolutional Neural Networks (CNNs) that perform well in the low-noise, feature-dense space of camera-based imagery. The efficacy of PI-ML will be evaluated using ResNet, EfficientNet, and other networks, using both traditional training techniques and all-SAR transfer learning.
Process-structure linkage is one of the most important topics in materials science due to the fact that virtually all information related to the materials, including manufacturing processes, lies in the microstructure itself. Therefore, to learn more about the process, one must start by thoroughly examining the microstructure. This gives rise to inverse problems in the context of process-structure linkages, which attempt to identify the processes that were used to manufacturing the given microstructure. In this work, we present an inverse problem for structure-process linkages which we solve using asynchronous parallel Bayesian optimization which exploits parallel computing resources. We demonstrate the effectiveness of the method using kinetic Monte Carlo model for grain growth simulation.
We measure the photoelastic constants of piezo-optomechanical photonic integrated circuits incorporating a specially formulated, silicon-depleted silicon nitride thin films using a laser doppler vibrometer to calibrate the strain produced by the integrated piezoelectric actuators.
Industrial and security communities leverage x-ray computed tomography for several applications in non-destructive evaluation such as material detection and metrology. Many of these applications ultimately reach a limit as most x-ray systems have a nonlinear mathematical operator due to the Bremsstrahlung radiation emitted from the x-ray source. This work proposes a design of a multi-metal pattered anode coupled with a hyperspectral X-ray detector to improve spatial resolution, absorption signal, and overall data quality for various quantitative. The union of a multi-metal pattered anode x-ray source with an energy-resolved photon counting detector permits the generation and detection of a preferential set of X-ray energy peaks. When photons about the peaks are detected, while rejecting photons outside this neighborhood, the overall quality of the image is improved by linearizing the operator that defines the image formation. Additionally, the effective X-ray focal spot size allows for further improvement of the image quality by increasing resolution. Previous works use machine learning techniques to analyze the hyperspectral computed tomography signal and reliably identify and discriminate a wide range of materials based on a material's composition, improving data quality through a multi-material pattern anode will further enhance these identification and classification methods. This work presents initial investigations of a multi-metal patterned anode along with a hyperspectral detector using a general-purpose Monte Carlo particle transport code known as PHITS version 3.24. If successful, these results will have tremendous impact on several nondestructive evaluation applications in industry, security, and medicine.