Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “uncertainty”

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 739 records · Page 41

Reducing the sensitivity of IMPT treatment plans to setup errors and range uncertainties via probabilistic treatment planning

Treatment plans optimized for intensity modulated proton therapy (IMPT) may be very sensitive to setup errors and range uncertainties. If these errors are not accounted for during treatment planning, the dose distribution realized in the patient may by strongly degraded compared to the planned dose distribution. The authors implemented the probabilistic approach to incorporate uncertainties directly into the optimization of an intensity modulated treatment plan. Following this approach, the dose distribution depends on a set of random variables which parameterize the uncertainty, as does the objective function used to optimize the treatment plan. The authors optimize the expected value of the objective function. They investigate IMPT treatment planning regarding range uncertainties and setup errors. They demonstrate that incorporating these uncertainties into the optimization yields qualitatively different treatment plans compared to conventional plans which do not account for uncertainty. The sensitivity of an IMPT plan depends on the dose contributions of individual beam directions. Roughly speaking, steep dose gradients in beam direction make treatment plans sensitive to range errors. Steep lateral dose gradients make plans sensitive to setup errors. More robust treatment plans are obtained by redistributing dose among different beam directions. This can be achieved by the probabilistic approach. In contrast, the safety margin approach as widely applied in photon therapy fails in IMPT and is neither suitable for handling range variations nor setup errors.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

SU-E-T-367: Evaluate the Dosimetric Impacts On Patient Specific Treatment Plans Due to Set Up Uncertainties During LINAC Annual QA and Commission

Purpose: To quantify the dosimetric impact on patient’s specific treatment plans due to set up uncertainties during LINAC commission and annual QA and to determine the maximum set up uncertainty allowance range. Methods: A 60×60×60 cm{sup 3} solid water cube was created in Varian Eclipse TPS. Beam data profiles (crossline and diagonal) and PDDs for field sizes ranging from 2×2 cm{sup 2} to 40×40 cm{sup 2} were simulated. Three main uncertainty scenarios were purposely introduced for gantry position tilts (0–5°), source axis distance changes (100–105 cm), and iso-center position shifts (0–5 mm) during the simulation. A gamma analysis was used to compare the correct simulated profiles with the profiles for each scenario. Two static IMRT treatment plans (H&N and GYN) with tumors at 5 cm and 15 cm depths were compared using similar set up uncertainties. Results: A gamma analysis using ±3%/±3mm with 90% passing rate criteria is included to show the passing rate for each scenario. Crossline and diagonal profiles showed a gamma passing rating of ≥ 90% at depth ≤10 cm for these scenarios: gantry tilted from 0–5°, SAD changed from 100–105 cm, and iso-center shifted ≤ 4 mm. From 10 to 20 cm depths, all three scenarios failed with gamma passing ≤ 90% excepted for diagonal profiles at Gantry =2°, SAD =1 cm, and iso-center =1 mm off center. Diagonal profiles showed a higher gamma passing rating compared to crossline profiles for all three scenarios. PDD differences also increased as depth increased. For patient’s specific treatment plans, maximum uncertainties allowed to obtain a ≥90% gamma passing rating are: gantry tilts ±1 degree, SAD shifts ±2 cm, and iso-center moves ±3 mm. Conclusion: This study validated AAPM TG 142 recommendations on the mechanical and dosimetry uncertainties and provided proofs on maximum acceptance tolerances for LINAC annual QA and commission.

60 APPLIED LIFE SCIENCES↗

SU-F-T-14: Dosimetric Impacts of Various Uncertainties in Cervical Cancer HDR Brachytherapy: Are Conventional Point Doses Good Surrogates for 3D Dosimetry?

Purpose: In the context of evaluating dosimetric impacts of a variety of uncertainties involved in HDR Tandem-and-Ovoid treatment, to study the correlations between conventional point doses and 3D volumetric doses. Methods: For 5 cervical cancer patients treated with HDR T&O, 150 plans were retrospectively created to study dosimetric impacts of the following uncertainties: (1) inter-fractional applicator displacement between two treatment fractions within a single insertion by applying Fraction#1 plan to Fraction#2 CT; (2) positional dwell error simulated from −5mm to 5mm in 1mm steps; (3) simulated temporal dwell error of 0.05s, 0.1s, 0.5s, and 1s. The original plans were based on point dose prescription, from which the volume covered by the prescription dose was generated as the pseudo target volume to study the 3D target dose effect. OARs were contoured. The point and volumetric dose errors were calculated by taking the differences between original and simulated plans. The correlations between the point and volumetric dose errors were analyzed. Results: For the most clinically relevant positional dwell uncertainty of 1mm, temporal uncertainty of 0.05s, and inter-fractional applicator displacement within the same insertion, the mean target D90 and V100 deviation were within 1%. Among these uncertainties, the applicator displacement showed the largest potential target coverage impact (2.6% on D90) as well as the OAR dose impact (2.5% and 3.4% on bladder D2cc and rectum D2cc). The Spearman correlation analysis shows a correlation coefficient of 0.43 with a p-value of 0.11 between target D90 coverage and H point dose. Conclusion: With the most clinically relevant positional and temporal dwell uncertainties and patient interfractional applicator displacement within the same insertion, the dose error is within clinical acceptable range. The lack of correlation between H point and 3D volumetric dose errors is a motivator for the use of 3D treatment planning in cervical HDR brachytherapy.

60 APPLIED LIFE SCIENCES↗

SU-F-T-221: An Assessment of the Potential for Improved Local Control of Skull- Base Chordomas Via Reduction of the Proton Beam Range Uncertainty

Purpose: The beam range uncertainty presents a special challenge for proton therapy. Novel technologies currently under development offer strategies to reduce the range uncertainty [1,2]. This work quantifies the potential advantages that could be realized by such a reduction for dosimetrically challenging chordomas at the base of skull. Therapeutic improvement was assessed by evaluating tumor control probabilities (TCP) and normal tissue complication probabilities (NTCP). Methods: Treatment plans were made for a modulated-scanned proton delivery technique using the Eclipse treatment planning system. The prescription dose was 7920 cGy to the CTV. Three different range uncertainty scenarios were considered: 5 mm (3.5% of the beam range + 1 mm, representing current clinical practice, “Curr”), 2 mm (1.3%), and 1 mm (0.7%). For each of 4 patients, 3 different PTVs were defined via uniform expansion of the CTV by the value of the range uncertainty. Tumor control probability (TCP) and normal tissue complication probabilities (NTCPs) for organs-at-risk (OARs) were calculated using the Lyman-Kutcher-Burman[3] formalism and published model parameters [ref Terahara[4], quantec S10, Burman Red Journal v21 pp 123]. Our plan optimization strategy was to achieve PTV close to prescription while maintaining OAR NTCP values at or better than the Curr plan. Results: The average TCP values for the 5, 2, and 1 mm range uncertainty scenarios are 51%, 55% and 65%. The improvement in TCP for patients was between 4 and 30%, depending primarily on the proximity of the GTV to OAR. The average NTCPs for the brainstem and cord were about 4% and 1%, respectively, for all target margins. Conclusion: For base of skull chordomas, reduced target margins can substantially increase the TCP without increasing the NTCP. This work demonstrates the potential significance of a reduction in the range uncertainty for proton beams.

60 APPLIED LIFE SCIENCES↗

SU-G-BRA-09: Estimation of Motion Tracking Uncertainty for Real-Time Adaptive Imaging

Purpose: kV fluoroscopic imaging combined with MV treatment beam imaging has been investigated for intrafractional motion monitoring and correction. It is, however, subject to additional kV imaging dose to normal tissue. To balance tracking accuracy and imaging dose, we previously proposed an adaptive imaging strategy to dynamically decide future imaging type and moments based on motion tracking uncertainty. kV imaging may be used continuously for maximal accuracy or only when the position uncertainty (probability of out of threshold) is high if a preset imaging dose limit is considered. In this work, we propose more accurate methods to estimate tracking uncertainty through analyzing acquired data in real-time. Methods: We simulated motion tracking process based on a previously developed imaging framework (MV + initial seconds of kV imaging) using real-time breathing data from 42 patients. Motion tracking errors for each time point were collected together with the time point’s corresponding features, such as tumor motion speed and 2D tracking error of previous time points, etc. We tested three methods for error uncertainty estimation based on the features: conditional probability distribution, logistic regression modeling, and support vector machine (SVM) classification to detect errors exceeding a threshold. Results: For conditional probability distribution, polynomial regressions on three features (previous tracking error, prediction quality, and cosine of the angle between the trajectory and the treatment beam) showed strong correlation with the variation (uncertainty) of the mean 3D tracking error and its standard deviation: R-square = 0.94 and 0.90, respectively. The logistic regression and SVM classification successfully identified about 95% of tracking errors exceeding 2.5mm threshold. Conclusion: The proposed methods can reliably estimate the motion tracking uncertainty in real-time, which can be used to guide adaptive additional imaging to confirm the tumor is within the margin or initialize motion compensation if it is out of the margin.

60 APPLIED LIFE SCIENCES↗

On the Power of Uncertainties in Microbial System Modeling: No Need To Hide Them Anymore

For decades, microbiologists have considered uncertainties as an undesired side effect of experimental protocols. As a consequence, standard microbial system modeling strives to hide uncertainties for the sake of deterministic understanding. However, recent studies have highlighted greater experimental variability than expected and emphasized uncertainties not as a weakness but as a necessary feature of complex microbial systems. We therefore advocate that biological uncertainties need to be considered foundational facets that must be incorporated in models. Not only will understanding these uncertainties improve our understanding and identification of microbial traits, it will also provide fundamental insights on microbial systems as a whole. Taking into account uncertainties within microbial models calls for new validation techniques. Formal verification already overcomes this shortcoming by proposing modeling frameworks and validation techniques dedicated to probabilistic models. However, further work remains to extract the full potential of such techniques in the context of microbial models. Herein, we demonstrate how statistical model checking can enhance the development of microbial models by building confidence in the estimation of critical parameters and through improved sensitivity analyses.

59 BASIC BIOLOGICAL SCIENCES↗

An experimentalist’s view of the uncertainties in understanding heavy element synthesis

The overall uncertainties in predicting heavy element synthesis cross sections are examined in terms of the uncertainties associated with the calculations of capture cross sections, fusion probabilities and survival probabilities. Here, the values of the capture cross sections are known within a factor of two, the uncertainty in the values of the survival probabilities is no less than an order of magnitude with similar uncertainties in the fusion probabilities. Canceling errors may reduce the uncertainty in the overall predicted cross sections although for the heaviest nuclei (Z = 119 -120), the uncertainties in the predicted cross sections are a few orders of magnitude

208Pb↗

A cautionary tale of decorrelating theory uncertainties

A variety of techniques have been proposed to train machine learning classifiers that are independent of a given feature. While this can be an essential technique for enabling background estimation, it may also be useful for reducing uncertainties. We carefully examine theory uncertainties, which typically do not have a statistical origin. We will provide explicit examples of two-point (fragmentation modeling) and continuous (higher-order corrections) uncertainties where decorrelating significantly reduces the apparent uncertainty while the true uncertainty is much larger. These results suggest that caution should be taken when using decorrelation for these types of uncertainties as long as we do not have a complete decomposition into statistically meaningful components.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

PSL Uncertainty Calculator v. 1.2

The PSL Uncertainty Calculator can be used to determine the total uncertainty of a measurement system. In many measurement applications, the quantity of interest is not measured directly, but computed from a combination of several measured values. While the uncertainty of each measurand is typically straightforward to determine, total uncertainty of the output value requires calculation. This software calculates uncertainty using the framework provided by the Guide to Expression of Uncertainty in Measurements, JCGM100-2008E (GUM), and a Monte Carlo simulation as described by Supplement 1 to the GUM, by providing an easy to use graphical user interface and a Python package for running the calculations.

Delker, Collin↗

Patterns of Precipitation Change and Climatological Uncertainty among CMIP5 Models, with a Focus on the Midlatitude Pacific Storm Track

Projections of modeled precipitation ( P ) change in global warming scenarios demonstrate marked intermodel disagreement at regional scales. Empirical orthogonal functions (EOFs) and maximum covariance analysis (MCA) are used to diagnose spatial patterns of disagreement in the simulated climatology and end-of-century P changes in phase 5 of the Coupled Model Intercomparison Project (CMIP5) archive. The term principal uncertainty pattern (PUP) is used for any robust mode calculated when applying these techniques to a multimodel ensemble. For selected domains in the tropics, leading PUPs highlight features at the margins of convection zones and in the Pacific cold tongue. The midlatitude Pacific storm track is emphasized given its relevance to wintertime P projections over western North America. The first storm-track PUP identifies a sensitive region of disagreement in P increases over the eastern midlatitude Pacific where the storm track terminates, related to uncertainty in an eastward extension of the climatological jet. The second PUP portrays uncertainty in a zonally asymmetric meridional shift of storm-track P , related to uncertainty in the extent of a poleward jet shift in the western Pacific. Both of these modes appear to arise primarily from intermodel differences in the response to radiative forcing, distinct from sampling of internal variability. The leading storm-track PUPs for P and zonal wind change exhibit similarities to the leading uncertainty patterns for the historical climatology, indicating important and parallel sensitivities in the eastern Pacific storm-track terminus region. However, expansion coefficients for climatological uncertainties tend to be weakly correlated with those for end-of-century change.

54 ENVIRONMENTAL SCIENCES↗

Uncertainty Underprediction in Monte Carlo Eigenvalue Calculations

It is well-known that statistical estimates obtained from Monte Carlo criticality simulations can be adversely affected by cycle-to-cycle correlations in the fission source, which can lead to estimates of statistical uncertainties that are lower than the true uncertainty by a factor of 5 or more. However, several other more fundamental issues such as adequate source sampling over the fissionable regions and source convergence can have a significant impact on the uncertainties for the calculated eigenvalue and localized tally means, and these issues may be mistaken for effects resulting from cycle-to-cycle correlations. In worst-case scenarios, the uncertainty may be underpredicted by a factor of 40 or more. Since Monte Carlo methods are widely used in criticality safety applications and are increasingly being used for benchmarking reactor analyses, an in-depth understanding of the effects of these issues must be developed in order to support the practical use of Monte Carlo software packages. A rigorous statistical analysis of eigenvalue and localized tally results in Monte Carlo criticality calculations is presented using the SCALE/KENO-VI (continuous-energy version) and MCNP codes. The purpose of this analysis is to investigate the underprediction of uncertainty and its sensitivity to problem characteristics and calculational parameters using two of the most widely used Monte Carlo criticality codes. For the problems considered here, which are fuel rod and fuel assembly problems with reflecting boundary conditions on all four horizontal sides, we show that adequate source convergence along with proper specification of Monte Carlo parameters can reduce the magnitude of uncertainty underprediction to reasonable levels, below a factor of 2 in most cases.

Mervin, Brenden T↗

Uncertainty quantification in ToxCast high throughput screening

High throughput screening (HTS) projects like the U.S. Environmental Protection Agency’s ToxCast program are required to address the large and rapidly increasing number of chemicals for which we have little to no toxicity measurements. Concentration-response parameters such as potency and efficacy are extracted from HTS data using nonlinear regression, and models and analyses built from these parameters are used to predict in vivo and in vitro toxicity of thousands of chemicals. How these predictions are impacted by uncertainties that stem from parameter estimation and propagated through the models and analyses has not been well explored. While data size and complexity makes uncertainty quantification computationally expensive for HTS datasets, continued advancements in computational resources have allowed these computational challenges to be met. This study uses nonparametric bootstrap resampling to calculate uncertainties in concentration-response parameters from a variety of HTS assays. Using the ToxCast estrogen receptor model for bioactivity as a case study, we highlight how these uncertainties can be propagated through models to quantify the uncertainty in model outputs. Uncertainty quantification in model outputs is used to identify potential false positives and false negatives and to determine the distribution of model values around semi-arbitrary activity cutoffs, increasing confidence in model predictions. At the individual chemical-assay level, curves with high variability are flagged for manual inspection or retesting, focusing subject-matter-expert time on results that need further input. This work improves the confidence of predictions made using HTS data, increasing the ability to use this data in risk assessment.

59 BASIC BIOLOGICAL SCIENCES↗

SALSA3D: A Tomographic Model of Compressional Wave Slowness in the Earth’s Mantle for Improved Travel-Time Prediction and Travel-Time Prediction Uncertainty

The task of monitoring the Earth for nuclear explosions relies heavily on seismic data to detect, locate, and characterize suspected nuclear tests. In this study, motivated by the need to locate suspected explosions as accurately and precisely as possible, we developed a tomographic model of the compressional wave slowness in the Earth’s mantle with primary focus on the accuracy and precision of travel-time predictions for P and Pn ray paths through the model. Path-dependent travel-time prediction uncertainties are obtained by computing the full 3D model covariance matrix and then integrating slowness variance and covariance along ray paths from source to receiver. Path-dependent travel-time prediction uncertainties reflect the amount of seismic data that was used in tomography with very low values for paths represented by abundant data in the tomographic data set and very high values for paths through portions of the model that were poorly sampled by the tomography data set. The pattern of travel-time prediction uncertainty is a direct result of the off-diagonal terms of the model covariance matrix and underscores the importance of incorporating the full model covariance matrix in the determination of travel-time prediction uncertainty. In addition, the computed pattern of uncertainty differs significantly from that of 1D distance-dependent travel-time uncertainties computed using traditional methods, which are only appropriate for use with travel times computed through 1D velocity models.

58 GEOSCIENCES↗

UNCERTAINTIES OF ANION AND TOC MEASUREMENTS AT THE DWPF LABORATORY

The Savannah River Remediation (SRR) Defense Waste Processing Facility (DWPF) has identified a technical issue related to the amount of antifoam added to the Chemical Process Cell (CPC). Specifically, due to the long duration of the concentration and reflux cycles for the Sludge Receipt and Adjustment Tank (SRAT), additional antifoam has been required. The additional antifoam has been found to impact the melter flammability analysis as an additional source of carbon and hydrogen. To better understand and control the carbon and hydrogen contributors to the melter flammability analysis, SRR's Waste Solidification Engineering (WSE) has requested, via a Technical Task Request (TTR), that the Savannah River National Laboratory (SRNL) conduct an error evaluation of the measurements of key Slurry Mix Evaporator (SME) anions. SRNL issued a Task Technical and Quality Assurance Plan (TTQAP) [2] in response to that request, and the work reported here was conducted under the auspices of that TTQAP. The TTR instructs SRNL to conduct an error evaluation of anion measurements generated by the DWPF Laboratory using Ion Chromatography (IC) performed on SME samples. The anions of interest include nitrate, oxalate, and formate. Recent measurements of SME samples for these anions as well as measurements of total organic carbon (TOC) were provided to SRNL by DWPF Laboratory Operations (Lab OPS) personnel for this evaluation. This work was closely coordinated with the efforts of others within SRNL that are investigating the Chemical Process Cell (CPC) contributions to the melter flammability. The objective of that investigation was to develop a more comprehensive melter flammability control strategy that when implemented in DWPF will rely on process measurements. Accounting for the uncertainty of the measurements is necessary for successful implementation. The error evaluations conducted as part of this task will facilitate the integration of appropriate uncertainties for the measurements utilized in that control strategy. The flammability control strategy presented in relies on SME measurements of TOC and nitrate while one of the uses by WSE of the oxalate and formate measurement data will be the estimation of the amount of carbon coming from antifoam additions. The estimation is to be conducted by backing out contributions to the measured TOC concentration in the SME from the oxalate and the formate concentrations that are measured in the SME. The resulting adjusted TOC value will provide a basis for WSE to estimate the amount of antifoam that was added for that SME batch. The uncertainties of the oxalate, formate, and TOC measurements provided by the evaluations conducted as part of this task will allow for the propagation of their uncertainties into the estimated quantity of carbon coming from the added antifoam. The purpose of this technical report is to present the measurements generated by the DWPF Laboratory for recent SME batches, to conduct an evaluation of their uncertainties, and to provide the approach for propagating the uncertainties associated with these measurements into DWPF's strategies for controlling melter flammability and for monitoring antifoam additions. JMP Version 7.0.2 was used to support the analyses presented in this report.

08 HYDROGEN↗

Adjoint-Based Uncertainty Quantification with MCNP

This work serves to quantify the instantaneous uncertainties in neutron transport simulations born from nuclear data and statistical counting uncertainties. Perturbation and adjoint theories are used to derive implicit sensitivity expressions. These expressions are transformed into forms that are convenient for construction with MCNP6, creating the ability to perform adjoint-based uncertainty quantification with MCNP6. These new tools are exercised on the depleted-uranium hybrid LIFE blanket, quantifying its sensitivities and uncertainties to important figures of merit. Overall, these uncertainty estimates are small (< 2%). Having quantified the sensitivities and uncertainties, physical understanding of the system is gained and some confidence in the simulation is acquired.

42 ENGINEERING↗

Area 2: Inexpensive Monitoring and Uncertainty Assessment of CO 2 Plume Migration using Injection Data

In-depth understanding of the long-term fate of CO₂ in the subsurface requires study and analysis of the reservoir formation, the overlaying caprock formation, and adjacent faults. Because there is significant uncertainty in predicting the location and extent of geologic heterogeneity that can impact the future migration of CO₂ in the subsurface, there is a need to develop algorithms that can reliably quantify this uncertainty in plume migration. This project is focused on the development of a model selection algorithm that refines an initial suite of subsurface models representing the prior uncertainty to create a posterior set of subsurface models that reflect injection performance consistent with that observed. Such posterior models can be used to represent uncertainty in the future migration of the CO₂ plume. Because only injection data is required, the method provides a very inexpensive method to map the migration of the plume and the associated uncertainty in migration paths. The model selection method developed as part of this project mainly consists of assessing the connectivity/dynamic characteristics of a large prior ensemble of models, grouping the models on the basis of their expected dynamic response, selecting the subgroup of models that most closely yield dynamic response closest to the observed dynamic data, and finally quantifying the uncertainty in plume migration using the selected subset of models. The main accomplishment of the project is the development of a software module within the SGEMS earth modeling software package that implements the model selection methodology. This software module was subsequently applied to analyze CO₂ plume migration in two field projects – the In Salah CO₂ Injection project in Algeria and CO₂ injection into the Utsira formation in Norway. These applications of the software revealed that the proxies developed in this project for quickly assessing the dynamic characteristics of the reservoir were highly efficient and yielded accurate grouping of reservoir models. The plume migration paths probabilistically assessed by the method were confirmed by field observations and auxiliary data. The report also documents the application of the software to answer practical questions such as the optimum location of monitoring wells to reliably assess the migration of CO₂ plume, the effect of CO₂-rock interactions on plume migration and the ability to detect the plume under those conditions and the effect of a slow, unresolved leak on the predictions of plume migration.

54 ENVIRONMENTAL SCIENCES↗

An Uncertainty Analysis of the Hydrogen Source Term for a Station Blackout Accident in Sequoyah Using MELCOR 1.8.5

A methodology for using the MELCOR code with the Latin Hypercube Sampling method was developed to estimate uncertainty in various predicted quantities such as hydrogen generation or release of fission products under severe accident conditions. In this case, the emphasis was on estimating the range of hydrogen sources in station blackout conditions in the Sequoyah Ice Condenser plant, taking into account uncertainties in the modeled physics known to affect hydrogen generation. The method uses user-specified likelihood distributions for uncertain model parameters, which may include uncertainties of a stochastic nature, to produce a collection of code calculations, or realizations, characterizing the range of possible outcomes. Forty MELCOR code realizations of Sequoyah were conducted that included 10 uncertain parameters, producing a range of in-vessel hydrogen quantities. The range of total hydrogen produced was approximately 583kg ± 131kg. Sensitivity analyses revealed expected trends with respected to the parameters of greatest importance, however, considerable scatter in results when plotted against any of the uncertain parameters was observed, with no parameter manifesting dominant effects on hydrogen generation. It is concluded that, with respect to the physics parameters investigated, in order to further reduce predicted hydrogen uncertainty, it would be necessary to reduce all physics parameter uncertainties similarly, bearing in mind that some parameters are inherently uncertain within a range. It is suspected that some residual uncertainty associated with modeling complex, coupled and synergistic phenomena, is an inherent aspect of complex systems and cannot be reduced to point value estimates. The probabilistic analyses such as the one demonstrated in this work are important to properly characterize response of complex systems such as severe accident progression in nuclear power plants.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Extreme-Scale Bayesian Inference for Uncertainty Quantification of Complex Simulations

Uncertainty quantification (UQ)—that is, quantifying uncertainties in complex mathematical models and their large-scale computational implementations—is widely viewed as one of the outstanding challenges facing the field of CS&E over the coming decade. The EUREKA project set to address the most difficult class of UQ problems: those for which both the underlying PDE model as well as the uncertain parameters are of extreme scale. In the project we worked on these extreme-scale challenges in the following four areas: 1. Scalable parallel algorithms for sampling and characterizing the posterior distribution that exploit the structure of the underlying PDEs and parameter-to-observable map. These include structure-exploiting versions of the randomized maximum likelihood method, which aims to overcome the intractability of employing conventional MCMC methods for solving extreme-scale Bayesian inversion problems by appealing to and adapting ideas from large-scale PDE-constrained optimization, which have been very successful at exploring high-dimensional spaces. 2. Scalable parallel algorithms for construction of prior and likelihood functions based on learning methods and non-parametric density estimation. Constructing problem-specific priors remains a critical challenge in Bayesian inference, and more so in high dimensions. Another challenge is construction of likelihood functions that capture unmodeled couplings between observations and parameters. We will create parallel algorithms for non-parametric density estimation using high dimensional N-body methods and combine them with supervised learning techniques for the construction of priors and likelihood functions. 3. Bayesian inadequacy models, which augment physics models with stochastic models that represent their imperfections. The success of the Bayesian inference framework depends on the ability to represent the uncertainty due to imperfections of the mathematical model of the phenomena of interest. This is a central challenge in UQ, especially for large-scale models. We propose to develop the mathematical tools to address these challenges in the context of extreme-scale problems. 4. Parallel scalable algorithms for Bayesian optimal experimental design (OED). Bayesian inversion yields quantified uncertainties in the model parameters, which can be propagated forward through the model to yield uncertainty in outputs of interest. This opens the way for designing new experiments to reduce the uncertainties in the model parameters and model predictions. Such experimental design problems have been intractable for large-scale problems using conventional methods; we will create OED algorithms that exploit the structure of the PDE model and the parameter-to-output map to overcome these challenges. Parallel algorithms for these four problems were created, analyzed, prototyped, implemented, tuned, and scaled up for leading-edge supercomputers, including UT-Austin’s own 10 petaflops Stampede system, ANL’s Mira system, and ORNL’s Titan system. While our focus is on fundamental mathematical/computational methods and algorithms, we will assess our methods on model problems derived from several DOE mission applications, including multiscale mechanics and ice sheet dynamics.

97 MATHEMATICS AND COMPUTING↗