Search NASA⌕ Search

SEARCH · Search NASA

Results for “Robust optimization”

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 1,045 records · Page 58

Automatically Finding the Control Variables for Complex System Behavior

Testing large-scale systems is expensive in terms of both time and money. Running simulations early in the process is a proven method of finding the design faults likely to lead to critical system failures, but determining the exact cause of those errors is still time-consuming and requires access to a limited number of domain experts. It is desirable to find an automated method that explores the large number of combinations and is able to isolate likely fault points. Treatment learning is a subset of minimal contrast-set learning that, rather than classifying data into distinct categories, focuses on finding the unique factors that lead to a particular classification. That is, they find the smallest change to the data that causes the largest change in the class distribution. These treatments, when imposed, are able to identify the factors most likely to cause a mission-critical failure. The goal of this research is to comparatively assess treatment learning against state-of-the-art numerical optimization techniques. To achieve this, this paper benchmarks the TAR3 and TAR4.1 treatment learners against optimization techniques across three complex systems, including two projects from the Robust Software Engineering (RSE) group within the National Aeronautics and Space Administration (NASA) Ames Research Center. The results clearly show that treatment learning is both faster and more accurate than traditional optimization methods.

Gay, Gregory↗

Progress in multirate digital control system design

A new methodology for multirate sampled-data control design based on a new generalized control law structure, two new parameter-optimization-based control law synthesis methods, and a new singular-value-based robustness analysis method are described. The control law structure can represent multirate sampled-data control laws of arbitrary structure and dynamic order, with arbitrarily prescribed sampling rates for all sensors and update rates for all processor states and actuators. The two control law synthesis methods employ numerical optimization to determine values for the control law parameters. The robustness analysis method is based on the multivariable Nyquist criterion applied to the loop transfer function for the sampling period equal to the period of repetition of the system's complete sampling/update schedule. The complete methodology is demonstrated by application to the design of a combination yaw damper and modal suppression system for a commercial aircraft.

Berg, Martin C.↗

SU-F-BRD-06: Robust Dose Calculation in Intensity Modulated Proton Therapy

Purpose: Commissioning data for intensity modulated proton therapy (IMPT) must be post-processed by fits to ad-hoc functions to derive the dose calculation kernel parameters in a treatment planning system (TPS). Whether from experimental measurement or Monte Carlo simulation, the limited and noisy nature of such data makes this task very challenging. We present a method to improve the modeling of the lateral dose distribution of clinical energy proton beams in water to commission an in-house IMPT dose calculation engine. Methods: A linear sum of three Gaussian distribution functions was fitted to the lateral dose data in logarithmic scale. Starting values of fitting solutions were determined from the Generalized Highland Approximation. We exhaustively optimized the combinations of data weights with upper bounds of the fitting solutions to minimize confidence intervals of the fitting solutions while maintaining the coefficient of determination (R{sup 2}). Results: Across all energies, average confidence bounds improved 72.88% [Max: 88.28%, Min: 55.05%] for small angle coulomb scattering, 114.25% [409.13%, 66.72%,] for nuclear scattering, and 68.66% [141.09%, 33.27%] for large angle coulomb scattering, while the coefficients of determination of the fits (R{sup 2}) remained comparable. On average R {sup 2} only changed 0.18% and were very close to 1 (approx. 0.999). Wilcoxon signed rank tests comparing unweighted/unbounded fits with weighted/bounded fits averaged 0.0146 (Max: 0.177, Min: 7.05×10−{sup 7}) for small angle Coulomb, 0.0903 (0.945, 7.05×10−{sup 7}) for nuclear, and 0.254 (0.871, 1.86×10−{sup 6}) for large angle Coulomb scattering. This allows rejection of the null hypothesis for small angle Coulomb scattering at the 0.015 level and nuclear interaction at the 0.1 level. Conclusion: Optimal weights assigned to IMPT lateral dose data minimized fitting to stochastic noise in the tail region. Optimizing the upper bounds of fitting parameters improved the robustness of fitting solutions. Both led to a more physically meaningful dose calculation kernel for the TPS. NIH/NCI K25CA168984, Eagles Cancer Research Career Development, The Lawrence W. and Marilyn W. Matteson Fund for Cancer Research, Mayo ASU Seed Grant, and The Kemper Marley Foundation.

60 APPLIED LIFE SCIENCES↗

Optimization of carbon capture systems using surrogate models of simulated processes.

With increasing demand placed on power generation plants to reduce carbon dioxide (CO2) emissions, processes to separate and capture CO2 for eventual sequestration are highly sought after. Carbon capture processes impart a parasitic load on the power plants; it is estimated that this would increase the cost of electricity from existing pulverized coal plants anywhere from 71-85 percent [1]. The National Energy and Technology Lab (NETL) is working to lower this to below a 30 percent increase. To reach this goal, work is being done not only to accurately simulate these processes, but also to leverage those accurate and detailed simulations to design optimal carbon capture processes. The major challenges include the lack of accurate algebraic models of the processes, computationally costly simulations, and insufficiently robust simulations. The first challenge bars the use of provable derivative-based optimization algorithms. The latter two can either lead to difficult or impossible direct derivative-free optimization. To overcome these difficulties, we take a more indirect method to solving this problem by, first, generating an accurate set of algebraic surrogate models from the simulation then using derivative-based solvers to optimize the surrogate models. We developed a method that uses derivative-based and derivative-free optimization alongside machine learning and statistical techniques to generate the set of low-complexity surrogate models using data sampled from detailed simulations. The models are validated and improved through the use of derivative-free solvers to adaptively sample new simulation points. The resulting surrogate models can then be used in a superstructure-based process synthesis and solved using derivative-based methods to optimize carbon capture processes.

01 COAL, LIGNITE, AND PEAT↗

Data-driven optimal charging decision making for connected and automated electric vehicles: A personal usage scenario

This study introduces an optimal charging decision making framework for connected and automated electric vehicles under a personal usage scenario. This framework aims to provide charging strategies, i.e. the choice of charging station and the amount of charged energy, by considering constraints from personal daily itineraries and existing charging infrastructure. A data-driven method is introduced to establish a stochastic energy consumption prediction model with consideration of realistic uncertainties. This is performed by analyzing a large scale electric vehicle data set. A real-time updating method is designed to construct this prediction model from new consecutive data points in an adaptive way for real-world applications. Based on this energy cost prediction framework from real electric vehicle data, multistage optimal charging decision making models are introduced, including a deterministic model for average outcome decision making and a robust model for safest charging strategies. A dynamic programming algorithm is proposed to find the optimal charging strategies. Detailed simulations and case studies demonstrate the performance of the proposed algorithms to find optimal charging strategies. They also show the potential capability of connected and automated electric vehicles to reduce the range anxiety and charging infrastructure dependency.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

Asynchronous parallel hybrid optimization combining DIRECT and GSS.

In this talk, we explore the benefits of hybrid optimization using parallel versions of DIRECT and asynchronous generating set search (GSS) for optimization. Both of these methods are derivative-free, making them useful for a variety of science and engineering problems. Our goal is to ideally find a global minimum, but practically to find a good local minimum in a small amount of time. DIRECT is a global search method that systematically divides the search space into ever-smaller rectangles, and GSS is a local search method. The combination of these method guarantees a good local minimum but is better than a purely local approach because it finds more global solutions. We compare the performance of hybrid and non-hybrid methods on a suite of standard global optimization test problems. Overall, the hybrid methods are more robust than the non-hybrid methods at the cost of more function evaluations. In terms of wall-clock time on a parallel system, the hybrid methods are actually less expensive due to nearly perfect parallel load balance and scaling.

97 MATHEMATICS AND COMPUTING↗

Optimal Cosmological Measurements with Weak Gravitational Lensing

The goal of the research carried out under this DOE award was to develop methodology to ensure that the upcoming Large Synoptic Survey Telescope (LSST) survey achieves its goals of robustly constraining dark energy using weak gravitational lensing. This included developing more optimal ways to use the data on its own or in combination with other surveys, better ways of detecting and reducing the impact of systematic errors, and demonstrating that methodology on existing data.

79 ASTRONOMY AND ASTROPHYSICS↗

Advocating Feedback Control for Human-Earth System Applications

This paper proposes a feedback control perspective for Human-Earth Systems (HESs) which essentially are complex systems that capture the interactions between humans and nature. Recent attention in HES research has been directed towards devising strategies for climate change mitigation and adaptation, aimed at achieving environmental and societal objectives. However, existing approaches heavily rely on HES models, which inherently suffer from inaccuracies due to the complexity of the system. Moreover, overly detailed models often prove impractical for optimization tasks. We propose a framework inheriting from feedback control strategies the robustness against model errors, because inaccuracies are mitigated using measurements retrieved from the field. The framework comprises two nested control loops. The outer loop computes the optimal inputs to the HES, which are then implemented by actuators controlled in the inner loop. Potential fields of applications are also identified and a numerical example is provided.

biological system modeling↗

SU-F-T-188: A Robust Treatment Planning Technique for Proton Pencil Beam Scanning Cranial Spinal Irradiation

Purpose: To propose a proton pencil beam scanning (PBS) cranial spinal irradiation (CSI) treatment planning technique robust against patient roll, isocenter offset and proton range uncertainty. Method: Proton PBS plans were created (Eclipse V11) for three previously treated CSI patients to 36 Gy (1.8 Gy/fractions). The target volume was separated into three regions: brain, upper spine and lower spine. One posterior-anterior (PA) beam was used for each spine region, and two posterior-oblique beams (15° apart from PA direction, denoted as 2PO-15) for the brain region. For comparison, another plan using one PA beam for the brain target (denoted as 1PA) was created. Using the same optimization objectives, 98% CTV was optimized to receive the prescription dose. To evaluate plan robustness against patient roll, the gantry angle was increased by 3° and dose was recalculated without changing the proton spot weights. On the re-calculated plan, doses were then calculated using 12 scenarios that are combinations of isocenter shift (±3mm in X, Y, and Z directions) and proton range variation (±3.5%). The worst-case-scenario (WCS) brain CTV dosimetric metrics were compared to the nominal plan. Results: For both beam arrangements, the brain field(s) and upper-spine field overlap in the T2–T5 region depending on patient anatomy. The maximum monitor unit per spot were 48.7%, 47.2%, and 40.0% higher for 1PA plans than 2PO-15 plans for the three patients. The 2PO-15 plans have better dose conformity. At the same level of CTV coverage, the 2PO-15 plans have lower maximum dose and higher minimum dose to the CTV. The 2PO-15 plans also showed lower WCS maximum dose to CTV, while the WCS minimum dose to CTV were comparable between the two techniques. Conclusion: Our method of using two posterior-oblique beams for brain target provides improved dose conformity and homogeneity, and plan robustness including patient roll.

60 APPLIED LIFE SCIENCES↗

Robust neighboring extremal guidance for the advanced launch system

With the availability of modern flight computers, realtime neighboring extremal guidance seems feasible. To overcome sensitivity to unknown system parameters and environmental uncertainties, a robust neighboring extremal guidance scheme is proposed. About the optimal trajectory, the accessory problem in the calculus of variations is formed, generating a quadratic cost criterion in the perturbed states and controls. By formulating a disturbance attenuation problem based upon the second variation cost criterion, a differential game is formulated. The game theoretic cost criterion is minimized with respect to the perturbed control but maximized with respect to the unknown parameters in the linearized dynamics. The resulting differential game problem gives rise to a two-point boundary-value problem solved using the sweep method. The sweep method solution provides a linear robust neighboring extremal guidance scheme that is applied to the Advanced Launch System.

Bain, John↗

Robust controllers for the Middeck Active Control Experiment using Popov controller synthesis

Recent work in robust control with real parameter uncertainties has focused on absolute stability and its connections to real mu theory. In particular, the research has investigated the Popov stability criterion and its associated Lur'e-Postnikov Liapunov functions. State space representations of this Popov stability analysis tests are included in an H2 design formulation to provide a powerful technique for robust controller synthesis. This synthesis approach uses a state space optimization procedure to design controllers that minimize an overbound of an H2 cost functional and satisfy stability analysis tests based on the Popov multiplier. The controller and stability multiplier coefficients are optimized simultaneously, which avoids the iteration and curve-fitting procedures required by the D-K algorithm of mu synthesis. While previous work has demonstrated this synthesis approach on benchmark control problems, the purpose of this paper is to use Popov controller synthesis to design robust compensators for the Middeck Active Control Experiment (MACE).

How, Jonathan P.↗

Gust Load Alleviation Control and Gust Estimation for a High Aspect Ratio Wing Wind Tunnel Model

This paper presents the gust load alleviation (GLA) study of the aspect ratio 13.5 Common Research Model (CRM) wind-tunnel model. This study details the design of the GLA controller in preparation for wind-tunnel testing in the Transonic Dynamics Tunnel at NASA Langley Research Center. An aeroservolastic (ASE) model was first reduced using a model reduction method that takes advantage of the sinusoidal steady-state response. Then, the reduced model was used to design an extended-state Kalman filter which estimates the states and the sinusoidal gust input. The GLA control was then derived using the optimal control solution to a multi-objective cost function. The results of the GLA controller indicate a 75.6% reduction in wing-root strain while maintaining robust stability margins. The final paper will include performance under sub-optimal conditions, e.g., sensor noise, uncertainty in gust frequency, etc..

Christopher Forte↗

Liquid Crystal Orientation and Shape Optimization for the Active Response of Liquid Crystal Elastomers

Liquid crystal elastomers (LCEs) are responsive materials that can undergo large reversible deformations upon exposure to external stimuli, such as electrical and thermal fields. Controlling the alignment of their liquid crystals mesogens to achieve desired shape changes unlocks a new design paradigm that is unavailable when using traditional materials. While experimental measurements can provide valuable insights into their behavior, computational analysis is essential to exploit their full potential. Accurate simulation is not, however, the end goal; rather, it is the means to achieve their optimal design. Such design optimization problems are best solved with algorithms that require gradients, i.e., sensitivities, of the cost and constraint functions with respect to the design parameters, to efficiently traverse the design space. In this work, a nonlinear LCE model and adjoint sensitivity analysis are implemented in a scalable and flexible finite element-based open source framework and integrated into a gradient-based design optimization tool. To display the versatility of the computational framework, LCE design problems that optimize both the material, i.e., liquid crystal orientation, and structural shape to reach a target actuated shapes or maximize energy absorption are solved. Multiple parameterizations, customized to address fabrication limitations, are investigated in both 2D and 3D. The case studies are followed by a discussion on the simulation and design optimization hurdles, as well as potential avenues for improving the robustness of similar computational frameworks for applications of interest.

42 ENGINEERING↗

TH-AB-202-03: A Novel Tool for Computing Deliverable Doses in Dynamic MLC Tracking Treatments

Purpose: In tracked dynamic multi-leaf collimator (MLC) treatments, segments are continuously adapted to the target centroid motion in beams-eye-view. On-the-fly segment adaptation, however, potentially induces dosimetric errors due to the finite MLC leaf width and non-rigid target motion. In this study, we outline a novel tool for computing the 4d dose of lung SBRT plans delivered with MLC tracking. Methods: The following automated workflow was developed: A) centroid tracking, where the initial segments are morphed to each 4dCT phase based on the beams-eye-view GTV shift (followed by a dose calculation on each phase); B) re-optimized tracking, in which all morphed initial plans from (A) are further optimised (“warm-started”) in each 4dCT phase using the initial optimisation parameters but phase-specific volume definitions. Finally, both dose sets are accumulated to the reference phase using deformable image registration. Initial plans were generated according to the RTOG-1021 guideline (54Gy, 3-Fx, equidistant 9-beam IMRT) on the peak-exhale (reference) phase of a phase-binned 4dCT. Treatment planning and delivery simulations were performed in RayStation (research v4.6) using our in-house segment-morphing algorithm, which directly links to RayStation through a native C++ interface. Results: Computing the tracking plans and 4d dose distributions via the in-house interface takes 5 and 8 minutes respectively for centroid and re-optimized tracking. For a sample lung SBRT patient with 14mm peak-to-peak motion in sup-inf direction, mainly perpendicular leaf motion (0-collimator) resulted in small dose changes for PTV-D95 (−13cGy) and GTV-D98 (+18cGy) for the centroid tracking case compared to the initial plan. Modest reductions of OAR doses (e.g. spinal cord D2: −11cGy) were achieved in the idealized tracking case. Conclusion: This study presents an automated “1-click” workflow for computing deliverable MLC tracking doses in RayStation. Adding a non-deliverable re-optimized tracking scenario is expected to help quantify plan robustness for more challenging patients with anatomy deformations. We acknowledge support of the MLC tracking research from Elekta AB. MFF is supported by Cancer Research UK under Programme C33589/A19908. Research at ICR is also supported by Cancer Research UK under Programme C33589/A19727 and NHS funding to the NIHR Biomedical Research Centre at RMH and ICR.

60 APPLIED LIFE SCIENCES↗

An optimal estimation approach for enhanced identification of large space structures

A very robust technique for the realization/identification of structural dynamic systems is presented. This methodology combines the eigensystem realization algorithm (ERA) and minimum model error (MME) techniques to develop an extremely robust algorithm able to distinguish modes whose amplitudes are small fractions of the noise amplitude. As the amplitude of the noise grows, the enhanced robustness of the combined ERA/MME algorithm becomes more significant. This is a result of the MME being able to produce simulated measurements of much greater accuracy than the original measurements, for input to the ERA. The new method was able to determine correct model order at noise levels many times higher than the original ERA for the realization problem in which the model order must be derived.

Mook, D. Joseph↗

Predictive Capability for Strongly Correlated Systems

Diffusion Monte Carlo methods can give highly accurate results for correlated systems, provided that well optimized trial wave functions with accurate nodal surfaces are employed. The Cornell team developed powerful methods for optimizing all the parameters within a multi-determinant Slater-Jastrow form of the wave function. These include the Jastrow parameters within a flexible electron-electron-nucleus form of the Jastrow function, the parameters multiplying the configuration state functions, the orbital parameters and the basis exponents. The method optimizes a linear combination of the energy and the variance of the local energy. The optimal parameters are found iteratively by diagonalizing the Hamiltonian matrix in the space spanned by the wave function and its first-order derivatives, making use of a strong zero-variance principle. It is highly robust, has become the method of choice for correlated wave function optimization and has been adopted by other QMC groups. This optimization method was used on the first-row atoms and homonuclear diatomic molecules, demonstrating that molecular well depths can be obtained with near chemical accuracy quite systematically at the diffusion Monte Carlo level for these systems. In addition the complete ground-state potential energy curve of the C 2 molecule up to the dissociation limit was obtained, and, size consistency and broken spin-symmetry issues in quantum Monte Carlo calculations were studied. The method was used with a eight-electrons-in-eight-orbitals complete active space CAS(8,8) wave function to study the relative energies of the monocyclic and bicyclic forms of m-benzyne. The DMC calculations show that the monocyclic structure is lower in energy than the bicyclic structure by 1.92 kcal/ mole, which is in excellent agreement with the best coupled cluster results (CCSD(T)) and in disagreement with the CCSD results. QMC methods have for the most part been used only for ground states of a given symmetry. We developed a method for calculating low-lying excited states as well and tested it on the ground and lowest three adiabatic excited states of methylene with progressively larger JastrowSlater multideterminant complete active space CAS wave functions. The highest of these states has the same symmetry, 1 A 1 , as the first excited state. The DMC excitation energies obtained using any of the CAS wave functions are in excellent agreement with experiment. In contrast, single-determinant wave functions do not yield accurate DMC energies, indicating that it is important to include in the wave function Slater determinants that describe static strong correlation.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Co-optimizing water-alternating-carbon dioxide injection projects using a machine learning assisted computational framework

In this report, a robust machine-learning-based computational framework that couples multi-layer neural network (MLNN) proxies and a multi-objective particle swarm optimizer (MOPSO) to design water-alternating-carbon dioxide injection (CO 2 -WAG) projects is presented. The proposed optimization protocol considers various objectives, including oil recovery and CO 2 storage volume. Expert MLNN systems are trained and employed as surrogate models of the high-fidelity compositional simulator in the optimization workflow. When multiple objective functions are considered, two approaches are employed to treat the objectives: the weighted sum method and the Pareto-front-based scheme. A field-scale implementation focusing on tertiary recovery in the Morrow B formation at Farnsworth Unit (FWU) is presented. The developed Pareto-optimal solutions indicate the maximal available oil production can be 1.64 × 10 7 barrels and maximal carbon storage can achieve 2.35 × 10 7 tons. Trade-offs factor is defined to divide the constructed Pareto front into 4 sections with the trade-off factors’ value ranges from 0.35 to 49.9. This work also compares the optimum solution found by the aggregative objective function and the solution repository covered by the Pareto front that considers the physical and operational constraints and reduces uncertainties involved by the multi-objective optimization process. Our comparison indicates multiple solutions exist to satisfy the objective criteria of the WAG design, and these results cannot be found using the traditional weighted sum method. The Pareto front solution can provide more options for project designers, but decisions regarding necessary trade-offs must be made using the solution repository to balance the project economics and CO 2 storage amount.

10 SYNTHETIC FUELS↗

Projective Integral Updates for High-Dimensional Variational Inference

Variational inference is an approximation framework for Bayesian inference that seeks to improve quantified uncertainty in predictions by optimizing a simplified distribution over parameters to stand in for the full posterior. Capturing model variations that remain consistent with training data enables more robust predictions by reducing parameter sensitivity. This work introduces a fixed-point optimization for variational inference that is applicable when every feasible log density can be expressed as a linear combination of functions from a given basis. In such cases, the optimizer becomes a fixed-point of projective integral updates. When the basis spans univariate quadratics in each parameter, the feasible distributions are Gaussian mean-fields and the projective integral updates yield quasi-Newton variational Bayes (QNVB). Other bases and updates are also possible. Since these updates require high-dimensional integration, this work begins by proposing an efficient quasirandom sequence of quadratures for mean-field distributions. Each iterate of the sequence contains two evaluation points that combine to correctly integrate all univariate quadratic functions and, if the mean-field factors are symmetric, all univariate cubics. More importantly, averaging results over short subsequences achieves periodic exactness on a much larger space of multivariate polynomials of quadratic total degree. The corresponding variational updates require four loss evaluations with standard (not second-order) backpropagation to eliminate error terms from over half of all multivariate quadratic basis functions. Furthermore, this integration technique is motivated by first proposing stochastic blocked mean-field quadratures, which may be useful in other contexts. A PyTorch implementation of QNVB allows for better control over model uncertainty during training than competing methods. Experiments demonstrate superior generalizability for multiple learning problems and architectures.

Gaussian mean-field↗