Inverse Design of Materials with Lab Automation and AI-driven Experimentation
Explore the source record for details and available documents.
automated lab↗
SEARCH · Search NASA
Results for “inverse 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.
Denoising diffusion probabilistic models for generative alloy design
Inverse material design is an extremely challenging optimization task made difficult by, in part, the highly nonlinear relationship linking performance with composition. Quantitative approaches have improved significantly owing to advances in high throughput experimentation and computational thermodynamics. However, existing physics-based tools are mostly forward models; input a chemistry and obtain a prediction. More recently the materials community has leveraged advances in the machine learning community to establish novel inverse design frameworks. Very recently denoising diffusion probabilistic models have been shown to be extremely powerful generators producing synthetic data of various modalities e.g. images, text, audio, tables, etc.. In this work a novel framework for alloy design and optimization is proposed leveraging these class of models. Five key generative tasks are demonstrated (1) unconditional generation (2) composition conditioned generation (3) property conditioned generation (4) multi-feedstock conditioned generation and (5) generative optimization. These methods were tested on three case studies: high entropy alloy design, superalloy binder jet additive manufacturing, and in-situ dual-feedstock wire-arc additive manufacturing. Results indicate that the established models are extremely flexible, expressive, and robust. The architecture’s flexibility and training procedure empower the model to learn complex intra-compositional and composition-property relationships. Furthermore, the probabilistic nature of these models makes them well suited for addressing solution non-uniqueness and tackling uncertainty quantification tasks. While the fidelity and quantity of the underlying training data is paramount, we envision that future alloy design frameworks will make extensive use of these kinds of machine learning models as “search” tools bolstering the utility of experimental and computational approaches.
36 MATERIALS SCIENCE↗
Optimization and mechanisms of mistuning in cascades
In the present inverse design procedure for the optimum mistuning of a high bypass ratio shroudless fan that is modeled as a cascade of blades (each with a single torsional degree-of-freedom), linearized supersonic aerodynamic theory is used to compute the unsteady aerodynamic forces in the 'influence coefficient' form at a typical blade section. The mistuning pattern is then numerically optimized in order to achieve a specified increase in the aeroelastic stability margin with a minimum amount of mistuning. If the blades are self-damped, an optimized mistuning pattern can be found that achieves a given stability margin for a much lower level of mistuning than required for the alternate mistuning pattern, which requires only two blade frequencies and is relatively insensitive to implementation errors.
Crawley, E. F.↗
P- and L-Band Retrieval of Subsurface Soil Moisture and Temperature Profiles as First-Order Polynomial Function
This paper demonstrates the potential use of P and L band passive measurements to determine root zone soil moisture (SM) and soil temperature(ST). SM and ST data have been taken as a function of depth during the NASA GSFC PLEX19 experiment in the summer of 2019 at Beltsville, MD, USA. Using these data, a coherent model has been used to compute H and V brightness temperatures at frequencies of 0.8 and 1.4 GHz with an observation angle of 35 degrees. These synthetic brightness data are then used to estimate the SM and ST profiles which are represented by linear polynomials. The inversion problem is formulated as a least square problem that is solved by a global optimization method known as the Adaptive Simulated Annealing(ASA) method. Four inversion examples having different SM and ST profiles are presented. Selected results show that the standard deviation between the retrieved and measured data is less than 0.077 cm3/cm3 for SM, and 2.245 °C for ST.
Ming Li↗
Optimal mistuning for enhanced aeroelastic stability of transonic fans
An inverse design procedure was developed for the design of a mistuned rotor. The design requirements are that the stability margin of the eigenvalues of the aeroelastic system be greater than or equal to some minimum stability margin, and that the mass added to each blade be positive. The objective was to achieve these requirements with a minimal amount of mistuning. Hence, the problem was posed as a constrained optimization problem. The constrained minimization problem was solved by the technique of mathematical programming via augmented Lagrangians. The unconstrained minimization phase of this technique was solved by the variable metric method. The bladed disk was modelled as being composed of a rigid disk mounted on a rigid shaft. Each of the blades were modelled with a single tosional degree of freedom.
Hall, K. C.↗
Optimization as a Tool for Consistency Maintenance in Multi-Resolution Simulation
The need for new approaches to the consistent simulation of related phenomena at multiple levels of resolution is great. While many fields of application would benefit from a complete and approachable solution to this problem, such solutions have proven extremely difficult. We present a multi-resolution simulation methodology that uses numerical optimization as a tool for maintaining external consistency between models of the same phenomena operating at different levels of temporal and/or spatial resolution. Our approach follows from previous work in the disparate fields of inverse modeling and spacetime constraint-based animation. As a case study, our methodology is applied to two environmental models of forest canopy processes that make overlapping predictions under unique sets of operating assumptions, and which execute at different temporal resolutions. Experimental results are presented and future directions are addressed.
TEMPORAL RESOLUTION↗
Polynomial compensation, inversion, and approximation of discrete time linear systems
The least-squares transformation of a discrete-time multivariable linear system into a desired one by convolving the first with a polynomial system yields optimal polynomial solutions to the problems of system compensation, inversion, and approximation. The polynomial coefficients are obtained from the solution to a so-called normal linear matrix equation, whose coefficients are shown to be the weighting patterns of certain linear systems. These, in turn, can be used in the recursive solution of the normal equation.
Baram, Yoram↗
Modified Cascading Generalized Inverse Control Allocation
The current aviation revolution towards electric propulsion aircraft (e.g., electric vertical takeoff-and-landing) brings unique control challenges. These vehicles are typically over-actuated (more effectors than desired control outcomes), may require control strategies for the three phases of flight (hover, transition and cruise), and currently have limited electric power availability. These vehicle challenges bring the need for optimal control allocation to the forefront of research. A leading control allocation algorithm, used in current flight vehicles, is the Cascading Generalized Inverse (CGI). Unfortunately, the Cascading Generalized Inverse algorithm is unable to achieve some desired outcomes, it intermittently provides non-optimal allocations, and it may fail to preserve moment direction near maximal achievable outcomes. In this research, the shortcomings of the Cascading Generalized Inverse algorithm are addressed by augmenting the algorithm with Scalar Difference Quadratic unsaturation identification and location at each iteration. Rigorous theory is shown that the Modified Cascading Generalized Inverse performs better at obtaining optimal allocations for all attainable outcomes. Numerical case studies for over-actuated vehicles demonstrate resolution to the aforementioned deficiencies.
Control Allocation↗
Modified Cascading Generalized Inverse Control Allocation
The current aviation revolution towards electric propulsion aircraft (e.g., electric vertical takeoff-and-landing) brings unique control challenges. These vehicles are typically over-actuated (more effectors than desired control outcomes), may require control strategies for the three phases of flight (hover, transition and cruise), and currently have limited electric power availability. These vehicle challenges bring the need for optimal control allocation to the forefront of research. A leading control allocation algorithm, used in current flight vehicles, is the Cascading Generalized Inverse (CGI). Unfortunately, the Cascading Generalized Inverse algorithm is unable to achieve some desired outcomes, it intermittently provides non-optimal allocations, and it may fail to preserve moment direction near maximal achievable outcomes. In this research, the shortcomings of the Cascading Generalized Inverse algorithm are addressed by augmenting the algorithm with Scalar Difference Quadratic unsaturation identification and location at each iteration. Rigorous theory is shown that the Modified Cascading Generalized Inverse performs better at obtaining optimal allocations for all attainable outcomes. Numerical case studies for over-actuated vehicles demonstrate resolution to the aforementioned deficiencies.
Control Allocation↗
Exergy-based Sensitivity Analysis of the Generic Hypersonic Vehicle using FUN3D
In this paper, the implementation of an exergy-based objective function and its adjoint gradient into NASA’s FUN3D solver is discussed and verified. In order to verify that the exergy-based functional is properly implemented, it is used to predict the drag of the Generic Hypersonic Vehicle (GHV), which is then compared to more traditional force-based drag predictions. In addition to the functional implementation, FUN3D’s adjoint capability was extended to obtain sensitivities. Results were verified using FUN3D’s native complex step method for di↵erentiation using a generic wing configuration. The complex and adjoint gradients yielded discrete agreement demonstrating correct implementation and that the functional can be used for gradient-based multidisciplinary analysis and optimization. Next, various trade studies are conducted on the GHV to understand the design space of the vehicle. Finally, an inverse design problem is solved to verify the utilized design optimization framework which is ready to be deployed for exergy-based optimizations in future work.
Neal L. Novotny↗
AutoTandemML: Active Learning Enhanced Tandem Neural Networks for Inverse Design Problems
Inverse design in science and engineering involves determining optimal design parameters that achieve desired performance outcomes, a process often hindered by the complexity and high dimensionality of design spaces, leading to significant computational costs. To tackle this challenge, we propose a novel hybrid approach that combines active learning with Tandem Neural Networks to enhance the efficiency and effectiveness of solving inverse design problems. Active learning allows to selectively sample the most informative data points, reducing the required dataset size without compromising accuracy. We investigate this approach using three benchmark problems: airfoil inverse design, photonic surface inverse design, and scalar boundary condition reconstruction in diffusion partial differential equations. We demonstrate that integrating active learning with Tandem Neural Networks outperforms standard approaches across the benchmark suite, achieving better accuracy with fewer training samples.
97 MATHEMATICS AND COMPUTING↗
Engine/airframe integration for waverider cruise vehicles
Engine-airframe integrated hypersonic waveriders have been optimized previously for both cruising flight and accelerators using inverse design flowfields to provide desirable aerodynamic performance and inlet properties. With engine installation, cruisers retain the high L/D advantages of the basic waverider form. To provide an understanding of waverider optimizer trends, an analytical development is presented which includes the presentation of a simplified model of a generic waverider shape. The model is shown to have good agreement with computational results, and it is used to demonstrate engineering tradeoffs in L/D, volume, and volume efficiency. Based on the demonstrated performance trends, and considering the desirability of uniform inlet flow, an improved waverider shape for engine-airframe integration is proposed using a hybrid cone-wedge generating flowfield.
Lewis, Mark J.↗
Efficient Neural Network Approaches for Conditional Optimal Transport with Applications in Bayesian Inference
In this work, we present two neural network approaches that approximate the solutions of static and dynamic conditional optimal transport (COT) problems. Both approaches enable conditional sampling and conditional density estimation, which are core tasks in Bayesian inference—particularly in the simulation-based (“likelihood-free”) setting. Our methods represent the target conditional distribution as a transformation of a tractable reference distribution. Obtaining such a transformation, chosen here to be an approximation of the COT map, is computationally challenging even in moderate dimensions. To improve scalability, our numerical algorithms use neural networks to parameterize candidate maps and further exploit the structure of the COT problem. Our static approach approximates the map as the gradient of a partially input convex neural network. It uses a novel numerical implementation to increase computational efficiency compared to state-of-the-art alternatives. Our dynamic approach approximates the conditional optimal transport via the flow map of a regularized neural ODE; compared to the static approach, it is slower to train but offers more modeling choices and can lead to faster sampling. We demonstrate both algorithms numerically, comparing them with competing state-of-the-art approaches, using benchmark datasets and simulation-based Bayesian inverse problems.
97 MATHEMATICS AND COMPUTING↗
Optimal low-thrust takeoff from an orbit about an oblate planet
Future space missions to the outer planets may depend upon the use of low-thrust propulsion systems. As these planets are decidedly oblate, the question of the effect of that oblateness on a low-thrust trajectory is of some interest. In this paper the problem of optimal energy increase is attacked under the assumption that the coefficients for the second zonal harmonic, and the nondimensional thrust acceleration are the same order of magnitude. By means of a two-variable asymptotic expansion technique, a near optimal control program is generated and the first-order uniformly valid approximation for the corresponding trajectory is obtained. Tangential thrust is shown to be a good near-optimal thrust program even in the presence of oblateness effects. The optimal control program is found to be oscillatory and quite similar to the optimal control for energy increase in an inverse square gravitational field.
Jacobson, R. A.↗
Analytical modeling of structure-soil systems for lunar bases
The study of the behavior of granular materials in a reduced gravity environment and under low effective stresses became a subject of great interest in the mid 1960's when NASA's Surveyor missions to the Moon began the first extraterrestrial investigation and it was found that Lunar soils exhibited properties quite unlike those on Earth. This subject gained interest during the years of the Apollo missions and more recently due to NASA's plans for future exploration and colonization of Moon and Mars. It has since been clear that a good understanding of the mechanical properties of granular materials under reduced gravity and at low effective stress levels is of paramount importance for the design and construction of surface and buried structures on these bodies. In order to achieve such an understanding it is desirable to develop a set of constitutive equations that describes the response of such materials as they are subjected to tractions and displacements. This presentation examines issues associated with conducting experiments on highly nonlinear granular materials under high and low effective stresses. The friction and dilatancy properties which affect the behavior of granular soils with low cohesion values are assessed. In order to simulate the highly nonlinear strength and stress-strain behavior of soils at low as well as high effective stresses, a versatile isotropic, pressure sensitive, third stress invariant dependent, cone-cap elasto-plastic constitutive model was proposed. The integration of the constitutive relations is performed via a fully implicit Backward Euler technique known as the Closest Point Projection Method. The model was implemented into a finite element code in order to study nonlinear boundary value problems associated with homogeneous as well as nonhomogeneous deformations at low as well as high effective stresses. The effect of gravity (self-weight) on the stress-strain-strength response of these materials is evaluated. The calibration of the model is performed via three techniques: (1) physical identification, (2) optimized calibration at the constitutive level, and (3) optimized calibration at the finite element level (Inverse Identification). Activities are summarized in graphic and outline form.
Macari-Pasqualino, Jose Emir↗
Design of a mixer for the thrust-vectoring system on the high-alpha research vehicle
One of the advanced control concepts being investigated on the High-Alpha Research Vehicle (HARV) is multi-axis thrust vectoring using an experimental thrust-vectoring (TV) system consisting of three hydraulically actuated vanes per engine. A mixer is used to translate the pitch-, roll-, and yaw-TV commands into the appropriate TV-vane commands for distribution to the vane actuators. A computer-aided optimization process was developed to perform the inversion of the thrust-vectoring effectiveness data for use by the mixer in performing this command translation. Using this process a new mixer was designed for the HARV and evaluated in simulation and flight. An important element of the Mixer is the priority logic, which determines priority among the pitch-, roll-, and yaw-TV commands.
Pahle, Joseph W.↗
Intelligent Control Approaches for UAVs
This talk will present overviews of various intelligent control technologies currently being developed and studied at the NASA Ames Research Center as applicable to Unmanned Aerial Vehicles (UAVs), Mars flyers, and to the next generation of flight controllers for manned aircraft. The approaches being examined include: (a) direct adaptive dynamic inverse controller, (b) adaptive critic-based optimal trajectory generator; (c) optimal allocation technique based on linear programming; (4) immunized maneuvering using autopilot building blocks. These approaches can utilize, but do not require, fault detection and isolation information. Piloted and unmanned simulation studies are performed to examine if the intelligent flight control techniques adequately: 1) match flying qualities of modem fly-by-wire flight controllers under nominal conditions; 2) improve performance under failure conditions when sufficient control authority is available; and 3) achieve intelligent maneuvering capabilities for unmanned vehicles. Results obtained so far will be presented and discussed.
KrisnaKumar, Kalmanje↗
A Comprehensive Model of Earth's Magnetic Field Determined from 4 Years of Swarm Satellite Observations
The European Space Agency's three-satellite constellation Swarm, launched in November 2013, has provided unprecedented monitoring of Earth's magnetic field via a unique set of gradiometric and multi-satellite measurements from low Earth orbit. In order to exploit these measurements, an advanced "Comprehensive Inversion" (CI) algorithm has been developed to optimally separate the various major magnetic field sources in the near-Earth regime. The CI algorithm is used to determine Swarm Level-2 (L2) magnetic field data products that include the core, lithospheric, ionospheric, magnetospheric, and associated induced sources. In addition, it has become apparent that the CI is capable of extracting the magnetic signal associated with the oceanic principal lunar semi-diurnal tidal constituent M(sub 2) to such an extent that it has been added to the L2 data product line. This paper presents the parent model of the Swarm L2 CI products derived with measurements from the first four years of the Swarm mission and from ground observatories, denoted as "CIY4", including the new product describing the magnetic signal of the M(sub 2) oceanic tide.
Swarm Satellites↗