A Proximal Quasi-Newton Trust-Region Method for Nonsmooth Regularized Optimization
Not provided.
SEARCH · Engineering Papers
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.
Not provided.
The SPAR thermal analyzer, a system of finite-element processors for performing steady-state and transient thermal analyses, is described. The processors communicate with each other through the SPAR random access data base. As each processor is executed, all pertinent source data is extracted from the data base and results are stored in the data base. Steady state temperature distributions are determined by a direct solution method for linear problems and a modified Newton-Raphson method for nonlinear problems. An explicit and several implicit methods are available for the solution of transient heat transfer problems. Finite element plotting capability is available for model checkout and verification.
The simulation of spacecraft attitude dynamics and control using the generic, multi-body code called TREETOPS and other codes written especially to simulate particular systems is discussed. Differences in the methods used to derive equations of motion--Kane's method for TREETOPS and the Lagrangian and Newton-Euler methods, respectively, for the other two codes--are considered. Simulation results from the TREETOPS code are compared with those from the other two codes for two example systems. One system is a chain of rigid bodies; the other consists of two rigid bodies attached to a flexible base body. Since the computer codes were developed independently, consistent results serve as a verification of the correctness of all the programs. Differences in the results are discussed. Results for the two-rigid-body, one-flexible-body system are useful also as information on multi-body, flexible, pointing payload dynamics.
Abstract not provided.
Explore the source record for details and available documents.
This paper describes an efficient and nonlinearly consistent parallel solution methodology for solving coupled nonlinear thermal transport problems that occur in nuclear reactor applications over hundreds of individual 3D physical subdomains. Efficiency is obtained by leveraging knowledge of the physical domains, the physics on individual domains, and the couplings between them for preconditioning within a Jacobian Free Newton Krylov method. Details of the computational infrastructure that enabled this work, namely the open source Advanced Multi-Physics (AMP) package developed by the authors is described. Details of verification and validation experiments, and parallel performance analysis in weak and strong scaling studies demonstrating the achieved efficiency of the algorithm are presented. Furthermore, numerical experiments demonstrate that the preconditioner developed is independent of the number of fuel subdomains in a fuel rod, which is particularly important when simulating different types of fuel rods. Finally, we demonstrate the power of the coupling methodology by considering problems with couplings between surface and volume physics and coupling of nonlinear thermal transport in fuel rods to an external radiation transport code.
Here, this study evaluates the performance of multiple machine learning (ML) algorithms and electrical resistivity (ER) arrays for inversion with comparison to a conventional Gauss-Newton numerical inversion method. Four different ML models and four arrays were used for the estimation of only six variables for locating and characterizing hypothetical subsurface targets. The combination of dipole-dipole with Multilayer Perceptron Neural Network (MLP-NN) had the highest accuracy. Evaluation showed that both MLP-NN and Gauss-Newton methods performed well for estimating the matrix resistivity while target resistivity accuracy was lower, and MLP-NN produced sharper contrast at target boundaries for the field and hypothetical data. Both methods exhibited comparable target characterization performance, whereas MLP-NN had increased accuracy compared to Gauss-Newton in prediction of target width and height, which was attributed to numerical smoothing present in the Gauss-Newton approach. MLP-NN was also applied to a field dataset acquired at U.S. DOE Hanford site.
The objective of this paper is to develop a micromechanics approach to homogenizing finitely deformed viscoelastic-viscoplastic composites using the mechanics of structure genome. The incremental constitutive relation for glassy polymers, formulated in the spatial configuration, is implemented in the present approach.This involves (1) pulling-back the constitutive model to the material configuration and (2)choosing the deformation gradient tensor and the first Piola–Kirchhoff stress tensor as the strain and the stress measures during homogenization, respectively. An Euler–Newton predictor–corrector method is developed for homogenization. Each step involves formulating a variational statement using the mechanics of structure genome, discretizing the statement in a finite-dimensional space, and solving the problem using an Euler/multilevel Newton method. The present approach is demonstrated by homogenizing fiber- and particle-reinforced composites undergoing uniaxial, biaxial, or shear deformation, at different stain rates.
A new algorithm for solving the duals of separable convex optimization problems is presented. The algorithm is based on an active set strategy in conjunction with a variable metric method. This first order algorithm is more reliable than Newton's method used in DUAL-2 because it does not break down when the Hessian matrix becomes singular or nearly singular. A perturbation technique is introduced in order to remove the nondifferentiability of the dual function which arises when linear constraints are present in the approximate problem.
This paper focuses on a method to solve the static configuration for a lumped mass cable system. The method demonstrated here is intended to be used prior to performing a dynamics simulation of the cable. Conventional static analysis approaches resort to dynamics relaxation methods or root-finding algorithms (such as the Newton-Raphson method) to find the equilibrium profile. The alternative method demonstrated here is general enough for most cable configurations (slack or taut) and ranges of cable elasticity. The forces acting on the cable are attributable to elasticity, weight, buoyancy, and hydrodynamics. For the three-dimensional problem, the initial cable profile is obtained by solving three equations, regardless of the cable discretization resolution. This analysis discusses regions and circumstances under which failures in the method are encountered.
The basic principle in Newtonian Mechanics is based upon equal and opposite forces. Placing the vectors of velocity, acceleration, force and momentum of interacting objects along a single line satisfies the claim it is a linear or a 1-D concept. Classical Mechanics states that there are two main kinds of motion, linear and angular motion. Similarly placing the vectors of angular velocity, angular acceleration, torque and angular momentum along a line in the case of rotation in fact brings a plane 2-D interaction to the well known 1-D Newtonian concept. This adaptation transforms Classical Mechanics into a 1-D concept as well and presents a conformation that the linear concept is the only possible one. The Laws of Conservation of Momentum and Angular Momentum are results of the 1-D concept. But the world contains 3 geometrical spatial dimensions. Within the 3-D world there can exist 1-D, 2-D and 3-D kinds of interaction. The question is how to believe that the 3-D world can really be composed of a 1-D interaction or interactions made equal to the 1-D concept only? Examine a gyroscope - the only one mechanical device that is capable of performing 3-D behavior. The problem is that a gyroscope cannot perform three permanent and unidirectional torques that are fixed in space acting about perpendicular axes. This impossibility conforms to a 1-D concept. The idea is to find a solution that can be achieved for the 3-D concept.
The Multiphysics Object-Oriented Simulation Environment (MOOSE) enables a wide range of advanced nuclear reactor simulations.[6] Under the guidance of MOOSE’s Thermal Hydraulics Team,I worked to expand the capabilities of Tabulated Fluid Properties (TFP) in the fluid properties module. The fluid properties module allows the user to determine a variety of fluid properties by interpolating points between tabulated data. I implemented the ability to use bilinear interpolation instead of bicubic interpolation for interpolating tabulated data. I also changed the method of variable set inversions to use a 2-dimensional Newton’s Method utility that I created. Variable set inversions are often done from (v,e) to (p,T), where v is specific volume, e is specific internal energy, p is pressure and T is temperature. New routines have also been added into TFP such that it can be used with more applications, such as the Navier Stokes and Thermal Hydraulics modules in MOOSE for Pronghorn[5] and RELAP-7[1] respectively. This work was spurred by interest from NASA in testing a Nuclear Thermal Propulsion (NTP) engine system. NTP engines have drastically different fluid properties throughout the engine and Tabulated Fluid Properties provides the flexibility needed to properly simulate and test these engines.
Many applications require minimizing the sum of smooth and nonsmooth functions. For example, basis pursuit denoising problems in data science require minimizing a measure of data misfit plus an $\ell^1$-regularizer. Similar problems arise in the optimal control of partial differential equations (PDEs) when sparsity of the control is desired. Here, we develop a novel trust-region method to minimize the sum of a smooth nonconvex function and a nonsmooth convex function. Our method is unique in that it permits and systematically controls the use of inexact objective function and derivative evaluations. When using a quadratic Taylor model for the trust-region subproblem, our algorithm is an inexact, matrix-free proximal Newton-type method that permits indefinite Hessians. We prove global convergence of our method in Hilbert space and demonstrate its efficacy on three examples from data science and PDE-constrained optimization.
Low thrust trajectory optimization, using Newton- Raphson method to solve nonlinear two-point boundary value problem
Low thrust trajectory optimization, using Newton- Raphson method to solve nonlinear two-point boundary value problem
Newton-Raphson method generalized for solution of two-point boundary value problems of nonlinear optimal control theory for digital solution
The paper describes algorithms for equilibrium and steady-state problems with coefficients in the expansions derived by the Galerkin weighted residual method and calculated from the resulting sets of nonlinear algebraic equations by the Newton-Raphson method. Initial approximations are obtained from nearby solutions by continuation techniques as parameters are varied. The Newton-Raphson technique is preferred because the Jacobian of the solution is useful for continuation, for analyzing the stability of solutions, for detecting bifurcation of solution families, and for computing asymptotic estimates of the effects on any solution of small changes in parameters, boundary conditions, and boundary shape.
Reversibility of a computational method is the ability to execute the method forward as well as backward. Reversible computational methods are generally useful in undoing incorrect computation in a speculative execution setting designed for efficient parallel processing. Here, reversibility is explored of a common component in scientific codes, namely, the Newton-Raphson root-finding method. A reverse method is proposed that is aimed at retracing the sequence of points that are visited by the forward method during forward iterations. When given the root, along with the number of iterations, of the forward method, this reverse method is aimed at backtracking along the reverse sequence of points to finally recover the original starting point of the forward method. The operation of this reverse method is illustrated on a few example functions, serving to highlight the method's strengths and shortcomings.