Implementation of the Jacobian-free Newton-Krylov
نویسندگان
چکیده
14 We have implemented the Jacobian-free Newton-Krylov (JFNK) method 15 for solving the first-order ice sheet momentum equation in order to improve 16 the numerical performance of the Community Ice Sheet Model (CISM), the 17 land ice component of the Community Earth System Model (CESM). Our 18 JFNK implementation is based on significant re-use of existing code. For ex19 ample, our physics-based preconditioner uses the original Picard linear solver 20 in CISM. For several test cases spanning a range of geometries and bound21 ary conditions, our JFNK implementation is 1.84-3.62 times more efficient 22 than the standard Picard solver in CISM. Importantly, this computational 23 gain of JFNK over the Picard solver increases when refining the grid. Global 24 convergence of the JFNK solver has been significantly improved by rescaling 25 the equation for the basal boundary condition and through the use of an in26 exact Newton method. While a diverse set of test cases show that our JFNK 27 implementation is usually robust, for some problems it may fail to converge 28 with increasing resolution (as does the Picard solver). Globalization through 29 parameter continuation did not remedy this problem and future work to im30 prove robustness will explore a combination of Picard and JFNK and the use 31 of homotopy methods. 32
منابع مشابه
On Using Approximate Finite Differences in Matrix-Free Newton-Krylov Methods
A Newton–Krylov method is an implementation of Newton’s method in which a Krylov subspace method is used to solve approximately the linear systems that characterize steps of Newton’s method. Newton–Krylov methods are often implemented in “matrix-free” form, in which the Jacobian-vector products required by the Krylov solver are approximated by finite differences. Here we consider using approxim...
متن کاملA Jacobian-free Newton-krylov Iterative Scheme for Criticality Calculations Based on the Neutron Diffusion Equation
Newton-Krylov methods, primarily using the Jacobian-Free Newton-Krylov (JFNK) approximation, are examined as an alternative to the traditional power iteration method for the calculation of the fundamental eigenmode in reactor analysis applications based on diffusion theory. One JFNK approach can be considered an acceleration technique for the standard power iteration as it is “wrapped around” t...
متن کاملMatrix-free second-order methods in implicit time integration for compressible flows using automatic differentiation
In the numerical simulation of inviscid and viscous compressible fluid flow, implicit Newton-Krylov methods are frequently used. A crucial ingredient of Krylov subspace methods is the evaluation of the product of the Jacobian matrix of the spatial operator, e.g., fluxes, and a Krylov vector. In this article we consider a matrix-free implementation of the Jacobian-vector product within the flow ...
متن کاملNonsymmetric Preconditioner Updates in Newton-Krylov Methods for Nonlinear Systems
Newton-Krylov methods, combination of Newton-like methods and Krylov subspace methods for solving the Newton equations, often need adequate preconditioning in order to be successful. Approximations of the Jacobian matrices are required to form preconditioners and this step is very often the dominant cost of Newton-Krylov methods. Therefore, working with preconditioners destroys in principle the...
متن کاملRevisiting Individual Discipline Feasible using matrix-free Inexact-Newton-Krylov
The individual-discipline-feasible (IDF) formulation was proposed to simplify the implementation of MDO problems. The IDF formulation introduces coupling variables into the optimization problem that eliminate the need for a full multidisciplinary analysis at each optimization iteration; this simplifies the solution of MDO problems by maintaining modularity of the discipline software. Historical...
متن کاملSolving Nonlinear Equations with Newton-Krylov Method Based on Automatic Differentiation
The Jacobian-free Newton-Krylov method is widely used in solving nonlinear equations arising in many applications. However, an effective preconditioner is required each iteration and determining such may be hard or expensive. In this paper, we propose an efficient two-sided bicoloring method to determine the lower triangular half of the sparse Jacobian matrix via automatic differentiation. Then...
متن کامل