The Newton-Raphson method accelerated by using a line search - comparison between energy functional and residual minimization

A line search was combined with the Newton-Raphson method to accelerate the convergence of the iterative calculation in nonlinear magnetic field analysis. As a method for determining a step size for update, the minimization of an energy functional and a square of 2-norm of residual obtained from the finite-element discretization was investigated. It was demonstrated that the energy functional minimization is superior to the residual minimization from the viewpoint of computational cost. The line search is effective even in the magnetic vector potential formulation, which is said to be stable usually.