A computational method for finding all the roots of a vector function (Q918635)

The problem of finding all solutions of a system of nonlinear algebraic equations \(f(x)=0\); f: \({\mathbb{R}}^ n\to {\mathbb{R}}^ n\), \(f\in C^ 1\) is considered. The theory of autonomous dynamical system is employed to help to locate the roots of the vector function f. The dynamical system is such that its equilibrium points are the roots of f, which are guaranteed to be asymptotically stable. This implies that there exist trajectories in the state space which approach these equilibrium solutions asymptotically. The computational approach utilizes the simple cell-mapping technique, which exploits the idea of discretization of the state space and avoids repetitive, time consuming calculation of trajectories of dynamical systems. The cell-mapping method is also capable of locating stable solutions of strongly nonlinear systems and is computationally well suited for determining the global behavior of the nonlinear systems. It is shown that the cell-mapping method is an efficient and powerful computational tool, not only for locating all the zeros of f, but also for delineating the regions of attraction associated with each of them. The number of the roots located by the method depends on the extent of the state space discretized and fineness of discretization determined by the cell size. Other numerical techniques based on iterative and homotopy methods can make use of these attraction domains to choose appropriate initial guesses. Singular manifolds play an important role in limiting the extent of these domains of attraction. Based on the similar discretization of state space, a novel approach for locating the singular manifolds is developed. Analytical results are established to support the validity of the computational algorithm. Examples are given to illustrate the computational approaches.