A self-adaptive time integration algorithm for solving partial differential equations (Q1126657)

A new adaptive time integration method is investigated for solving problems with locally fast-changing physical phenomena in space. For these problems non-uniform spatial grids are commonly used, but a traditional explicit time integration scheme is restricted by the stability criterion which is dictated by the smallest grid spacing and requires excessively small time steps. Small space grid is usually unnecessary in the whole domain. This fact is the base of a new idea to use different time stepsizes at different spatial grid points. The stability criterion is still satisfied at all grid points by using different time stepsizes. Complexity analysis and implementation details are discussed. As a numerical example the nonlinear Burgers equation is used. Complexity analysis and numerical computations as well demonstrate significant improvement of computational efficiency.