| Abstract: | The convergence and stability of step-by-step integration schemes used in the inelastic dynamic analysis of structures and their corresponding criteria were studied for a restoring force model with negative-stiffness. Convergence conditions and stability conditions 1, 2 or 3 were established. The numerical stability of the integration under negative-stiffness belongs to the category of relative stability; consequently, the concepts and the conclusions concerning numerical stability in the case of positive-stiffness (which belongs to absolute stability) cannot be used. Research into several step-by-step integration methods usually employed in inelastic dynamic analysis has shown great differences in numerical stability for models with negative-stiffness as compared with positive-stiffness models. The central difference method is convergent and unconditionally stable in the case of negative-stiffness, though it is only conditionally stable in the case of positive-stiffness. The Houbolt method satisfies the requirement for convergence; its stability, however, depends not only on the integration step size Δt but also on the stiffness ratio β for the model with negative-stiffness, unlike the unconditional stability for the model with positive-stiffness. The Newmark constant acceleration method is convergent and unconditionally stable in the case of negative-stiffness just like it is in the case of positive-stiffness. |
