In many mechanical systems, the tendency of sliding components to intermittently stick and slip leads to undesirable performance, vibration, and control behaviors. Computer simulations of mechanical systems with friction are difficult because of the strongly nonlinear behavior of the friction force near zero sliding velocity. In this paper, two improved friction models are proposed. One model is based on the force-balance method and the other model uses a spring-damper during sticking. The models are tested on hundreds of lumped mass-spring-damper systems with time-varying excitation and normal contact forces for both one-dimensional and two-dimensional stick-slip motions on a planar surface. Piece-wise continuous analytical solutions are compared with solutions using other published force-balance and spring-damper friction models. A method has been developed to set the size of the velocity window for Karnopp’s friction model. The extensive test results show that the new force-balance algorithm gives much lower sticking velocity errors compared to the original method and that the new spring-damper algorithm exhibits no spikes at the beginning of sticking. Weibull distributions of the sticking velocity errors enable maximum errors to be estimated a priori.