A continuously rising level of virtualization throughout all development phases plays an important key role to reduce real hardware tests and the corresponding amount of expensive prototypes. Highly accurate models are necessary to transfer test drives on the road to complex multibody simulations or to enable a subjective evaluation on a dynamic driving simulator, respectively. Simultaneously, highly automated and autonomous driving leads to increased customer requirements regarding ride comfort characteristics, especially due to road excited vibrations. Since the automotive shock absorber is a central element in the process of suspension tuning, modeling its complex nonlinear dynamic behavior is crucial. Therefore, a new modular and real-time capable modeling concept based on local linear model networks is proposed in this contribution. Extensive component measurements of various shock absorber setups focusing on different secondary ride comfort characteristics were carried out and serve as a reference for both parameterization and the definition of required model accuracy. An efficient process of model parameter and topology optimization is shown. In this context, the model is validated by using stochastic excitations derived from real full vehicle measurements on representative roads.