Parameter and force identification through multibody model based virtual sensing on a vehicle suspension

Virtual sensing is an emerging technology that allows to make use of simulation models in a testing context to better understand the vehicle behavior, by for instance estimating hard or expensive to measure quantities in operational conditions. Some examples involve dynamic loads or system parameters, as well as response quantities. State estimators (e.g. Kalman filters) are typically used to develop virtual sensing algorithms and the focus of this paper is on extending a previously proposed framework for coupled state-input estimation for multibody systems to also concurrently perform parameter estimation. Such existing framework describes the multibody equations with Natural Flexible Coordinates and enforces constraints with a Penalty formulation, while the state-input estimation is performed by means of an augmented extended Kalman filter. The extension towards parameter estimation is obtainable for such framework by further augmenting the system with states related to the unknown parameters and relative random walk models. The potential of the approach is then shown on a numerical example in which some parameters (related e.g. to friction) of a suspension system are estimated.