Tire parameter identification (PI) for multi-body-system (MBS) tire models is a time-consuming task. Since MBS tire simulations implement highly abstract, effective models, the tire model parameters cannot be interpreted as physical (material) properties directly, they must be seen as macroscopic entities in an integral sense. In a set of typically 30 and more parameters with varying orders of magnitude, optimal values are to be found that make various tire simulation results fit best to experimental data in some sense.

Interpreted as an optimization program, some numerical schemes can be applied to reduced (sub)-sets of programs - but this approach always includes making an initial guess based on physical tire properties or experience. Also, repeated manual interactions take place during a typical program - monitoring the current progress by visual inspection and iterating over this process until a certain optimum is reached. Due to the variety of parameter combinations and the relatively long simulation time, the global optimum for all available measurements usually cannot be found with acceptable effort (if at all); instead a subspace strategy is typically applied.

Hence the definition of “optimum” often lies in the eye of the beholder, i.e. identical PI tasks done by different experts generally lead to different optima. Real test data are acquired by static experiments (e.g. vertical/longitudinal stiffness), stationary experiments (longitudinal/lateral slip) and dynamic experiments (e.g. cleat runs). The way of comparing experimental and simulation results also influences the detected optimum and leads to the problem of defining suitable error criteria (i.e. what is “good” PI).

This paper focuses on a new programmatic approach developed at Fraunhofer LBF relying on a rule based expert system. The new PI-Tool makes PI faster, more standardized, as automatable as possible and the results at least as good as achieved by current processes. Special focus was put on developing error criteria which try to mimic the human skill while comparing measurement and simulation results - a multi-layered and complex process. Various local signal properties and signal processing algorithms are derived to simulate the experiment-conform, visual human inspection.

Together with the presentation of this new PI method, results achieved with our new tool will be shown using the tire model CDTire for an already parameterized tire.