Identification and identifiability of unknown parameters in multibody dynamic systems

Assessing design changes in mechanical systems from simulation results requires both accurate dynamic models and accurate values for parameters in the models. Model parameters are often unavailable or difficult to measure. This study details an identification procedure for determining optimal values for unknown or estimated model parameters from experimental test data. The resulting optimization problem is solved by Levenberg-Marquardt methods. Partial derivative matrices needed for the optimization are computed through sensitivity analysis. The sensitivity equations to be solved are generated analytically. Unfortunately, not all parameters can be uniquely determined using an identification procedure. An issue of parameter identifiability remains. Since a global identifiability test is impractical for even the simplest models, a local identifiability test is developed. Two examples are provided. The first example highlights the test for parameter identifiability, while the second shows the usefulness of parameter identification by determining vehicle suspension parameters from experimentally measured data.