Efficient parameter identification in nonlinear multi-body dynamics through frequency response and harmonic distortion analysis

This paper presents a methodology to identify sensitive parameters in nonlinear multi-body dynamic systems. It aims to reduce computational cost by minimising the number of numeric calculations required for statistical evaluation, while reliably identifying significant influencing parameters. The proposed approach consists of three major steps: In the first step, a design space analysis based on the frequency response function pre-filters parameters with negligible influence. It verifies the model’s validity across the parameter space for linear dependencies. The second step applies the design of experiments based on the total harmonic distortion to investigate the nonlinear effects of the retained parameters. The third step uses a multiple linear regression analysis to estimate and validate the non-measurable or virtual parameters by identifying linear and multifactorial relationships. This integrated approach reduces the required number of simulations, while the application of the frequency response function and the total harmonic distortion as part of a cost function enables robust statistical evaluations. The methodology is validated by comparing simulation and experiment using a consistent setup, in which a small shaker excites the system with a logarithmic chirp signal. The resulting spectrograms of simulated and measured acceleration signals reveal comparable trends, including observable higher harmonics. These similarities indicate the validity of the method, while discrepancies in the absolute values between simulation and experiment can be attributed to inaccurate model assumptions. Thus, future research could consider testing a simplified or refined benchmark system to further investigate the applicability of the approach.