When is a rigid pelvis–trunk approximation sufficient? An output-specific model-reduction protocol applied to lumbar mobility in multibody biomechanics

This preprint presents an output-specific model-reduction protocol for evaluating when a simplified rigid pelvis–trunk representation is sufficient in multibody biomechanics. Three mass-equivalent architectures are compared: an exact lumbar lock, axial-yaw-only lumbar mobility, and three-coordinate lumbar mobility. The study uses the public ModelicaHumanBodyPArts v0.10.0 RC1 library and a separate reproducible experiment package. The verification campaign includes six mirrored nominal cases, a 45-case operating sweep, 20 passive-lumbar sensitivity simulations, 18 numerical-refinement runs, and an analytical two-inertia benchmark. Model reduction is evaluated using both absolute deviations and an output-specific locked-normalized error ratio. The results show that reduction suitability depends on the output and operating range: some global rotational quantities are reproduced closely by the reduced model, while contact-related quantities are less transferable. The study is computational and is not a human or clinical validation. Complete reproduction data and executable experiment material are archived separately on Zenodo at DOI 10.5281/zenodo.22957144. The underlying ModelicaHumanBodyPArts v0.10.0 RC1 software is archived at DOI 10.5281/zenodo.22941517.