Automation of Vibration Mode Identification in Real Vehicle Bodies

Abstract

[Abstract] The early development of a vehicle body relies on modal analysis to evaluate many structural variants, whose global vibration modes (torsion, bending, and the like) must be identified and tracked against design targets. In current industrial practice this is done manually, mode by mode, which is slow, subjective, and hard to reproduce. This project, carried out in collaboration with SEAT S.A., automates the identification of the global modes of a representative vehicle body, taking the torsion mode as the primary case study because of its central role in NVH. Two complementary methods are developed and validated on three models of increasing fidelity: a simplified beam-frame chassis, and a representative vehicle body provided by SEAT in its Body in White and Trimmed Body configurations. The first is a reference-based pipeline built on the Modal Assurance Criterion (MAC), extended with four weighting variants and three optional pre-processing steps. A systematic comparison shows that the most meaningful criterion is not the absolute MAC value but its separation from the next most torsional candidate, and that the mass-weighted MAC with rigid-body removal is the most reliable; on the most detailed model, however, the dense set of local shell modes pushes this approach to its limit. To overcome this, a reference-free geometric torsion score is proposed as an original contribution, scoring a mode directly from the geometry of its displacement field. It identifies the same torsion mode as the MAC on all three models, a strong mutual cross-validation, and resolves it cleanly even where the MAC falters. The complete pipeline is implemented in Python and made openly available, offering an automatic and reproducible alternative to manual mode identification on industrial-scale vehicle bodies.