360° full-field vibration measurement of cylindrical shells based on 3D-DIC: FOV-constrained optimization and non-overlapping mode shape stitching

Abstract

The accurate extraction of full-field dynamic parameters of large cylindrical shell structures is important for structural design optimization and health monitoring. To address the difficulty of reconciling wide-FOV coverage with high spatial resolution in visual measurements of complex curved surfaces, together with the cumulative errors readily introduced by stitching overlapping regions across multiple views, this paper proposes a FOV-constrained optimization method and a non-overlapping full-field stitching method based on partitioned mode shapes. Specifically, instead of relying on cascaded features in overlapping adjacent measurement regions, the algorithm independently maps and stitches the mode shape of each partition in a pre-established global coordinate system to achieve 360° full-field coverage of the cylindrical shell. First, by jointly considering the frequency-domain signal-to-noise ratio requirement for micro-amplitude vibration measurement and the depth-of-field boundary of curved-surface imaging, a FOV optimization model under multi-physical constraints is established, and a partitioning criterion for maximizing the effective single-view FOV is derived. Second, based on the fact that mode shapes are inherent spatial properties of linear time-invariant systems, the local mode shapes acquired from different partitions are assembled into full-field 3D mode shapes through global coordinate mapping, excitation-energy normalization, and unified phase reference. A stitching error analysis model based on the modal assurance criterion (MAC) is also established. Experiments on an aluminum alloy cylindrical shell demonstrate that the natural frequencies and damping ratios extracted by the vision-based method agree well with accelerometer measurements and finite element results. Furthermore, cross-validation between the reconstructed full-field mode shapes and 60 accelerometer measurement points yields MAC values above 0.90, verifying the effectiveness of the proposed method.