Leveraging Vision for Structural Identification: A Digital Image Correlation Based Approach

This paper describes a case study on a large scale structural test specimen using 3D-DIC as an image-based metrology approach for structural identification (St-ID). For the identification process, a commercial FEA software package was interfaced with MATLAB to converge on the optimal unknown/uncertain system parameters of the experimental setup. The 3D-DIC results provided a rich full-field dataset that was used in the identification process, which was compared against ground-truth measurements derived from traditional physical in-place sensors typically used in St-ID. For the identification, a novel hybrid algorithm, incorporating a combination of a genetic algorithm and a gradient-based scheme was utilized for updating the FEA model and obtaining the optimal values of the selected parameters. Results demonstrated that deflection, strain, and rotation measurements derived from DIC mirrored those from the ground-truth sensors and allowed for convergence of the updating with a variety of measurement responses that are challenging to acquire in typical applications.