Abstract
Identification of material parameters is critical for providing reliable finite element analysis (FEA) and advancing predictive modelling in engineering applications. Optical measurement techniques are increasingly used in experimental mechanics, with digital image correlation (DIC) being one of the most widely adopted methods for capturing full-field displacement and strain. These data-rich measurements offer significant potential for validating numerical models and identifying constitutive parameters. However, this potential is often underutilised, as traditional approaches still rely on point-wise measurements (e.g., from virtual extensometers).
A promising strategy to exploit full-field data is the DIC-based finite element method (FEM) approach, an inverse framework that integrates DIC with FEM for material parameter identification. By using a DIC implementation that aligns with FE formulations and two-dimensional meshes, this approach bridges experimental measurements and numerical simulations. This enables simultaneous identification of multiple parameters, accommodates complex geometries and heterogeneous materials, and reduces the number of physical tests required for calibration.
To advance the DIC-based FEM, we introduce a virtual laboratory (VL) framework based on synthetic data. Inverse methods must be robust, computationally efficient, and capable of delivering accurate results even under noisy or uncertain experimental conditions. An essential step in the method development is the use of synthetic data to decouple and quantify errors originating from measurement noise (e.g., camera artifacts) and those inherent to correlation algorithms. Simulated displacement fields are applied to speckle patterns to generate numerically deformed images, providing ground truth for systematic error quantification and sensitivity analysis. This VL framework is essential for validating DIC-based FEM and for designing specimens that activate all relevant material parameters.
This work demonstrates how VLs enhance robustness, allow for determining complex sample geometries, and improve parameter identification for advanced constitutive models.