Uncertainties of digital image correlation near strain localizations

Abstract

Estimates of the strain errors resulting from digital image correlation (DIC) measurements are desired for many uses. One application is the measurement of the strain localization near failure in forming limit testing of sheet metals. This work measures and statistically characterizes the displacement measurement uncertainties for a typical DIC system. These uncertainties are found to have nearly Normal probability distributions and were used as inputs into a Monte Carlo analysis to determine the resulting strain uncertainty. A limited parameter study was made using the Monte Carlo analysis. The results demonstrate that the strain measurement uncertainty is quantifiable, and reducing the virtual gauge length (over which the strain is determined) tends to increase the strain measurement uncertainty. Based on the results, curves relating the strain uncertainty to the DIC analysis parameters can be developed. These curves suggest a balance must be chosen between the optimum processing parameters to minimize the strain error or the parameters to minimize the virtual gauge length. For some engineering applications (e.g. forming limit testing), a smaller virtual gauge length might be preferred even if it results in a higher strain uncertainty, as long as that uncertainty is quantifiable.