Abstract
During fatigue tests of tidal turbine blades, digital image correlation (DIC) is used to collect vital information about the specimen. DIC provides high-resolution displacement and strain maps of selected blade sections; however, continuous operation is hindered by the need to acquire, transfer, and process large volumes of high-resolution images, precluding real-time use during long tests. We address this problem by optimising sparse sensing locations on the blade surface so that full-field maps can be accurately reconstructed from a small subset of pixel measurements. In contrast to most DIC improvements found in the literature, which focus on accelerating the processing stage, this approach circumvents the need to collect high-resolution data. We evaluate this approach in a case study at FastBlade, a dedicated testing facility for tidal turbine blades. With less than 1% of the original pixels measured, the mean relative error evaluated on the dataset is 0.4% and 16% for displacement and strain maps, respectively, with the larger strain error reflecting the higher spatial complexity of strain fields. The optimised layouts outperform random and grid-like arrangements. The framework enables real-time monitoring and, subject to relevant validation, might be applied to reconstruct high-resolution strain maps directly from strain-gauge readings, potentially extending to in-ocean blade monitoring. Given the high accuracy of deflection reconstructions, using them to derive strain fields is suggested as a direction for further study.