Online evaluation of microscale fatigue crack propagation using a real-time digital image correlation method

To achieve quasi-real-time crack growth evaluation in in-situ fatigue tests, an automated approach is proposed that could potentially benefit failure prevention by addressing fatigue crack propagation. This method significantly reduces the manual effort required to extract geometric characteristics, such as crack paths and lengths, while enhancing the efficiency of fatigue crack studies in the early stages. The framework consists of a series of operations, including image filtering, correlation search, static feature recognition, connected domain analysis, and crack block connection, to automatically identify microscale cracks and determine growth rates in nearly real time. The influence of input identification parameters on the effectiveness of the recognition process is also examined. Additionally, the framework proves effective in identifying cracks under various complex surface conditions, demonstrating its robustness. Finally, three methods based on this framework are evaluated by comparing their results with those obtained through human labeling in reverse time order.