Dual-Speed, Dual-Resolution Digital Image Correlation Method Based on Physics-Informed Neural Networks

Due to memory and bandwidth limitations, it is difficult for conventional high-speed cameras to meet simultaneously the demands of high frame rate and high resolution testing. In dynamic loading tests that require high-speed image acquisition (e.g., explosion and impact tests), limiting the image resolution will lead to insufficient displacement field resolution accuracy of the DIC algorithm. To address the above problems, a dual-speed, dual-resolved DIC method was proposed based on PINN (PINN-DSDR-DIC). The method was arranged to realize dual-speed dual-resolution sparse image acquisition with a spectral optical path, and to realize high-precision resolution with an algorithm for high-speed high-resolution displacement field. Firstly, incorporating the temporal and spatial continuity of displacement as well as the gray scale consistency into PINN, the algorithm was designed to solve the high-speed, high-resolution and high-precision displacement field by optimizing the network parameters. And then, some numerical experiments were carried out to prove the feasibility of the PINN-DSDR-DIC algorithm and to analyze the affecting factors on the accuracy of the algorithm. Finally, a spectral path system was built and the PINN-DSDR-DIC method was applied and verified based on four-point bending experiments of graphite with grooved nuclei.