Usage of Digital Image Correlation in Analysis of Cracking Processes

Abstract

In this paper, the analysis of the possibilities of using Digital Image Correlation (DIC) based on Graphics Processing Unit (GPU) for strain analysis in fatigue cracking processes is presented. The basic assumption for the discussed displacement and strain measurement method under time variable loads was obtaining high measurement sensitivity by simultaneously minimizing the measurement time consumption. For this purpose special computing procedures based on multiprocessor graphic cards were developed, which significantly reduced the total time of displacement and strain analysis. The developed digital procedure for correlation of images has been used for an example of displacement analysis in the method of fatigue crack propagation testing in airplane riveted joints. In this paper are presented the results of the researches of the team run by professor Antoni Zabłudowski

High accuracy digital image correlation powered by GPU-based parallel computing

“…In recent years, the compute unified device architecture (CUDA) developed by NVIDIA has become the most popular parallel computing platform and programming model, which allows flexible uses of their GPU products for high performance computing. Very recently, researchers introduced CUDAbased parallel computing into the integer-pixel DIC algorithms [14][15][16]. It was found that the fast Fourier transform-based cross correlation (FFT-CC) could be accelerated by up to 23.4 times [16].…”

Heterogeneous parallel computing accelerated iterative subpixel digital image correlation

“…In comparison with the global DIC and DVC methods based on finite elements, local DIC and DVC methods based on subsets (or sub-volumes) seem to be more suitable for parallel computing technology due to their nature of windowed processing. GPU parallel computing can be simply used in cross-correlation-based integer-pixel DIC algorithms for magnetic resonance image processing [22], crack propagation analysis [23] and deformation measurement of human body (wrist and calf) [24]. The speed-up ratio of the GPU version to the serial CPU version attained in refs.…”

“…The speed-up ratio of the GPU version to the serial CPU version attained in refs. [21][22][23][24] varies from 10 to 23.…”

Towards high-speed optical measurement of shape and deformation

“…GPU-accelerated integer-pixel level DIC method has been proposed by Gembris et al [115], and applied in cracking analysis [116], biomechanics [49], and optical coherent tomography (OCT) [50]. The ZNCC can be either parallelized straightforwardly in space domain [50,117] or in frequency domain using the fast Fourier transform (FFT) ZNCC (FFT-CC) algorithm [49,116]. The FFT-CC algorithm takes the advantage that crosscorrelation calculations in the frequency domain become pointwise complex number multiplications, which well fit the pixelwise pattern.…”

“…If all POIs in DIC were processed simultaneously, a dramatic speedup could be expected. The parallelization, in fact, has been successfully realized in integer-pixel DIC methods [49,59,115,116], in which the FFT-CC algorithm has been well accelerated on the GPU. However, as announced in [48], the heaviest computation burden of DIC resides in its iterative sub-pixel registration…”