This paper presents the speed enhancement of displacement and strain field measurement using graphics processing unit (GPU) towards real-time field measurement. The displacement and strain field measurement technique to be enhanced is a vision-based dot centroid tracking (DCT) method which measures the darkness of each pixel in gray scale, identifies dots marked on a specimen, derives dot centroids using pixel darkness information and derives displacement and strain fields by tracking the centroids and deriving and interpolating the nodal displacements and strains. The proposed enhancement begins by finding parallel processes in both the image analysis phase and the field estimation phase. The centroid derivation and tracking of marked dots in the image analysis phase are computed independently for each marked dot whereas the derivation, interpolation and estimation of the displacement and strain field are computed parallel as well with respect to the interpolated points which covers the evaluated field. The proposed enhancement then accelerates the displacement and strain field estimation by storing the pixel and dot information as a predefined index look-up table and a new measurement in a shared buffer in the GPU’s shared memory. The use of the shared buffer with a look-up table avoids the duplicated memory allocation and contributes to not only speed but also memory usage. Numerical examples have first investigated the validity and performance of the proposed enhancement through parametric studies. A comparison with field measurement using only a central processing unit (CPU) shows that the proposed enhancement gains at least 10 times speedup when the same accuracy is attained. The proposed approach was then applied to the displacement and strain field measurement of a rail in a three-point bending test under relatively fast periodic loadings, and its ability to capture real-time behavior has been demonstrated.

This content is only available via PDF.
You do not currently have access to this content.