Fraunhofer-Gesellschaft

Publica

Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Marker-free GPU-based digital image correlation system for high-temperature strain-controlled fatigue measurements

Presentation held at Annual International DIC Society Conference 2017, November 6-9, 2017, Barcelona, Spain
 
: Blug, Andreas; Regina, David Joel; Eckmann, Stefan; Senn, M.; Eberl, Christoph; Bertz, Alexander; Carl, Daniel

:
Präsentation urn:nbn:de:0011-n-4770841 (3.7 MByte PDF)
MD5 Fingerprint: 169e2ae3d049090329142ff4aeb56bb8
Erstellt am: 16.12.2017


2017, 21 Folien
International Digital Image Correlation Society (DIC Conference) <2017, Barcelona>
Englisch
Vortrag, Elektronische Publikation
Fraunhofer IPM ()
Fraunhofer IWM ()
Werkstoffprüfung; Kurzzeitfestigkeit; Dehnungsregelung; Bildverarbeitung; material testing; low-cycle fatigue; image processing; Bildkorrelation; digital image correlation

Abstract
In contrast to tactile extensometers 2D digital image correlation (2D-DIC) works contactless and thus without slip. It measures not only the average strain between two points, it also allows for full-field analysis, e.g. the analysis of material failure cause. While these advantages also apply to the standard DIC systems available today, such systems so far have a significant disadvantage – their slow measuring speed with maximum sampling rates below 100 Hz. Strain-controlled fatigue measurements with such sensors are possible but limited by their low speed. According to ASTM E 606, at least 400 images per fatigue cycle are necessary to resolve strain amplitude better than 1 %. Hence the fatigue cycle frequency is limited to 0.25 Hz for such DIC systems resulting in a measurement time of more than 11 h for a typical strain-controlled low cycle fatigue experiment with 10.000 cycles. So far, mechanical extensometers have sampling frequencies in the kHz range and thus accelerate those experiments to less than 1 h. To overcome this limitation of slow optical measurement, the 2D-DIC evaluation was implemented on a NVIDIA GeForce GTX 1080 graphics processing unit (GPU) allowing for up to 25.000 FFT evaluations per second (both, forward and backward) of 256 x 256 pixel ROIs. For subpixel displacement calculation, 2D polynomials of second order are fitted to the 3 x 3 pixel area surrounding the correlation maximum. The standard deviation of the “Low Contrast Subpixel Contrast Images” was below 0.01 pixels (1 sigma) for a 30 x 30 pixel kernel size. This high-speed GPU evaluation was paired with a high-resolution telecentric lens with 10 mm FOV minimizing the out-of-plane error, coaxial LED illumination allowing for an exposure time of 300 µs, and a fast CameraLink camera acquiring 2040 x 256 pixel images with 1.3 kHz. This is sufficient to resolve the microstructure even on polished cylindrical samples. A blue LED was used to separate the blackbody radiation from the illumination light by a short pass filter. The result is a high-speed DIC system with a strain measurement rate of 1.2 kHz, 6 ms delay time and a total error in the range of 2*10(-5) (1 sigma). The strain signals were even less noisy than those measured mechanically and they were sufficient to resolve the turning points of a 10 Hz triangular force controlled measurement. So the DIC sensor is well suited for strain-controlled measurements combing the advantages of both, marker-free optical and fast tactile mechanical extensometers.

: http://publica.fraunhofer.de/dokumente/N-477084.html