PhD Thesis, University of Cambridge, 2006
H J Prentice
The challenge of this research was to develop a versatile, inexpensive and portable optical system to obtain full‑field displacement measurements. Two single‑camera, stereoscopic systems were paired with digital speckle photography to obtain temporally‑synchronised image pairs of speckle patterns using split‑frame photography; the three‑dimensional surface displacements were measured non‑invasively by correlating digitised image pairs, using a common reference image to improve image correspondence. Such measurements are crucial to the measurement of asymmetric deformations and are ideally suited to numerical model validation.
Chapter 1 introduces the speckle technique and gives its history. The correlation algorithm used is examined in Chapter 2, identifying the importance of pre‑processing data to remove poorly correlated points, revealing that the mean displacement is measured in strained sub‑images, and determining its accuracy in the applications presented. Chapter 3 introduces stereo‑photography and describes the logistic and synchronisation advantages of using a single camera.
Chapters 4 and 5 describe and compare two techniques to acquire stereo images with a single camera. The optical wedge stereo system uses inclined prisms to refract together spatially separated images formed by stereo lenses. Improved resolution compared with the corresponding two‑camera system was achieved though systematic and random distortions incurred though the optical components led to errors up to ± 50 %. The prisms were additionally large, heavy and difficult to align accurately. Superior accuracy is acquired (± 3 %) with the stereo mirror technique which uses a periscopic assembly of rotating mirrors to reflect angled images together with negligible distortion. This system was simple to align and offered a novel, unobtrusive calibration procedure. Its range was limited by the camera’s resolution and framing rate and by the decorrelation of the speckle patterns for large distortions.
Chapter 6 discusses important results and models pertaining to the spherical impact of plates and the fundamentals of hydrocode modelling are introduced in Chapter 7. The stereo DSP mirror system was used in Chapter 8 to verify numerical model predictions (using QinetiQ’s DYNA‑3D and GRIM hydrocodes) of the response of Mild Steel and Copper plates to normal spherical impact. The plates were observed to bulge locally at the impact site, with rings of failure, and dished globally. The experimental displacement uncertainty was ± 0.05 mm. Bulging was most localised in the thinnest, Copper plate owing to greater ductility. Opposing displacement fields across the bulge peak led to a decrease in peak heights with increasing sub‑image size.
A sensitivity investigation of the hydrocode material models employed revealed pre‑stress level to be the most significant factor in determining the response of Copper. Other parameters produced differences comparable with experimental error. Close agreement between experiment and model was obtained for Mild Steel. Stereo measurements for Copper were compared with path‑dependent models for annealed and pre‑stressed (200 MPa) Copper; whilst single‑point measurements of peak displacement history were insufficiently resolved to differentiate between the two models, the full‑field data exposed clear differences in profile form, with greater global deformation and closer agreement with experiment for the pre‑stressed model. This both illustrates an advantage of a full‑field measuring technique such as stereo DSP and underlines the need to characterise materials completely and accurately before modelling. Imprecise modelling of the plate confinement led to differences between the calculated and post‑mortem profile measurements.
Thin metallic plates are often present in layered targets. The internal response of a sugarmock slab, confined between two thin Mild Steel plates, to impact was investigated using speckle radiography (Chapter 9). Shear and pinching mechanisms were observed. Film alignment errors were reduced by pre‑processing data to remove decorrelated and spurious points (typically ± 5 %) and could be further reduced by using double exposure speckle radiography. Future research should combine the stereo speckle method described with speckle radiography to determine the full response of composite targets.
Thesis available upon request.
For local users: edit