Measurement of Concrete Samples using Magnetic Resonance Imaging

Emanuele Marfisi

The overall intention of this study is to support the use of Magnetic Resonance Imaging (MRI) for studying mechanical behaviour of concrete. Two separate MRI acquisitions were proposed during previous work to observe the internal structure of concrete and the fractures produced by mechanical loading. Knowledge about the accuracy of the detection was required to use these images. Both proton density and signal relaxivity affect the image intensity and the characterisation of the MR signals is the first step of this investigation.

The observation of cement shows, even at the spatial resolution under study (~ 150 micron), the in-homogeneities of the microstructure; the changes in image intensity could be used to infer the presence of defects, capillaries and weaker regions within the sample. NMR measurements are used to support this study. The properties of the signal define the imaging protocol and therefore the practical limitations of future investigations.

The characterisation of the signal from water-saturated fractures highlighted the presence of variable signal relaxivity, which was partially associated with the geometry of fractures. Model-system samples were used to validate this finding. The ability to obtain proton density images is discussed and the imaging protocol redesigned accordingly. The fact that each fracture thickness has a unique MR response suggested NMR for on-line monitoring of fracture propagation. This can support and assist future MRI investigations.

Tailored image post-processing software was developed to achieve some of the measurements required. Fractures are often smaller than the voxel but the use of proton density images allowed sub-voxel volumetric measurements. These measurements combined with geometrical assumptions can be used to estimate the size of fractures. This analysis was finally compared with the destructive observation of the sample, which confirmed the quality of the results.

The conclusion of this work suggests a revised experimental method and
proposes a number of future investigations.