| The objectives of current research can be summarized as follows. (1) Examining the compressive fracture processes in cement-based materials. The variables considered include composition of materials, strength of materials, length of specimens, effect of end conditions, and testing conditions. (2) Developing a non-destructive evaluation (NDE) method which is suitable for obtaining accurate full-field information on compressive failure. (3) Establishing theoretical predictions with an approach based on fracture mechanics.;Mechanical experiments under compressive loads are conducted on more than 50 cylindrical concrete specimens which have different strengths and lengths. Fracture response such as pre-peak behavior, peak loads, strain softening, localization, and failure patterns are examined. The effects of testing conditions which are examined include the rotation of loading platens and feedback signal.;A non-destructive evaluation system which supplies full-field deformations with an accuracy within the micron range is developed. The system is based on digital image correlation scheme (often called computer vision). Using this measurement system, the failure patterns and crack propagations on prismatic specimens (cement paste, mortar, and concrete) are examined. In cement paste, a single crack initiates around the peak load from a loading surface and propagates in the loading direction. In concrete, cracks initiate from inside of a specimen as well as from the loading surface. Inclined cracks are observed at the interface of aggregates and vertical (loading direction) cracks are observed in the matrix.;Theoretical predictions on crack growth are made using fracture mechanics based approaches in conjunction with experimental observations. Two different approaches for the computation of the stress intensity factors are made; one (for cement paste) is based on discrete cracks in a three-dimensional body, and the other (for concrete) is based on multiple sliding cracks relying on damage mechanics concept. The theoretically predicted values agree well with the measured data. |