Key points are not available for this paper at this time.
Abstract Determining the compression strength of advanced ceramics would appear to be a simple and straightforward process; however, it is not. During compression testing, tensile stresses typically develop at the specimen/load platen interface, leading to the formation of macrocracks at the specimen ends. These macrocracks then propagate through the specimen parallel to the loading direction. This type of failure is “end splitting,” which does not represent the true compressive failure process in ceramics and commonly results in misleading, low‐strength values with a significant scatter in the data. Cuboids and/or cylinders are the common specimen geometries used to determine compression strength, yet end splitting is the regularly observed fracture process with these specimens. A dumbbell‐shaped specimen has been shown to drastically reduce the likelihood of end splitting by promoting fracture in the reduced diameter section, leading to higher and more consistent strength values. This manuscript reviews the accepted compressive failure process in ceramics, examines issues related to the use of cuboids and cylinders as test specimens, the various standard test methods and specimens available, the development of a dumbbell‐shaped specimen, the applicability of the dumbbell geometry over a range of strain rates (0.0001/s to ∼300/s), previous studies comparing strength values, and the different failure process observed in ceramics when different specimen geometries are used. A summary and discussion of compressive strength data generated over different strain rates using a dumbbell specimen on a wide range of advanced ceramics is presented. Where available and appropriate, this new data is compared with previous values obtained using cuboids and cylinders. It is definitely shown that a dumbbell shape is the most appropriate and only specimen geometry that should be used to generate compression strength values for ceramics at strain rates between 10 −4 /s and 10 2 /s.
Swab et al. (Wed,) studied this question.