Abstract
Drained extension tests on cylindrical specimens indicate that strain localization is consistently encountered in the form of specimen necking. The cause of the strain localization, which initiate early in the test, is the inherent instability in the axisymmetric extension test, which allows stresses, and therefore deformations, to concentrate at the weakest part of the specimen. This instability is the result of the inward radial strains experienced in these tests. The conventional extension test is therefore unreliable for determining soil strength in extension. A method is employed to enforce uniform strains in extension tests on cylindrical specimens by the use of curved steel plates separated by lubricated latex membranes. Using such harnesses, two series of tests were performed, one with short plates and one with long plates. The results of these tests are compared with results from conventional extension tests. The extension tests with the long plate harness are most successful in maintaining uniform strains, and they result in higher rates of dilation and higher strengths than obtained in the conventional extension tests. Detailed comparisons between strain-localized and uniform strain tests are presented. Surprisingly, the strengths obtained from the conventional extension tests after proper area correction are much lower than the strengths produced in extension tests on specimens with both long and short harnesses. It is concluded that the sand behaves differently in tests with soft boundaries than in tests with stiff boundaries, which enforce uniform strains.
