Effects of testing procedures on the laboratory determination of swell pressure of expansive soils

There is lack of agreement on laboratory procedures that should be used to measure swell pressure. In this study, effects of apparatus compressibility and applied initial net normal stress on measured swell pressure were revisited, with emphasis on undisturbed specimens. Samples covering a wide range of properties were collected from Texas and Arizona, and effects were explored by examination for consistency of trends in behavior for several "companion" sets of undisturbed specimens. Failure to account for apparatus compressibility results in unintended swell of the specimen, reducing the constant volume swell pressure by as much as half for specimens of this study. Applied initial net normal stress level also affects measured swell pressure. For specimens of this study, swell pressures measured at low initial net normal stress (token load) were found to be about 40%, on average, of swell pressures measured for specimens loaded initially equivalent to field overburden stress. This difference is believed to be primarily due to initial specimen compression. Commonly adopted sampling disturbance corrections to constant volume tests conducted at initial light confining stress (∼ 1-7 kPa) were investigated, and found to give inconsistent values of swell pressure. Comparisons of constant volume to load-back swell pressure are also presented. Findings suggest that application of an initial confining stress level corresponding to field conditions and application of apparatus compressibility corrections as the swell pressure develops are key to the consistent measurement of constant volume swell pressure, and that measured swell pressures from constant volume tests, when carefully performed, are more consistent than swell pressures estimated using methods to account for sampling disturbance.