One of the common sources of error in triaxial tests is the penetration of
the latex membrane into the peripheral voids of the specimen as confining p
ressure is applied. Existing analytical solutions developed to predict the
magnitude of this membrane penetration have all assumed that the peripheral
sand particles in contact with the surrounding membrane are arranged so th
at the centroids of the sand-membrane contact areas form a square pattern o
f side length D-50 on a plane parallel to the undeflected membrane. Measure
ments made on digital images of sand-membrane contact patterns for triaxial
specimens of varying characteristics have shown this assumption to be inad
equate. The analysis presented herein shows how actual contact pattern can
be reasonably replicated by simulated patterns. These simulated patterns ar
e generated by adding randomly selected distances to the centroid co-ordina
tes of contact areas initially placed in a systematic grid pattern. The res
ults indicate that it is possible to systematically model sand-membrane con
tact patterns as a function of triaxial specimen characteristics such as re
lative density and preparation method. The ability to simulate the real pat
terns eliminates the reliance of existing analytical solutions on a single
assumed square contact pattern and hence can serve as the basis for improvi
ng the accuracy of future solutions to account for the effects of membrane
penetration. Copyright (C) 2000 John Wiley & Sons, Ltd.