When instalhng suction caissons in sand, the suction process will induce an upward flow, which reduces the effective stress in the soil inside the caisson. The suction pressure cannot be increased indefinitely. I f a critical suction pressure is exceeded, liquefaction, boiling and/or piping occurs. This will halt the installation process. This paper presents results from a novel laboratory upward flow test (LUFT), to investigate the soil mechanisms affecting the critical suction pressure. The LUFT apparatus is based on a conventional permeameter. As suggested by the results of published full/small scale installation tests, LUFT results confirm that, in dense to very dense sand, most conventional critical suction prediction models underestimate the values observed in reality. It is argued that soil arching contributes to the achievable high values for critical
suction pressure. Higher allowable suction pressures may be cost effective.