The effects of variability in the parameters used to calculate prestress loss are evaluated by a probabilistic prestress loss computer program. The program accounts for the time-dependent effects caused by creep, shrinkage, and steel relaxation. Losses caused by elastic shortening of the prestressed concrete member are also calculated. Statistical information for the parameters of the program is taken from the literature or from experimental results. Numerical examples show that the variability in the prestress losses exceeds the loss calculated by deterministic methods. This increase in prestress loss may then lead to stresses that exceed allowable stresses. Though the losses determined probabilistically can exceed deterministic losses by 50 percent, it is recommended that deterministic nominal losses be increased by only 25 percent when checking final stresses. This reasoning is based on the fact that allowable stresses also have variability. The study also determined that a normal distribution, with a bias of 1.25, models prestress losses fairly accurately.
