Investigation of long-term prestress losses in pretensioned high performance concrete girders.

Effective determination of long-term prestress losses is important in the design of prestressed concrete bridges. Over-predicting prestress losses results in an overly conservative design for service load stresses, and under-predicting prestress losses, can result in cracking at service loads. Creep and shrinkage produce the most significant time-dependent effect on prestress losses, and research has shown that high performance and high strength concretes (HPC and HSC) exhibit less creep and shrinkage than conventional concrete. For this reason, the majority of traditional creep and shrinkage models and methods for estimating prestress losses, over-predict the prestress losses of HPC and HSC girders. Nine HPC girders, with design compressive strengths ranging from 8,000 psi to 10,000 psi, and three 8,000 psi lightweight HPC (HPLWC) girders were instrumented to determine the changes in strain and prestress losses. Several creep and shrinkage models were used to model the instrumented girders. For the HPLWC, each model over-predicted the long-term strains, and the Shams and Kahn model was the best predictor of the measured strains. For the normal weight HPC, the models under-estimated the measured strains at early ages and over-estimated the measured strains at later ages, and the B3 model was the best-predictor of the measured strains. The PCI-BDM model was the most consistent model across all of the instrumented girders. Several methods for estimating prestress losses were also investigated. The methods correlated to high strength concrete, the PCI-BDM and NCHRP 496 methods, predicted the total losses more accurately than the methods provided in the AASHTO Specifications. The newer methods over-predicted the total losses of the HPLWC girders by no more than 8 ksi, and although they under-predicted the total losses of the normal weight HPC girders, they did so by less than 5 ksi