This research focuses on deviations from the linear viscoelastic behavior of concrete occuring at high stress levels (from 0.5 f’c to 0.7 f’c), at early age loading (1 to 2 days) and in case of unloading implying strain reversal. A large series of creep tests was performed on high strength concrete specimens undergoing creep under constant stress, followed by a period of recording of the creep recovery after complete unloading. Some specimens were heat cured before loading. Some nonlinear effects at very early age have been observed. After unloading, experimental data show that the creep recovery deviates strongly from the numerical predictions obtained by the application of the principle of superposition but seems to conform rather well to the recovery model proposed by Yue and Taerwe3. This model was then applied, through a step-by-step approach, for the time-dependent structural analysis of a precast composite prestressed bridge deck with 26 m span. The application of the recovery model yielded computed strains which are in good agreement with in situ measured strains, and in better agreement than the strains computed by the application of the principle of superposition. This enhanced approach was then used to optimize the phases of construction of this kind of structure. Thanks to this research, the age at transfer of prestress could be significantly reduced.

This paper illustrates how stress relaxation can be used to obtain valuable information regarding the behavior of concrete at early ages. Five concrete mixtures were investigated using a so-called discretized restrained shrinkage (DRS) testing device, allowing the determination (from the time of casting) of the increase in load induced by autogenous shrinkage and the evaluation of the different strain components (free shrinkage, elastic strain, creep). Test results indicate that the stress due to early-age restrained autogenous shrinkage is quite variable, in good part due to the variation in the relaxation capacity of the mixtures. Both the relaxation ratio, defined as the stress generated divided by the theoretical stress, and the relative relaxation, defined as the absolute value of stress relaxation divided by the average applied stress, can be used to illustrate and analyze the variation of the relaxation phenomena as a function of the type of mixture tested.

Creep and shrinkage data for two high strength lightweight aggregate concretes were collected over a two-year period. The concretes, with unit weight of 1922 kg/m3 (120 pcf), were developed using expanded slate as coarse aggregate. Strengths of 55.2 MPa (8,000-psi) and 69.0 MPa (10,000-psi) were obtained at 56 days. Creep specimens were loaded to 40 or 60 percent of the initial compressive strength at 16 or 24 hours after casting. Based on this preliminary study, AASHTO-LRFD creep estimates of high strength, lightweight aggregate concrete were within 20% accuracy for ages later than one month. ACI-209 estimated creep of the 55.2 MPa lightweight concrete and shrinkage of the 69.0 MPa concrete within 20% accuracy, but greatly underestimated shrinkage of the 55.2 MPa mix. When compared with normal weight, high strength concrete of similar strength and similar cement paste content from previous research, the 69.0 MPa lightweight mix experienced lower total strain after two years.

This paper presents the results of a 10-year instrumentation and monitoring program on the North Halawa Valley Viaduct, a major prestressed box girder viaduct on the Island of Oahu, Hawaii. The long-term monitoring program was initiated in 1994 during construction of the long-span post-tensioned box-girder viaduct. Over 200 electrical strain, displacement, temperature and load sensors were installed in one unit of the structure and have been monitored continuously since. These instruments monitor vertical deflections, span shortening, prestress loss, longitudinal strains and temperature in the box-girder concrete. The long-term response of this structure is presented and compared with the initial predictions made during the design process. Modified material properties based on short-term shrinkage and creep tests were incorporated into the long-term prediction model to produce significantly improved comparisons. A procedure is proposed for prediction of upper and lower bounds for the long-term response of long-span prestressed concrete bridges. This improved prediction model is applied to the other five units making up the NHVV to verify its performance as a design tool. The results of this study were then incorporated into the development of an instrumentation system for the planned Kealakaha Bridge on the Island of Hawaii. Application of the prediction model is demonstrated using shrinkage and creep data determined from short-term tests performed on the concrete mixture proposed for this new long-span box-girder bridge structure.

Continuous highway overpass structures are often governed by serviceability rather than ultimate conditions. Deflection prediction and control is vital to avoid cracking. A two span overpass in Calgary was chosen as a case study. Deflections and strains in two precast prestressed girders were monitored from fabrication to erection, and a comprehensive material testing program was done on the concrete mix. The results of the case study show that the CEB MC-90 model code underestimated the time-dependent response by a maximum of 16% while ACI 209 overestimated by 19%. By tuning ACI 209 and CEB MC-90 to the concrete material testing data, predictions were increased to within 8% and 7%, respectively. A variability analysis on the two tuned models showed that while they give nearly the same prediction, the CEB MC-90 format induces less uncertainty in predictions. In addition, extrapolation to long-term ages shows a substantial divergence between predictions of the two models.