Prestress Losses in High-Performance SCC Prestressed Bridge Girders

By Brian Kukay, Ph.D.; Paul J. Barr , Ph.D.; and Marvin W. Halling , Ph.D. | Photo and illustrations courtesy of Montana Tech

Self-consolidating concrete (SCC) requires no vibration and fills in voids under its own weight when placed. SCC’s primary benefits are its ease of placement and reduced labor costs. Other advantages of SCC include resistance to segregation and high deformability. High-performance concrete (HPC) provides enhanced performance properties such as increased strength and improved durability.

Research review

Hueste and Cuadros3 reported that 85% of the responding state Departments of Transportation (DOTs) used HPC for bridge construction with specified compressive strengths in the range of 6,000 psi to 8,000 psi (41 to 55 MPa). The benefits of HPC for precast/prestressed girders include: the use of fewer girders per span; longer spans; or girders with reduced  height where grade clearance is an issue. HPC can result in a savings of $235 per yard ($216 per meter) of bridge length when used in place of normal-strength concrete; this savings is based on a study of bridge spans ranging from 39.4 feet to 141 feet (12.0 to 43.0 meters).

Results of the Hueste and Cuadros3 survey indicated that over half of the responding DOTs had some concerns with the use of HPC; seven  respondents implemented some type of design modification when HPC was used in prestressed girders. The AASHTO LRFD-04 provisions developed for prestress losses in normal-strength concrete may not provide reliable estimates for high-strength concrete (HSC) bridge girders.5 Prior to the AASHTO LRFD-07 method, appreciable differences between predicted and measured behavior of precast/prestressed bridge girders were expected.

Ahlborn et al. found that design specifications overestimated the modulus of elasticity of HSC, resulting in underestimated elastic-shortening losses and over-prediction of the creep and shrinkage losses. Roller et al. found that high-strength, prestressed concrete girders perform adequately when designed according to the AASHTO Standard Specifications. In 2001, Kowalsky et al. indicated that total prestress losses ranged from 12.9% to 19.1% of the initial jacking stress for HPC prestressed girders. In 2008, Barr, Kukay, and Halling determined prestress losses on in-service members of a Washington state bridge to be 28% of the jacking stress. Other HPC bridge research may be found in Shams and Kahn,10 Lopez et al. and Waldron.

Bridge description

The bridge instrumented for this study is located in Logan Canyon near Logan, Utah. The bridge consists of two 89.2-foot long (27.2-meter) spans. The 8-inch-thick (200-mm) bridge deck was supported with six precast/prestressed concrete AASHTO Type IV girders spaced at 8.5 feet (2.6 meters) on-center (Figure 1).

The specified concrete release strength for the girders was 4,500 psi (31.0 MPa) with a 28-day strength of 5,500 psi (37.9 MPa). Each girder contained 32 harped prestressing strands (Figure 1). The center of gravity of these 0.5-inch-diameter (13-mm) prestressing strands was 5.2 inches (133 mm) from the girder bottom. The prestressing strands were stressed to an initial jacking stress of 202 ksi (1,396 MPa) and were harped at 0.4 times the span length. The girder was designed as simply supported for girder and deck self-weights and continuously supported for an HL-93 live load in  accordance with the AASHTO LRFD Specifications.

Specifications for the girders called for relatively lower strength concrete but the precaster elected to use a HSSCC mix for the girders. The average diameter of the self-consolidating slump test for the four girders was 21 inches (525 mm). The average compressive strength at release and at 28 days for the instrumented girders was 8,580 psi and 11.5 ksi (59.2 and 79.2 MPa), respectively. The concrete required no external vibration and was placed with minimal labor. As part of the Logan Canyon bridge superstructure, these girders were among the first to be constructed in Utah using HPSCC.

Results

Curing: Figure 2 presents the measured concrete temperatures for a typical girder during the casting and curing stages. Each girder was instrumented with four strain gauges with embedded thermistors. Gauges 1ABL and 1ABR reflects the strain and temperature gauges to the left and right sides at the centroid of the prestressing steel. Recordings for gauges 1AGC and 1ACC are measurements of the strain and temperature at the center of the girder web and near the top of the web, respectively.

The temperature readings over the first 20 hours reflect ambient conditions as the concrete was not placed during this time. The hydration process reached a maximum temperature of 151 F (66 C) at about 18 hours into the hydration process; this temperature was recorded with the thermistor gauge nearest to the top of the girder. An accelerated curing temperature of 149 F (65 C) is common among precast/prestressed plants (Mokhtarzadeh, et al.13). The temperature measurements returned to ambient conditions after about 80 hours (1.5 days after the girder was removed from the casting bed).

A maximum curing temperature differential over the height of the girder and between the gauges of Δ11 F (Δ6 C) was recorded (Figure 2). Therefore, it was reasonable to assume that a consistent concrete strength was achieved throughout the depth of the prestressed girder with respect to curing. In a related study,14 a temperature differential of Δ25 F (Δ14 C)
during the curing process was reported between match-cured cylinders and girders; the lower curing temperature of the cylinders in the related study was shown to underestimate the compressive strength of the girder concrete by as much as 10% (Roller et al.14).

Short-term losses: Measured values for the elastic-shortening losses are based on an average change in strain readings prior to cutting the first strand and immediately after cutting the last strand. An average change in strain (304 microstrain) results in an elastic-shortening loss of 8,660 psi (59.7 MPa) or 4.3% of the jacking stress. The entire destressing process was completed over 97 minutes and is shown graphically in Figure 3.

The measured losses associated with elastic shortening are compared to predictive values using the AASHTO LRFD-04 method and the AASHTO LRFD-07 methods. The AASHTO LRFD-04 methodologies for calculating prestress losses due to elastic shortening, creep and shrinkage were based on the observed behavior of conventional concrete with strengths typically below 6,000 psi (41.4 MPa).

The methodology presented in the more recent AASHTO LRFD-07 Specifications incorporates gross- and transformed-section properties of both the precast and composite sections for predicting prestress losses. The modulus of elasticity has been shown to be highly influential when  predicting prestress losses; in light of this, the AASHTO LRFD-07 methodology can account for regional differences in aggregate type. When specific regional information is not available, AASHTO recommends assuming a design value of unity for the modulus-of-elasticity parameters; this design assumption was used in the study reported herein.

As noted previously, while the girder concrete was specified as relatively lower-strength concrete, the precaster elected to use a HPSCC mix for the girders. Accordingly, when the design compressive strengths (4,500 psi [31.0 MPa] at release and 5,500 psi [37.9 MPa] at 28 days) were used to estimate elastic-shortening losses, both specifications were shown to overestimate the field measurements. On average, the AASHTO LRFD-04 method overestimated field measurements by 24% and the AASHTO LRFD-07 method overestimated field measurements by 33%. However, when the measured compressive strengths (8,580 psi [59.2 MPa]) were used to estimate the elastic-shortening losses, The AASHTO LRFD-04 method underestimated the average field measurements by 4% and the AASHTO LRFD-07 method overestimated the average measured values by 6%.

Long-term losses: After destressing, the prestress force continues to decrease as a result of concrete creep and shrinkage over time. Effects such as temperature loss and relaxation also occur but are not addressed in this study. Additionally, there is an increase in the prestress force during deck placement; this is a result of the strain in the prestressing strand increasing when the deck self-weight is applied. Because total strain is measured in this study, the effects due to creep and shrinkage cannot be easily isolated in  field measurements; consequently, these effects are presented in terms of a total long-term prestress loss. Figure 4 shows a comparison of measured and AASHTO-calculated prestress losses over time. The measured losses are presented jointly with the total calculated prestress losses according to the AASHTO LRFD-04 and AASHTO LRFD-07 methodologies. For this study, the actual and specified compressive strengths at release were used when calculating the specification-based predictions. In all, 500 days worth of data were recorded and presented.

The initial jump in prestress loss observed in Figure 4 is due to elastic-shortening losses. Over time, the rate of prestress losses was initially large and then decreased until deck casting at Day 130, at which time the creep and shrinkage effects decreased. After deck casting, the magnitude of prestress loss increased until approximately Day 225, after which stress values leveled off; this increase in prestress losses after deck casting is presumably due to differential shrinkage and was consistent for all instrumented girders.

As shown in Figure 4, over all time periods, the calculated losses using the actual concrete compressive strengths  corresponded more closely to the measured results regardless of prediction method. The predicted losses, using the AASHTO LRFD-04, provide reasonable estimates of the stress losses through Day 75; after this point in time, the differences in the calculated and measured values continue to increase. At Day 500, prestress losses in measured values were 11.4% (23 ksi [160 MPa]) of the jacking  stress, on average. Using the AASHTO LRFD-04 method in conjunction with the actual compressive strengths for this girder, the estimated prestress loss at Day 500 was 15% of the jacking stress (30.5 ksi [210 MPa]). When  specified compressive strengths were used in conjunction with this method, the estimated prestress losses at Day 500 were determined to be about 16%
of the jacking stress to (33.2 ksi [229 MPa]).

In instances where the specified compressive strength of concrete is all that is known, both AASHTO methods yielded virtually identical results at Day 500. For this case, the measured prestress losses were overestimated by 21.5%, on average. However, these results reflect a special set of circumstances wherein the specified compressive strengths were for normal-strength concrete but where HPC was used instead.

The AASHTO LRFD-07 method more accurately predicted the residual prestressing force after the deck was cast when actual compressive strengths are used in the calculations; this method was developed based on material testing of HPC bridge girders. This method2 led to an  overestimation of various creep coefficients and separation difference of  about 4% of the jacking stress (8.0 ksi [55 MPa]) when compared to the plot using actual compressive strengths. In either instance, the results were shown to be conservative. Using the methodology outlined in the AASHTO LRFD-07 Specification in conjunction with the actual compressive strengths resulted in prestress losses at Day 500 that were about 12% of the jacking stress (23.8 ksi [164 MPa]). When the lower specified compressive strength was used in conjunction with this method, the estimated prestress losses at Day 500 grew to 16% of the jacking stress (33.0 ksi [227 MPa]). Individually, the measured prestress losses as of Day 500, expressed as a percentage of jacking stress were: 12% (24.4ksi [168 MPa]); 11% (22.2 ksi [153 MPa]); 11% (22.5 ksi [155.5 MPa]); and 11.6% (23.3 ksi [161 MPa]) for girders 1A, 2A, 1B and 2B, respectively.

Overall, the AASHTO LRFD-07 method was shown to produce the most realistic prediction of prestress losses over time when actual concrete compressive strengths were used. Up to and through the deck placement (Day 130), the calculated values were on average 13% larger than the measured values when actual concrete properties were used. Similarly, the AASHTO LRFD-04 method was shown to overestimate the average measured prestress losses by 6% using the actual compressive strengths. The prestress losses were overestimated by 17% with the AASHTO LRFD-04 method in conjunction with the specified concrete compressive strength.

In addition to the prestress losses, a gain in strand stress occurs at deck casting (Day 130). This gain in stress is a result of the prestressing steel elongation when the additional deck self-weight is applied. Figure 5 presents the measured changes in strand stress in comparison with the AASHTO  LRFD-07 method during the time the deck was placed (Days 123 through  137).

As shown in Figure 5, there is a reduction in the prestress losses at Day 130 due to deck casting, constituting a stress gain. Gains can also be noted when the future wearing surface and parapet are cast. Measured stress gains that resulted from the deck placement for the four instrumented girders were:
0.8% (1.64 ksi [11.3 MPa]); 0.8% (1.8 ksi [12.3 MPa]); 1.3% (2.8 ksi [19 MPa]); and 1.1% (2.1 ksi [14.6 MPa]) for girders 1A, 2A, 1B and 2B, respectively.

The AASHTO LRFD-07 method predicted a 0.5% stress gain relative to the jacking stress (1,000 psi [7.1 MPa]). Because the AASHTO-LRFD-07 method overestimated the creep, shrinkage, and relaxation prior to deck placement, the total stress loss was overestimated by approximately 1.5% of the jacking stress. Referring to Figure 5, it can be seen that this overestimation in the first 130 days of the bridge’s life resulted in very accurate prestress losses from days 130 through 500; namely, an average overestimation of 6% of the measured values.

Study conclusions

Four precast prestressed girders made with HPSCC were instrumented and monitored for prestress losses for 500 days after the time of casting. The observed values of prestress losses were compared with values calculated using the AASHTO LRFD-04 and the AASHTO LRFD-07 Specifications.

Excluding the elastic-shortening losses, ratios that reflect the overall averages are presented in Figure 6 for both the AASHTO LRFD-07 and the AASHTO LRFD-04 methods. The measured results shown in Figure 6 were calculated by averaging the data collected on all girders together, over the 500 days. By using the AASHTO LRFD-07 method in conjunction with actual compressive strengths, the measured values were on average 91.3% of calculated values.

When elastic-shortening losses were included, the AASHTO LRFD-07 method overestimated the total average prestress losses by 8.7%. When specified compressive strengths were used, the AASHTO LRFD-07 method was shown to overestimate the measured values by 33%. Using the AASHTO LRFD-04 Specifications and the actual compressive strengths, the average measured values were overestimated by 12.4%, compared to an average overestimate of 21.6% when specified values were used.

Based on these results, the methodology in the AASHTO LRFD-07 Specifications is shown to more accurately predict prestress losses for the instrumented girders than the 2004 AASHTO method. Overall trends indicate that the 2007 AASHTO Specification is conservative and that predicted values converge toward measured values after deck placement. The measured prestress losses for all four instrumented girders were predominantly overestimated during the first phase of bridge construction. Significant findings are as follows:

1. After 500 days worth of monitoring, the average measured prestress losses for each of the four instrumented girders is about 21 ksi [144 MPa], which corresponds to a total loss of 10.3% of the jacking stress. This relatively low percentage of total loss is due to a significantly higher actual concrete strength that was required for the design.
2. In terms of predicted elastic-shortening losses and using measured compressive strengths, the AASHTO LRFD-07 predictions overestimated the average measured values by 6% whereas the AASHTO LRFD-04 predictions underestimated the measured values by 4%. Using specified compressive strengths, averages indicate measured elastic-shortening losses were overestimated by 33% and 24%, using the AASHTO LRFD-07 and AASHTO LRFD-04 Specifications, respectively.
3. When the measured compressive strengths were used to predict long-term losses (excluding elastic-shortening losses), measured losses were on average 91.3% of the AASHTO LRFD-07 predictions and 87.6% of the AASHTO LRFD-04 predictions (both methods were conservative). Using specified compressive strengths, measured losses were 67.3% of the AASHTO LRFD-07 predictions and over-predicted by 78.4% of AASHTO LRFD-04 predictions (both methods were conservative). Overall, the AASHTO LRFD-07 Specification more closely approximated the measured values when actual compressive strengths were used.
4. At the time the deck was placed, an average gain of 1% of the jacking stress (2,100 psi [14.5 MPa]) was recorded for tendon stress; this compares to a gain of 0.5% using the AASHTO LRFD-07 method.
5. With HPSCC, compressive strengths were observed to average 11.5 ksi [79.3 MPa] at 28 days. The use of HPSCC in prestressed bridge girders saved time and money as no vibration is required during fabrication and the total number of bridge girders may be reduced.

Acknowledgements

The authors gratefully acknowledge the Utah Department of Transportation for its financial support of this study. Appreciation is also expressed to Hanson Eagle Precast of Salt Lake City, and to Ralph L. Wadsworth Construction of Draper, Utah.

Brian Kukay, Ph.D., is an assistant professor in the Department of General Engineering at Montana Tech, Butte, Mont. His research interests include  the development and testing of destructive and nondestructive techniques for concrete and wood members.

Paul Barr, Ph.D., is an associate professor in the Department of Civil and Environmental Engineering at Utah State University, Logan, Utah. His research interests include the nondestructive evaluation of bridges due to earthquake loads, live loads, changes in temperature and prestress losses.

Marvin Halling, Ph.D., is an associate professor in the Department of Civil and Environmental Engineering at Utah State University, Logan, Utah. His research interests include structural health monitoring and damage detection of bridges.