Investigating design methodology for a highperformance concrete bridge

For its significant economical savings and greater design flexibility, high-performance concrete (HPC) is becoming more widely used in highway bridge structures. To implement more widespread use of HPC in Missouri, the Missouri Department of Transportation (MoDOT) co-sponsored a research study to investigate both the early-age and later-age performance of a widely used prestressed concrete bridge system in Missouri that includes the use of HPC and larger prestressing strands.

For prestressed concrete highway bridge girders, prestress losses are an important factor related to serviceability conditions. Prestress losses have a direct impact on concrete stress development and deflection behavior of the member.

A poor estimate of prestress losses can result in a structure where allowable stresses are exceeded or camber and deflection behavior is poorly predicted, such that the serviceability of a structure may be adversely impacted.

There are several sources of loss in prestressed concrete highway bridge girders. For pretensioned strands, these sources include strand relaxation, elastic shortening of concrete, thermal effect, creep, and shrinkage of concrete. Actual losses are difficult to predict and depend on a number of factors, including actual time elapsed, exposure conditions, and dimensions of the member.

To investigate prestress losses in the actual beam members, an instrumentation program was developed and implemented.

This article reports the prestress losses observed throughout fabrication, shipment, erection, and the first two years of service for the first HPC superstructure bridge constructed in Missouri.

Bridge description

Bridge A6130 was designed as a fivespan bridge on Route 412, spanning drainage ditch No. 8 in Pemiscot County located near Hayti, Mo. The span lengths of the bridge are 50.9 feet, 55.8 feet, 55.8 feet, 55.8 feet and 50.9 feet, respectively. This is the first bridge in Missouri that fully implemented HPC into the superstructure, including the girders and bridge deck.

Precast, prestressed beams were designed to incorporate high-strength concrete (HSC). The required 56-day design compressive strength was 70 mega Pascals (MPa)—approximately 10,160 pounds per square inch (psi)—with a required release strength of 52 MPa (7,550 psi). The design used 0.6- inch-diameter pretensioned strands to make full use of the HSC. All 20 main span girders used in the bridge are MoDOT type II girders. Dimensions of these girders vary slightly from standard AASHTO type II girders. Low-permeability HPC was used in the 9.1-inchthick cast-in-place deck. The abutment and bent lines are projected on a skew at an angle of 48 degrees.

Instrumentation program

An instrumentation program was developed to monitor components of the bridge superstructure during earlyage and later-ages to identify trends in the measured and observed behavior. A total of 16 internal thermocouples, 64 internal vibrating wire strain gauges (VWSG), and 14 internal-bonded electrical resistance strain gauges (ERSG) were embedded in the precast girders and cast-in-place deck. A data acquisition system (DAS) with sufficient channels was designed and assembled for the project. In total, six girders and four locations in the deck were instrumented.

VWSG and ERSG were embedded in six girders at mid-span section and near support section. Gauge locations along the height of the section include the top flange, top web, middle web, center gravity of I-section, center gravity of strands, and bottom flange.

Concrete temperatures were recorded by thermocouples and thermistors integrated within the VWSG. Temperature data were used to investigate thermal gradients, extreme seasonal bridge temperatures, hydration temperatures, and corrections for strain and deflection measurements.

A 445 kilo Newton (100 kips) capacity load cell was placed at the bulkhead when jacking. Forces in selected strands were measured from stressing to release for three of the concrete casting dates. All six instrumented girders were included in the three casting dates; two girders were cast each date. For each instrumented strand, the load cell was placed between the strand chuck and the bulkhead at the non-stressing end of the bed.

Measurements of prestress losses were indirect since changes in stress were not actually measured. Instead, changes in concrete strain at the level of the centroid of pretensioned strands were measured using embedded strain gauges. The results were converted to prestress losses by multiplying the modulus of elasticity of the prestressed steel. Modulus of elasticity of the prestressing steel was taken as 193,000 MPa (28,000 kips per square inch (ksi)), which is also the value provided by the prestressing steel manufacturer and used for design.

Prestress losses before release

Before release of the pretensioned strands, there are three main losses— relaxation of the strand, temperature effects, and concrete shrinkage. These losses directly affect the level of prestress applied to the member.

Measured losses obtained through the load cell with thermal correction ranged from 1.58 to 18.75 MPa (0.23 to 2.72 ksi), with an average of 8.34 MPa (1.21 ksi) or 0.60 percent of the nominal jacking stress. Calculated losses ranged from -4.62 to 11.58 MPa (-0.67 to 1.68 ksi), with an average of 0.28 MPa (1.93 ksi) or 0.14 percent of the nominal jacking stress. It also was noted that the initial measured stresses directly after the stressing operation were a little higher than required. The fabricator pulled strands individually, and the strand tension was determined by force and independently by measurement of elongation.

Initial measured stresses were higher than the nominal jacking stress by an average of 1.5 percent.

The prestress behavior after casting, but prior to release, is very complicated.

However, using a simplified analytical model utilizing a curve-matching process, the prestress losses during this period can be computed analytically.

For this project, the losses before release were very small, with an average of 1.95 percent of the nominal jacking stress. There are several reasons why the value is so low. First, the prestress strands were stressed at a higher air temperature of 80 to 90 degrees Fahrenheit, usually in the afternoon, but concrete was cast in the early morning when the air temperature was approximately 68 to 72 degrees Fahrenheit. In this way, the losses caused by strand relaxation and other factors were balanced by decreased temperature, and thus the losses locked in the member before casting were very small, about 0.6 percent of jacking stress. Second, the strands were all released in the morning with an air temperature very close to the air temperature when the concrete was cast. Thermal effects during this period therefore were small, as well. The strands were released shortly after the girder molds were removed before significant shrinkage occurred.

Compared with a study on Texan HPC bridges conducted by the University of Texas-Austin, the prestress losses before release in this study are much lower. The reason is that thermal effect contributed much more to the prestress losses before release in the Texan HPC bridges.More research is warranted on prestress losses prior to release to develop a simple approach for prestress loss estimates from jacking to release.

Elastic shortening at release

The importance of a better estimate of elastic shortening losses is widely recognized.

Significant research work has been completed on elastic shortening losses when using normal-strength concrete.

However, more research is warranted on the elastic shortening losses for members utilizing HPC and/or HSC. It should be noted that elastic shortening varies along the length of the member as a function of the self-weight moment and prestressing layout at each section.Technically, elastic shortening should be calculated at each critical section.

Elastic shortening losses are highly dependent on the modulus of elasticity of the concrete at release because losses are directly related to the stress in the concrete by the modular ratio. For all instrumented girders, strain measurements were made using VWSG embedded in the concrete at the level of the centroid of the prestressing strand.

In general, measured elastic shortening losses typically were higher than predicted. Restraint against shortening of the member prior to release affected the measurements and caused the measured losses to be augmented artificially.

Differences between the actual modulus of elasticity and the values obtained from tests on match-cured companion specimens also may have affected the measurements and require further study.

Predicted losses showed good correlation with measured losses. Approximate method using gross-section properties resulted in an acceptable estimation for losses at design stage. Though bridge designers normally use modulus of elasticity specified in AASHTO for normal-strength concrete, for this particular project the equation specified for HSC in ACI 363R-92 was used in design. It is thus recommended that the approximate method based on grosssection properties is permissible for the calculation of elastic shortening losses in HSC designs.

Total losses Long-term prestress losses were measured successfully using embedded VWSG. Pre-release losses (PR) were estimated as described above and added to the total measurements. Elasticshortening losses (ES) were discussed in the previous section. The relaxation losses (RE) were estimated analytically.

Time-dependent creep losses (CR) and shrinkage losses (SH) were measured for two years after release.

In the past decades, several different methods have been provided for calculating long-term prestress losses. The following nine methods were adopted for comparison:

• Incremental time-step with measured parameters (This method was developed by the authors using a computer spreadsheet program and used to predict the time-dependent prestress losses, camber, and deflection for all the instrumented girders. A typical time-step, long-term prestress losses after release is plotted and shown in Figure 1.);
• Actual girder design using BR200 by MoDOT;
• American Association of State Highway and Transportation Officials’ Load and Resistance Factor Design (AASHTO LRFD) time-dependent lump-sum;
• AASHTO LRFD components with design parameters including concrete strength, modulus of elasticity, and strand stress at different stages;
• AASHTO LRFD components with measured parameters;
• PCI Design Handbook with design parameters;
• PCI Design Handbook with measured parameters;
• Suggested Method by S. Gross with measured parameters done at the University of Texas-Austin; and
• National Cooperative Highway Research Program (NCHRP) Report 496 method with design parameters.

Measured and predicted total prestress losses including all components for a typical measured girder are shown in Figure 2. Total measured losses averaged 289.2 MPa (41.94 ksi), or 20.7 percent of the nominal jacking stress of 1,396 MPa (202.5 ksi). Elastic shortening accounted for the largest portion of the total loss up to 53.7 percent.Time-dependent loss due to creep and shrinkage was less than the elastic shortening for measured values in all monitored girders. It is important to notice that losses before release accounted for about 7 percent of the total measured losses. Measured losses for all four girders were very close.

From the time-history figure of measured losses (Figure 1), it can be seen that measured losses became stable one year later after release. Predicted timehistory of prestress losses correlated very well (4 to12 percent difference) with the measured values while using measured material properties in the time-step method. However, even though the total predicted losses were in the range of 4 to 10 percent greater than measured total losses, the predicted elastic shortening losses were 1 to 29 percent smaller than measured values, while losses due to creep and shrinkage were greater than measured data.

Prestress losses in actual design using program BR200 based on the AASHTO Standard Specification are much higher than measured total losses by 45 to 55 percent. The difference is caused mainly by estimated losses due to creep and shrinkage, while elastic shortening losses in design are very close to the measured values by using a match curing system.

Similarly, predicted losses by AASHTO LRFD are much higher than measured losses by 50 to 60 percent with overestimated losses due to creep and shrinkage.

Predicted long-term losses due to relaxation, creep, and shrinkage are even three times the corresponding measured losses.

However, predicted losses by AASHTO LRFD are close to the values in actual design. Using measured material properties resulted in a significantly better estimation than using designed material properties, as would be expected.

Predicted losses using the PCI Handbook method with design material properties are close to those computed using the AASHTO LRFD timedependent lump-sum. Both are higher than measured losses by 22 to 30 percent for different girders.The PCI Handbook method and AASHTO LRFD timedependent lump-sum method are better for design compared to the AASHTO Standard Specification method and AASHTO LRFD design methods.

Predicted losses using the PCI Handbook method with measured material properties are close to those computed by the time-step method and the method recommended by Gross.

Each of these methods is very close to the measured losses, with the difference ranging from 4 to 12 percent.However, it should be noted that in the measured losses, elastic shortening losses accounted for more than half of the total losses. As discussed in the previous section, the greater measured elastic shortening losses was due to the restraint from the casting bed before release. Therefore, the actual losses may be a little lower than the value computed from measured strain data.

Since the time-step method is based on the tested time-dependent material properties, which are not available at design stage, use of this method requires much more accurate empirical timedependent material properties. Even though pre-release losses are not estimated using the PCI method, it still yields a good estimation (less than 15 percent difference), as does the method recommended by Gross, which considered pre-release losses.

Based on the analysis above, for prestressed, precast HPC girders, PCI Handbook method, the method recommended by Gross, and NCHRP 496 method are recommended for prestress losses estimation in design stage. If the prestress losses before release could be considered and concrete material and mechanical properties can be estimated with a high-accuracy level in the design stage, these two methods are very reliable for calculation of exact losses, with less than 15 percent difference from measured values.

Conclusions

The following conclusions are based upon the two years of bridge monitoring.

The losses before release were very small, with an average of 1.95 percent of the nominal jacking stress as observed in this project. Approximate methods using gross-section properties resulted in an acceptable estimation for elastic shortening losses; therefore, we recommended that the approximate method be used for the calculation of elastic shortening losses in HSC designs. Total measured losses averaged 289.2 MPa (41.94 ksi), or 20.7 percent of the nominal jacking stress of 1,396 MPa (202.5 ksi); elastic shortening accounted for the largest portion of the total loss. For prestressed, precast HPC girders, the PCI Handbook method, the method recommended by Gross, and the NCHRP 496 method are recommended for prestress losses estimation in design stage. If the prestress losses before release could be considered and concrete material and construction mechanisms can be estimated in a high accuracy level in the design stage, these two methods are very reliable for calculation of exact losses, with less than 15 percent difference from measured values.

Yumin Yang, Ph.D., is a structural engineer at Metz & Associates, Inc., in Orlando, Fla. He can be reached at yyang@metzbridges.com. John Myers, Ph.D., P.E., is an assistant professor at the University of Missouri-Rolla.