High-Performance/High-Strength Lightweight Concrete for Bridge Girders and Decks (2013)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

10 3.1 Mix Designs and Material Properties Research Approach The mix design and material property portion of the project consisted of two phases. The first phase focused on selecting lightweight aggregate and developing concrete mix designs for use in bridge decks and girders. The second phase involved the determination of the material properties of the mix designs selected for use on a production basis. A descrip- tion of this work is given below and is followed by discus- sions of the materials used in the concrete mixtures, concrete mixing procedures, and fresh and hardened concrete test pro- cedures. Further details of the mix design and material prop- erties investigation can be found in Attachment F (available by searching for NCHRP Report 733 on the TRB website). Laboratory Material and Mixtures Screening Tests To facilitate the selection of aggregate, manufacturers of structural-grade lightweight aggregate were contacted and test data, typical mixture designs, and other information were acquired from each of the production facilities. Structural grade aggregates are those used in ready-mix or precast, prestressed concrete applications. Aggregate used primarily in concrete masonry units (CMUs) was not considered to be structural aggregate. A database of material properties (including elas- tic modulus, unit weight, specific gravity, absorption rates, and angularity of the aggregate) was compiled for lightweight aggregate sources in the United States. Based on data received, the highest documented compres- sive strength for each concrete mixture and the correspond- ing unit weight together with the highest aggregate absorption recorded in laboratory tests were determined for each aggre- gate source. Several aggregate sources were chosen to provide geographic distribution, to include at least one source from each of the three material types (clay [CL], shale [SH], and slate [SL]), and to include sources believed to be capable of achiev- ing the strength requirements for this research. Six candidate aggregates were selected for testing in this project: • CL1 - expanded clay source one • CL2 - expanded clay source two • SH1 - expanded shale source one • SH2 - expanded shale source two • SH3 - expanded shale source three • SL1 - expanded slate source one Concrete mixtures using each of these six aggregates were developed. Mixture proportions were designed for (1) water- cementitious material ratios (w/cm) of 0.30, representing high- strength, low-permeability beam mixtures; (2) w/cm of 0.40, representing moderate-strength and moderate-permeability beam mixtures; and (3) good quality concrete deck mixtures. Mineral admixtures used included fly ash (FA), slag, and silica fume (SF), which were used at typical rates to produce high- performance concrete mixtures. Screening tests were executed on these concrete mixtures to determine basic material prop- erties. Testing considered ¾ in. and ½ in. nominal coarse aggregate sizes, depending on availability. Table 1 presents the cementitious materials used in making concrete batches with aggregate from each source. Specimens from each screening batch were subjected to the tests listed in Table 2. Based on the results of these tests, mixtures for use in structural testing were identified. The selected mixture designs for deck concrete had a target compressive strength ( f ′c) of 4 ksi. The mixtures selected for girder concrete included one typical of present practice (f ′c = 8 ksi) with two different aggregates and one having a higher strength than typically used in bridges (f ′c = 10 ksi). The maxi- mum design strength that appears achievable on a production basis was 10 ksi. A typical normal weight high-performance concrete (NWHPC) mixture was used to cast control specimens C h a p t e r 3 Background and Research Approach

11 for structural test comparisons. Table 3 lists the types of con- crete mixes selected for use in the structural testing portion of this project. Based on the screening test results, two lightweight aggregates were selected for further use. These two aggregates were the shale aggregate from source three (SH3) and the slate aggregate source from source one (SL1). Production Material Property Verification Tests During casting of the beams and interface shear and full- size test specimens, samples of the concrete were obtained for material property tests. For these samples, either a compre- hensive test regime or a basic test regime was conducted. For the comprehensive test regime, in addition to the material properties performed in the screening phase, tests were per- formed to characterize time-dependent parameters that affect prestress loss (e.g., shrinkage) and bulk diffusion characteris- tics that influence durability. During fabrication at the precast plant, companion specimens were obtained from the con- crete used in the precast girders. Additionally, specimens were obtained from the ready-mix concrete used in the lab-cast beams, interface shear specimens, and decks. The test matrix is presented in Table 4. The basic test regime typically included fresh concrete tests (slump, air content, and unit weight) as well as compressive strength and modulus of elasticity. Mixture Description Number of Batches from Each Coarse Aggregate Type w/cm Binder 0.3 OPC 2 OPC+20%FA 2 OPC+40% SLAG 2 OPC+7%SF 2 0.4 OPC 2 OPC+20%FA 2 OPC+40% SLAG 2 Table 1. Cementitious materials for screening test batches. Material Property AASHTO Test Method Number of Samples Age (days) 7 28 56 90 Compressive Strength /Modulus of Elasticity T 22 2 3 3 2 Modulus of Rupture T 97 - 3 - - Split Tensile Strength T 198 2 3 3 2 Shrinkage T 160 3 - - - Permeability / Absorption T 277 - 2 2 - Table 2. Number of samples for material screening tests. Mix Designation* Aggregate Source Specimens Bridge Use NWHPC 1 (8 ksi) normal weight aggregate Lab cast, full size, interface shear girders NWHPC 2 (4ksi) normal weight aggregate Interface shear deck LWHPC 1 (8 ksi) SL1 Lab cast, full size, interface shear girders LWHPC 2 (8 ksi) SH3 Lab cast girders LWHPC 3 (10 ksi) SL1 Lab cast & full size girders LWHPC 4 (4 ksi) SL1 Full size & interface shear deck *Target f’c is shown in parentheses. Table 3. Concrete mixtures for structural testing. Property AASHTO Test Method Number of Specimens Age (days) 1 7 28 56 90 Compressive Strength / Modulus of Elasticity T 22 - 2 3 3 2 Modulus of Rupture T 97 - - 3 - - Split Tensile Strength T 198 - 2 3 3 2 Shrinkage T 160 3 - - - - Permeability / Absorption T 277 - - 2 2 - Freeze-thaw T 161 Proc. A - - 2 - - Table 4. Material property tests for full-scale production.

12 Relative Density (Specific Gravity) and Absorption of Coarse Aggregate. Although the test is specified only for normal weight aggregate, it was chosen because no standard exists for spe- cifically testing lightweight aggregate. Table 6 provides the results of specific gravity and absorption tests for the six aggregates. Fine Aggregate. Fine aggregate used throughout this research was obtained from a local ready-mixed concrete supplier. The source of the fine aggregate is an area known as Curles Neck, a natural deposit along the James River near Richmond, VA. Cement. Laboratory concrete was made using a Type I/II portland cement conforming to AASHTO M 85 (Standard Specification for Portland Cement). Slag Cement. The slag cement used for this project con- formed to the requirements of ASTM C 989, Standard Speci- fication for Slag Cement for Use in Concrete and Mortars. Fly Ash. Class F fly ash conforming to the requirements of ASTM C618, Standard Specification for Coal Fly Ash and Raw or Calcined Natural Pozzolan for Use in Concrete, was used in this project because of its pozzolanic reactivity with portland cement, as well as a relatively low calcium oxide (CaO) content. Silica Fume. Silica fume conforming to the require- ments of ASTM C1240, Standard Specification for Silica Fume Used in Cementitious Mixtures was used. Admixtures. Air-entraining admixture (AEA) and high- range water-reducing admixture (HRWR) were used. Mix Designs and Mixing Procedures Several concrete mixtures were designed to investigate use of different lightweight aggregates for producing concrete mix- tures appropriate for bridge girders (mixtures with a 0.30 w/cm ratio) and for decks (mixtures with 0.40 w/cm ratio). In developing the mixture designs, compressive strength and unit weight took precedence over other material proper- ties such as tensile strength, modulus, shrinkage, and perme- Creep characteristics were measured for a number of con- crete mix designs used during full-scale specimen production. Table 5 presents the number of creep tests performed on each test group. Each test group consisted of three sets of three concrete cylinders and one control cylinder. Materials The following materials were used for preparing the dif- ferent concrete mixtures used in the tests. Coarse Aggregate. Lightweight aggregates were chosen to represent a broad sample set from across the United States. Raw materials used to manufacture these expanded lightweight aggregates included shale, clay, and slate products. Six light- weight aggregates were selected which included three expanded shale (SH1, SH2, and SH3), two expanded clay (CL1 and CL2), and one expanded slate (SL1) (see Figure 1). After receiving the lightweight aggregates from the manu- facturers, a series of tests were conducted to confirm the physi- cal properties reported by the manufacturer of each aggregate. The tests conducted included gradation, specific gravity, and absorption. A gradation of ½ in. × No. 4 was required for this study, and deviations were recorded. If the nominal maximum aggregate size (NMAS) is taken as the first sieve which retains 5–15% of the material in a standard gradation test, then sam- ples from CL2 and SH2 were closer to a ³⁄8-inch NMAS. SL1 had the most retained on the ½-inch sieve at 30%. The specific gravity and absorption tests were conducted according to ASTM C127, Standard Test Method for Density, Mix Designs Use Number of Tests First Test Group Second Test Group Third Test Group Fourth Test Group NWHPC1 Girder 1 1 - 1 LWHPC 1 - 1 - - LWHPC 2 1 - - - LWHPC 3 1 1 - - LWHPC 4 Deck - - 3 2 Table 5. Creep tests for full-scale specimen production. Figure 1. Lightweight aggregates (from left to right): CL1, SH1, SH2, CL2, SH3, and SL1.

13 Test Results Concrete batches were mixed in the laboratory to evaluate the performance of the six aggregate sources for a range of mixture designs. Based on results of the screening tests, mix- tures were identified for use in large-scale testing in labora- tory and lab-cast beams, full-size girders, and deck segments. Three lightweight mixtures represented moderate- and high- strength lightweight girder mixtures and a lightweight deck mixture. In addition, two lightweight aggregates (SH3 and SL1) were selected for use in the large-scale test specimens. These aggregates, when used in laboratory mixtures, yielded the properties needed for structural concretes. A mixture using normal weight aggregate was also tested for compari- son with the lightweight mixtures. Cylinders and prisms were cast from these production mixtures to characterize the materials used in the structural test specimens. Table 7 lists the structural test specimens as well as concrete type, cast- ing location, and test regime (comprehensive and basic test regimes are labeled C and B, respectively). Table 8 provides ability. One goal of the research is to recommend changes to the AASHTO LRFD Bridge Design Specifications and the AASHTO LRFD Bridge Construction Specifications relevant to high-strength lightweight concrete girders and high- performance lightweight concrete decks. To address this goal, the research involved developing mix designs of lightweight concrete with design compressive strengths of 8 and 10 ksi for girders and 4 ksi for decks. The limiting factor on fresh concrete properties was that unit weight be no greater than 125 lb/ft3. The lightweight coarse aggregate was subjected to a 24-hour saturation soak and allowed to drain for an additional 24 hours to place the aggregate in a saturated surface-dry (SSD) condition, per standard industry practice. The CL1 aggregate, which contained a large quantity of fine material, required longer than 24 hours to drain. The stockpile of fine aggregate was somewhat moist, but a 500-g sample was oven dried and tested to determine the moisture state for each day of mixing and adjust the amount of mixing water to maintain the design w/cm ratio. Property Aggregate Designation CL1 CL2 SH1 SH2 SH3 SL1 Oven Dry Weight (g) 1850 2000 2240 2140 3040 2950 SSD Weight (g) 2120 2310 2560 2310 3500 3410 Weight in Water (g) 906 420 988 476 1510 1260 SG (Oven Dry) 1.51 1.06 1.43 1.17 1.53 1.37 SG (SSD) 1.74 1.22 1.63 1.26 1.76 1.59 SG (Apparent) 1.97 1.27 1.79 1.29 1.99 1.74 Absorption, % 15.1 15.4 14.3 7.9 15.1 15.4 Manufacturer SG (SSD) 1.75 1.25 1.73 1.20 1.55 1.52 Table 6. Specific gravity and absorption. Designation Location Specimen Type Bridge Use Test Regime NWHPC1 Lab 2 Lab cast beams Girders C LWHPC1 Lab 2 Lab cast beams Girders C LWHPC3 Lab 4 Lab cast beams Girders C NWHPC1 Lab 2 Lab cast beams & f h Girders C LWHPC2 Lab 2 Lab cast beams Girders C LWHPC1 Lab Interface shear Girders B LWHPC1 Plant 2 AASHTO II Girders C LWHPC1 Plant 1 PCBT-45 Girders C LWHPC3 Plant 2 PCBT-45 Girders C LWHPC4 Lab CIP Deck Deck C LWHPC4 Lab CIP Deck Deck C LWHPC4 Lab Interface shear Deck B LWHPC4 Lab Interface shear Deck B NWHPC2 Lab Interface shear Deck B LWHPC4 Lab CIP Deck Deck C LWHPC4 Lab CIP Deck Deck C LWHPC4 Lab CIP Deck Deck B NWHPC1 Plant 1 PCBT-45 Girders B LWHPC4 Lab CIP Deck Deck B Table 7. Concrete mixtures for structural test specimens.

14 Compressive Strength. Identifying a binary lightweight concrete mixture that yielded a high compressive strength proved to be difficult. Table 9 lists the measured average 28-day compressive strengths for lightweight mixtures with different aggregates and combinations of binder propor- tions. All test cylinders were stored and tested according to the AASHTO standard test requirements and capped using sulfur compound. The table shows results for mix designs using 0.40, 0.30, and 0.25 w/cm ratios. All mixtures with 100% ordi- nary portland cement (OPC) and a 0.40 w/cm ratio performed similarly, regardless of the lightweight aggregate used. The the mix designs used in the lab-cast beams, full-size girders, and deck segments. Material Properties Equilibrium Unit Weight. Testing for equilibrium unit weight was conducted according to ASTM C567, Standard Test Method for Determining Density of Structural Light- weight Concrete, of all laboratory mixtures. The design fresh unit weight for all tests was 122 lb/ft3. Measured unit weight ranged between 119 and 125 lb/ft3. Ingredients* Concrete Mix Designations (Aggregate Type) & Use NWHPC1 (NW) NWHPC 2 (NW) LWHPC 1 (SL1) LWHPC 2 (SH3) LWHPC 3 (SL1) LWHPC 4 (SL1) One PCBT- 45 girder Lab cast beams & Interface shear specimens Interface shear AASHTO II girders & one PCBT- 45 girder Lab cast beams & Interface shear Lab cast beams Lab cast beams Two PCBT-45 girders Interface shear specimens & cast-in-place decks Cement 720 652 560 480 740 480 540 495 535 Fly ash 180 140 60 90 140 Microsilica 53 Slag 320 320 360 315 Sand 1,035 1,229 1,085 1,368 1,343 1,297 1,253 1,318 1,305 NW stone 1,683 1,700 1,746 LW stone 920 910 956 874 920 875 Water 275 247 283 232 240 242 268 246 304 *All quantities are in pounds. NW= Normal weight LW= Lightweight Table 8. Mix designs for production concrete. w/cm TCM* Agg. Source OPC OPC + 20% FA OPC + 40% Slag OPC + 7 % SF (lb/yd 3 ) (ksi) (ksi) (ksi) (ksi) 0.40 7 52 CL1 5.54 4 .61 4.58 CL2 6.17 5 .46 6.00 SH1 6.26 5 .48 6.18 SH2 6.10 5 .32 6.18 SH3 5.83 5 .82 5.93 SL1 7 .18 6.24 7 .03 0.30 8 00 CL1 6.21 6 .16 6.63 6.27 CL2 7.37 6 .92 7.58 7.43 SH1 7.50 7 .24 7.48 7.31 SH2 7.00 6 .59 7.33 7.12 SH3 8.45 7 .90 8.55 8.90 SL1 9 .00 8.50 9 .77 9.35 850 SL1 8.74 9 .38 9.43 900 SL1 8 .61 9.83 0.25 SL1 9 .87 *TCM = total cementitious materials (sum of OPC, FA, Slag, and SF). Table 9. Average compressive strengths of HPLWC laboratory mixtures.

15 compressive strengths ranged from 5 to 6 ksi. The strength of the other mixtures with a 0.40 w/cm ratio also ranged from 5 to 6 ksi range, but with greater variation. For each combina- tion of binder proportions, concrete made with SL1 light- weight aggregate provided higher strength than comparable mixtures containing other aggregates. Concrete with SL1 and SH3 lightweight aggregates and a 0.30 w/cm ratio consistently provided higher compressive strength. None of the lightweight concrete mixtures pro- vided the expected strength of 10 ksi. The mixture contain- ing 40% slag cement and 7% silica fume exhibited the highest compressive strengths. The mixture with SL1 and 40% slag cement had compressive strength exceeding 9.5 ksi at 28 days, meeting the 8.5 ksi design strength requirement. Increase in the total cementitious material from 800 lb to 850 lb only slightly increased the maximum compres- sive strength of one of the three mixtures using lightweight slate aggregate (SL1). Further, there was a smaller strength increase using 900 lb of total cementitious material. Consid- ering the harshness of the mix, wherein workability was dra- matically reduced, and the small strength gain, this mixture would not make a good choice for use in the field. Based on the compressive strength test results, the SH3 and SL1 lightweight aggregates were chosen for use in the struc- tural concrete mixtures. Modulus of Elasticity. Approximately 80% and 50% of the cylinders tested for compressive and splitting tensile strength, respectively, were previously tested for modulus of elasticity. AASHTO LRFD Bridge Design Specifications (Section 5.4.2.4) provide the following equation to predict modulus of elasticity based on compressive strength and unit weight of concrete: E K w fc c c= ′33 000 1 1 5, . where K1 = correction factor for source of aggregate (taken as 1.0 unless determined by physical test, and as approved) wc = unit weight of concrete (kcf) f ′c = specified compressive strength of concrete (ksi) Modulus of elasticity values for SH3 and SL1 lightweight aggregate concretes ranged from 4,000 to 5,000 ksi. One concrete mixture exceeded 5,000 ksi: a SL1 mixture con- tained 40% slag cement and 60% portland cement for a total cementitious material content of 850 lb. However, the modu- lus of elasticity for all SH3 mixtures with a w/cm ratio of 0.30 and 50 lb/yd3 less cementitious material was nearly 5,000 ksi. Figure 2 shows a comparison of the measured results to predicted modulus of elasticity based on the equation pro- vided in AASHTO LRFD specifications with K1 = 1.0. As noted above, the AASHTO equation for modulus of elasticity includes a factor to adjust for aggregate type, such as lightweight aggregate. This factor, K1, can be calibrated from a series of tests for a particular source aggregate; it was deter- mined for each lightweight aggregate source by least-squares fit of the predictive equation to the measured values. Table 10 lists the K1 factor determined for each aggregate source based on laboratory tests. Figure 3 shows the estimated modu- lus values versus the measured values using the K1 listed in Table 10. Note that a least-squares fit across all lightweight aggregate types resulted in a K1 value of 1.00. The accuracy of the modulus of elasticity prediction models for the production mixtures was investigated. A scatter plot was constructed with the experimental results and the associated values predicted by the AASHTO equation with K1 values of 1.0. As depicted in Figure 4, the AASHTO equation provided reasonable predictions for lightweight concrete deck mixtures. Splitting Tensile Strength. Although AASHTO specifica- tions do not provide an equation for splitting tensile strength Figure 2. Measured modulus of elasticity versus AASHTO prediction. Lightweight Aggregate Source K1 factor CL1 0.87 CL2 1.06 SH1 0.87 SH2 1.08 SH3 1.00 SL1 1.06 All Lightweight 1.00 Table 10. Modulus of elasticity factors for lightweight aggregates.

16 of concrete (fct), it can be determined from information given in Article 5.8.2.2. This article requires that ′f c be replaced with 4.7 fct in shear design calculations given in Articles 5.8.2 and 5.8.3 when fct is specified. In addition, 0.85 ′f c should be substituted for the ′f c term when fct is not specified. The factor 0.85 is required to be used for sand lightweight con­ crete. The first provision (fct specified) indicates a relationship between compressive strength and splitting tensile strength of lightweight concretes (without regard to fine aggregate type), capped at ′f c , consistent with the relationship assumed for normal weight concrete. The second provision (fct not speci­ fied), gives a discount factor for lightweight concrete, with adjustment for fine aggregate type. Using ′f c as the upper bound for splitting tensile strength similar to that of normal weight in the first expression, the expression was considered as an equality and the terms rearranged to yield the following equation for fct as a function of f ′c : f fct c= ′ 4 7. The relationship between splitting tensile strength and ′f c was determined by introducing the factor “a”, where fct = a ′f c . Figure 5 shows average values of factor a for a range of mixtures, all with total cementitious material (TCM) contents ranging from 752 to 900 lb/yd3. The values of a were slightly higher for 0.40 w/cm mixtures than for the 0.30 w/cm mixtures. The average a values for the lightweight mixtures ranged from about 0.23 to 0.27, with an overall average of 0.25, which is higher than the value of 0.21 derived from Article 5.8.2.2, suggesting the consideration for modifying Section 5.8.2.2 of the AASHTO specifications. Modulus of Rupture. For every concrete mixture, flex­ ural strength specimens were cast and tested at 28 days of age according to AASHTO T 97, Standard Test Method for Flexural Strength of Concrete (Using Simple Beam with Third- Point Loading), to determine modulus of rupture (fr). The measured values of modulus of rupture of lightweight concrete were compared to those predicted by the AASHTO equation relating flexural strength to compressive strength. According to AASHTO Section 5.4.2.6, for sand lightweight concrete, fr = 0.20 ′f c , where fr and f ′c are expressed in ksi. For normal weight concrete, the factors given in AASHTO Section 5.4.2.6 to be used before the radical are 0.20 for esti­ mating cracking moment and calculating Vci and 0.24 for cal­ culating deflections. A graph of measured results from this research and those predicted for lightweight concrete is given in Figure 6. Comparison of the predicted results for flexural strength with the measured results at 28 days showed the default factor of 0.20 significantly underestimates the flexural strength for sand lightweight. A statistical analysis of the data was used to determine the 95% prediction interval shown in Figure 6. The lower bound of the interval is approximately 0.26 ′f c , which is larger than the current AASHTO equation 0 20. ′( )f c . Modulus of rupture for the production mixtures were compared to predictions using the current AASHTO equation 0 20. ′( )f c as well as 0.26 ′f c . Figure 7 shows that modulus of rupture of production mixtures was significantly under­ predicted by the current AASHTO equation, as was seen with the laboratory mixtures. The sample size used for this com­ parison was very small. However, adjusted predictions using Figure 3. Measured modulus of elasticity versus AASHTO prediction using K1 from Table 10. Figure 4. Measured versus predicted modulus of elasticity for production mixtures.

17 0.15 0.20 0.25 0.30 0.35 0.40 0.30 w/c @ 800lb OPC 0.30 w/cm @ 800 lb OPC+FA 0.30 w/cm @ 800 lb OPC+Slag 0.30 w/cm @ 800 lb OPC+SF 0.40 w/c @ 752 lb OPC 0.40 w/cm @ 752 lb OPC+FA 0.40 w/cm @ 752 lb OPC+Slag 0.30 w/c @ 850lb OPC 0.30 w/cm @ 850 lb OPC+Slag 0.30 w/cm @ 850 lb OPC+SF 0.25 w/c @ 900 lb OPC 0.30 w/cm @ 900 lb OPC+FA 0.30 w/cm @ 900 lb OPC+Slag Fa ct or a Cementitious blend by w/cm and TCM CA Clay NY Shale CO Shale IN Shale LA Clay NC Slate Figure 5. Factor a (where fct = a fc′) at 28 days. 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 12.0 f'c (ksi) f r (ks i) Measured Upper Prediction Interval Least Squares Fit (0.306) Lower Prediction Interval Fitted Lower Bound (0.26) AASHTO Equation (0.20) AASHTO Prediction cr ff ′20.0= cr ff ′26.0= cr ff ′31.0= Figure 6. Flexural strength versus compressive strength at 28 days of age.

18 0.26 ′f c resulted in a closer yet conservative prediction of modulus of rupture. Permeability. Rapid chloride ion permeability tests were conducted on the top 2 in. of a 4 in.-diameter cylinder at both 28 and 56 days from casting. Two specimens from each mix- ture were tested, and their values were averaged and assigned a permeability rating according to AASHTO T 277, Standard Test Method for Electrical Indication of Concrete’s Ability to Resist Chloride Ion Penetration. The rating system is shown in Table 11. The use of any of the supplementary cementitious materi- als (fly ash, silica fume, slag cement) substantially improved the lightweight concrete’s ability to resist the ingress of chlo- ride ions. Additionally, the measured air content of the mix- tures seemed to be no indication of a more- or less-permeable material. Therefore, use of supplementary cementitious materials is appropriate for situations requiring a low- permeability concrete, such as bridge decks in regions receiv- ing deicing treatments in winter months or bridge piers in harsh environments. A representative sample of permeability test results for mixtures containing the SL1 aggregate is shown in Figure 8. Concrete made with cement and no supplementary cementi- tious materials consistently gave the highest permeability values based on charge passed. Permeability values for light- weight concrete containing portland cement only and SL1 aggregate ranged from 2200 and 2800 coulombs. Consider- ing all aggregate types, the only mixtures to rate “Moderate” or “High” were mixtures with 100% portland cement binder. The permeability of concretes containing a cement replace- ment (either silica fume, fly ash, or slag) were much lower than for mixtures with 100% portland cement. The permeabil- ity of lightweight mixtures with some supplementary cementi- tious material was less than 1000 coulombs, with the lowest values obtained for mixtures containing silica fume. Freezing and Thawing Resistance. Concrete exposed to a combination of saturating moisture and cycles of freezing and thawing in service may be susceptible to damage. The concrete mixtures used for full-scale girders and decks were subjected to rapid freezing and thawing in water in accor- dance with AASHTO T 161, Procedure A, except that a 5% NaCl solution was applied instead of tap water to simulate conditions of exposure common with bridge decks subject to snowfall, deicing, and subsequent freezing. As shown in Table 12, girder mixtures exhibited high resistance to freez- ing and thawing. The deck mixtures showed variable (but in some cases low) mass loss. Poor performance of a few deck mixtures is attributed primarily to the low air content in the mixtures. However, better performances were obtained for the girder mixtures. 0.00 0.20 0.40 0.60 0.80 1.00 0.00 0.20 0.40 0.60 0.80 1.00 Measured Modulus of Rupture (ksi) Pr ed ic te d M od ul us o f R up tu re (k si) beams beams (adj) decks decks (adj) Figure 7. Measured versus predicted modulus of rupture for production mixtures. Charge Passed (coulombs) Chloride Ion Penetrability > 4,000 High 2,000 - 4,000 Moderate 1,000 - 2,000 Low 100 - 1,000 Very Low < 100 Negligible Table 11. Chloride ion penetrability based on charge passed (AASHTO T 277).

19 Shrinkage. Shrinkage testing was performed in accor- dance with ASTM C157. The amount of shrinkage was mea- sured on two prisms with a steel stud cast into each end to serve as a gage point for length change measurements. Imme- diately after removing from the molds, the specimens were stored in a moist curing room for 28 days. The measured shrinkage values for the production con- crete mixtures were compared to those obtained from three prediction models: the AASHTO LRFD model contained in the AASHTO LRFD Bridge Design Specifications (Sec- tion 5.4.2.3), ACI 209, and CEB MC90. None of these mod- els accounted for supplementary cementitious materials (SCMs) or type of aggregate (i.e., lightweight or normal weight) used. Scatter plots were assembled to evaluate the applicability of the three models for lightweight concrete. Plotted points are for measured and predicted values at a given age, ranging from 1 to 448 days. Figure 9 shows measured versus predicted val- ues using the AASHTO method for all the production mix- tures, which were classified as normal weight girder mixtures, lightweight girder mixtures (w/cm = 0.30), or lightweight deck mixtures (w/cm = 0.40). The AASHTO model under-predicted the shrinkage strain in most lightweight deck mixtures, slightly over-predicted shrinkage in LWHPC girder mixtures, and closely predicted shrinkage of the normal weight girder mixtures. Figure 10 generally shows that the ACI 209 model over- predicted the shrinkage strain for the lightweight girder mix- tures, slightly under-predicted the shrinkage for the normal weight girders, and reasonably predicted shrinkage strains of lightweight deck mixtures. As shown in Figure 11, the CEB MC90 model typically under-predicted shrinkage, particularly for the lightweight 0 500 1000 1500 2000 2500 3000 800 850 900 Total Cementitious Material (lb/y3) Ch ar ge Pa ss ed (C ou lo m bs ) Cementitious materials: OPC only OPC + FA OPC + SF OPC + slag Figure 8. Permeability of concrete containing SL1 slag aggregate. Mixture Designation Aggregate Weight Loss Durability Factor Surface Rating (0 to 5) Unit Weight (lb/ft3) Air Content (%) LWHPC 1 (8ksi girder) SL1 0.24% 100 0.6 - 5.8 1.87% 100 0.7 - 4.8 0.29% 100 0.4 115.3 7.5 LWHPC 2 (8ksi girder) SH3 0.41% 100 0.6 118 8.0 LWHPC 3 (10ksi girder) SL1 0.47% 100 0.7 121.9 4.0 LWHPC 4 (4ksi deck) SL1 4.70% 100 1.1 120 4.5 22.72% 52 2.7 123.4 3.5 12.50% 100 1.3 121.6 3.3 25.47% 100 2.7 121.6 2.5 59.77% 0 4.9 122 2.5 NWHPC 1 (8ksi girder) NW 0.14% 83 0.3 145 4.0 0.07% 100 0.1 - - Table 12. Rapid freezing and thawing durability results.

20 NW Girder LW Girder LW Deck Figure 9. Measured versus predicted shrinkage strain (AASHTO model). Figure 10. Measured versus predicted shrinkage strain (ACI 209 model). -1200 -1000 -800 -600 -400 -200 0 -1200-1000-800-600-400-2000 Pr ed ic te d Sh rin ka ge S tra in Measured Shrinkage Strain NW Girder LW Girder LW Deck

21 deck and normal weight girder mixtures. At early ages, the CEB MC90 model generally over-predicted shrinkage for lightweight girder mixtures but accurately predicted shrinkage for lightweight deck mixtures. In general, the AASHTO models reasonably predicted shrinkage in lightweight girder mixtures. At later ages, the AASHTO model was the closest predictor of shrinkage for lightweight girder mixtures. For the three normal weight mixtures tested, no single model consistently predicted the shrinkage strain over the full range of testing ages, although the AASHTO model provided the best estimate. Based on these comparisons, use of the AASHTO model for predicting shrinkage of lightweight concrete is justified. Creep. Concrete materials used in full-scale girders or decks were tested to determine their creep characteristics in accordance with ASTM C512/C512M-10, Standard Test Method for Creep of Concrete in Compression. Specimens from girder mixtures were loaded at 24 hours after steam cur- ing, and spec imen deck mixtures were loaded after 28 days of moist curing. Unsealed specimens were loaded in sets of three in hybrid hydraulic/spring-loaded frames to approxi- mately 20% of ultimate compressive strength, well within the range of elastic behavior. Creep frames were placed in a chamber maintained at 73 ± 3°F and 50 ± 4% relative humid- ity. The observed values from creep testing were grouped by mixture class (NW Girder, LW Girder, or LW Deck) and plot- ted versus predicted values as a function of time from one of the applicable models. Plotted values represent creep at ages 1 day through the end of testing. A comparison of predicted to measured creep strains for the AASHTO, ACI 209, and CEB MC90 models are presented in Figures 12 through 14, respec- tively. These figures show that creep coefficients for normal weight and lightweight girders overall can be best predicted by the AASHTO model. The lightweight (LWHPC) deck mixtures exhibited considerable variation in creep coefficients that were best predicted by the ACI 209 model. However, creep behavior of the deck is secondary in importance to the girder creep. Summary Based on these tests, the following conclusions were made: • The SH3 lightweight aggregate and the SL1 lightweight aggregate were selected for use in the large-scale test speci- mens. When used in laboratory mixtures, these aggregates Figure 11. Measured versus predicted shrinkage strain (CEB MC90 model). -1200 -1000 -800 -600 -400 -200 0 -1200-1000-800-600-400-2000 Pr ed ic te d Sh rin ka ge S tra in Measured Shrinkage Strain NW Girder LW Girder LW Deck

22 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 0 0.5 1 1.5 2 2.5 3 3.5 Measured Creep Coefficient Pr ed ic te d Cr ee p Co ef fic ie nt NW Girder LW Girder LW Deck Figure 13. Measured versus predicted creep coefficients (ACI 209 model). 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 0 0.5 1 1.5 2 2.5 3 3.5 Measured Creep Coefficient Pr ed ic te d Cr ee p Co ef fic ie nt NW Girder LW Girder LW Deck Figure 12. Measured versus predicted creep coefficients (AASHTO model).

23 yielded test results consistent with what is needed for structural concretes. • Lightweight concrete with a compressive strength of 7000 psi and a unit weight less than 125 lb/ft3 can be produced with a 0.30 w/cm and 800 lb of cementitious material with expanded shales and slates. The slate (SL1) mixture consistently produced the highest strength concretes. • The AASHTO LRFD equation for modulus of elasticity with K1 = 1.0 is appropriate for lightweight aggregates. Predictions can be improved by calibrating the K1 value for each aggregate type. • The average splitting tensile strength of lightweight concrete mixtures exceeded ′ ′f fc c4 7 0 21. .or . • The average modulus of rupture of the lightweight con- crete can be expressed as 0.31 ′f c , with a lower bound of 0.26 ′f c . • The use of supplementary cementitious materials resulted in low-permeability lightweight concrete. The best perme- ability results were achieved by using silica fume as a sup- plementary cementitious material. • The AASHTO model for shrinkage reasonably predicted the shrinkage of lightweight concrete. • The AASHTO model for creep reasonably predicted the creep coefficients of the lightweight girder mixtures. The creep coefficients of the deck concrete mixtures were rea- sonably predicted by the ACI 209 model. 3.2 Interface Shear Strength Review of AASHTO LRFD Bridge Design Specifications (5th Edition) The design procedure described in Section 5.8.4 of AASHTO LRFD Bridge Design Specifications was examined. The design for horizontal shear (as found in the specifications) is based on the following equations: ≥ = φ V V V V ri ui ri ni where Vri = factored interface shear resistance Vui = factored interface shear due to applicable load combinations Vni = nominal interface shear resistance f = resistance factor for shear with f = 0.90 for normal weight concrete f = 0.70 for lightweight concrete 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 0 0.5 1 1.5 2 2.5 3 3.5 Measured Creep Coefficient Pr ed ic te d Cr ee p Co ef fic ie nt NW Girder LW Girder LW Deck Figure 14. Measured versus predicted creep coefficients (CEB MC90 model).

24 The nominal shear resistance of the interface plane is calculated as follows: V cA A f P minimumof A orni cv vf y c 1 c cv= + +( ) ≤ ′µ K f A2 cvK where Acv = area of concrete engaged in interface shear transfer c = cohesion factor µ = friction factor A vf = area of shear reinforcement crossing shear plane within the area Acv fy = yield stress of reinforcement (not to exceed 60 ksi) Pc = permanent net compressive force normal to the shear plane f ′c = 28-day compressive strength of weaker concrete K1 = fraction of concrete strength available to resist inter- face shear K2 = limiting interface shear resistance The equation is based on experimental data from normal weight concrete with strengths ranging from 2.5 to 16.5 ksi and lightweight concrete with compressive strengths rang- ing from 2.0 to 6.0 ksi. The yield strength of the shear rein- forcing steel is limited to 60 ksi because experimental data showed that the equation over-estimates the horizontal shear strength of the interface for a higher steel yield strength. The second term in the equation, µ, stems from a shear friction model. The net clamping force is the sum of the ten- sion in the reinforcing steel, Avffy, and the normal force, Pc, that applies compression to the interface. Multiplying the net clamping force by a friction coefficient term transforms the clamping force to a shear force at the interface. The first term of the equation, cAcv, represents the shear strength pro- vided by cohesion and aggregate interlock at the interface. This modification was made because the experimental data showed that cohesion and aggregate interlock affect interface shear strength. Values of c, µ, K1, and K2 depend on surface preparation and the type of concrete (lightweight or normal weight). The term K1f ′c A cv gives the force required to prevent shearing or crushing of aggregate along the shear plane while the term K2Acv accounts for the lack of test data above the K2 values. Section 5.8.4.3 of AASHTO LRFD Bridge Design Specifica- tions stipulates the following values for a cast-in-place slab on a clean concrete girder, free of laitance with the surface intentionally roughened to an amplitude of 0.25 in.: • c = 0.28 ksi • µ = 1.0 • K1 = 0.3 • K2 = 1.8 ksi for normal weight concrete • K2 = 1.3 ksi for lightweight concrete Test Specimens To investigate the horizontal shear strength of lightweight concrete at the interface of a precast girder and cast-in-place concrete deck, 36 push-off tests were performed. The test arrangement allowed application of the horizontal force directly to the interface of the deck and girder. The typical push-off test specimen is illustrated in Figure 15. Specimen Fabrication Twelve different specimen configurations were tested. These configurations included three combinations of slab and girder concrete with four amounts of horizontal shear reinforcement as shown in Table 13. Each of the twelve configurations was repeated three times to provide multiple test results to exam- ine the variability in horizontal shear strength of lightweight and normal weight concrete. The three combinations of slab and girder concrete were as follows: • Normal weight girder with a normal weight deck • Normal weight girder with a lightweight deck • Lightweight girder with a lightweight deck The four amounts of horizontal shear reinforcement tested were as follows: • No reinforcement: r (Avf/Acv) = 0.000 • Minimum horizontal shear reinforcement: r (Avf/Acv) = 0.001 • Maximum horizontal shear reinforcement: r (Avf/Acv) = 0.012 • Intermediate horizontal shear reinforcement: r (Avf/Acv) = 0.005 Thirty-six tests were performed with three tests for each combination of variables. The proportions for the girder and the deck concrete mixtures are listed in Table 8. The minimum and maximum ratios of interface shear rein- forcement are those identified in the AASHTO LRFD Bridge Design Specifications. Specimens with the minimum amount of shear reinforcement had one No. 4 stirrup placed at the center of the interface. For the maximum amount of shear reinforcement, five No. 6 stirrups were used. The specimens with the intermediate amount of horizontal shear reinforce- ment contained three No. 5 stirrups. The bottom half of the specimen was cast first, mimicking a precast girder. The surface of the girder was raked to an ampli- tude of 0.25 in. to improve bond with the cast-in-place deck concrete. The slump of the concrete used in the bottom half of the specimen was 4 to 6 in. Once all of the 36 girder specimens were cast, the deck specimens were then cast directly on the girders, simulating typical construction of a precast concrete girder with a cast-in-place deck.

25 Test Setup Figure 16 shows a typical test setup. The testing frame con- sisted of two steel bulkheads bolted to the floor. A 200-ton actuator was bolted to one of the bulkheads and used to apply a horizontal load to the specimen. The other bulkhead was used as an abutment to prevent the deck side from moving the test specimen, and the specimen was supported on four steel rollers. A normal force of 2.5 kips was placed on top of the speci- men to represent the dead weight on this section of a girder. This dead weight was estimated for an 8.5-in.-thick deck and 10-ft-girder spacing and was provided by two concrete blocks held together by a steel frame that rested on top of the test specimen. A load cell was horizontally aligned at the height of the girder and deck interface. The load cell was placed in front of the actuator and bore against an 8-in.-by-8-in. steel plate which rested flat against the back of the girder side of the specimen. The center of the steel plate was also aligned with the center of the load cell at the height of the interface. An 8-in.-by-8-in. steel plate was welded to the abutment on the deck side and also aligned with the center of the girder and deck interface. Instrumentation. A load cell was used to monitor the applied horizontal shear load. Two 2-in. linear variable dif- ferential transformers (LVDTs) were used to measure the slip of the girder side relative to the deck side. Strain gages were placed on the horizontal shear stirrups to measure the strain in the stirrups while the horizontal load was being applied. One strain gage was placed on each leg of the shear stirrup, on the side of the stirrup that would exhibit ten- sion during the test. Two strain gages were used for specimens with one No. 4 bar. Four strain gages were used for the speci- mens with three No. 5 and five No. 6 bars: two gages were placed on the first shear stirrup and two gages on the last stirrup. Testing Procedure. Each specimen was centered in the test frame and placed on the four steel rollers. The center of the specimen was lined up with the center of the actuator and abut- ment. The load cell was then centered between the actuator and the girder end of the specimen. The normal force blocks were placed on top of the specimen and centered above the area of the interface. The testing instruments were then zeroed and the horizontal load was applied. The horizontal force was slowly applied in 20-kip incre- ments. The specimen was examined at each increment to note any cracks that might have occurred. Loading contin- ued until peak load was reached and the specimen cracked along the interface, signifying the bond breaking between Side View Section A-A Figure 15. Push-off test specimen. Girder Concrete* Deck Concrete* ρiv = Avf /Acv NWHPC1 (8 ksi) NWHPC2 (4 ksi) 0 0.001 0.005 0.012 NWHPC1 (8 ksi) LWHPC4 (4 ksi) 0 0.001 0.005 0.012 LWHPC1 (8 ksi) LWHPC4 (4 ksi) 0 0.001 0.005 0.012 *Design f ′c is shown in parentheses Table 13. Test matrix for push-off specimens.

26 the specimens. The horizontal force continued to be applied until about an inch of slip occurred between the deck and girder. Test Results All push-off specimens behaved similarly at the begin- ning of testing. The load increased steadily with little slip between the two specimens until the bond failure load was reached. The bond failure load was defined as the load at which the cohesion at the interface was broken. The load in the actuator then decreased and leveled off at different loads, depending on the amount of shear reinforcement provided at the interface. Specimens without shear reinforcement maintained a small amount of resistance once the bond at the interface was broken. Figure 17 presents the designation used for each test and a key to decipher what each part of the label represents. Table 14 presents the results of the push-off tests. The per- cent yield strain in the reinforcement (ey) was taken as the recorded strain in the reinforcement at peak load divided by a yield strain of 0.00207 for reinforcing steel with a reported Figure 16. Push-off test setup. Figure 17. Specimen designation.

27 yield stress of 60 ksi and elastic modulus of 29,000 ksi. The recorded strain in the reinforcement was not pure axial strain but also incorporated bending. Specimens with one reinforc- ing bar had two strain gages, and the percent yield strain reported in the table is an average of the two gages. Speci- mens with three reinforcing bars (LL-3, NN-3, and NL-3 in Table 14) had a total of four strain gages. The upper value is the average of the two front strain gages (on the left in Figure 16) and the lower value is the average of the two back strain gages (on the right in Figure 16). Other information in the table includes A cv = area of concrete at the interface A vf = total area of reinforcement crossing the interface fy = reported yield stress of reinforcement Pc = permanent net compressive force normal to the shear plane ey = percent of yield strain in the reinforcement measured at peak load The results of the heavily reinforced specimens (riv = 0.012) are not reported. Because of the very large area of steel crossing the interface, the failure of the specimen occurred Specimen Designation Acv (in2) Avf (in2) fy (ksi) Pc (kips) f ′c deck (ksi) f ′c girder (ksi) εy (%) Peak Load (kips) LL-0-A 384 0 60 2.54 6.25 11.1 132 LL-0-B 384 0 60 2.54 6.25 11.1 140 LL-0-C 384 0 60 2.54 6.25 11.1 178 NN-0-A 384 0 60 2.54 6.15 7.78 153 NN-0-B 384 0 60 2.54 6.15 7.78 160 NN-0-C 384 0 60 2.54 6.15 7.78 156 NL-0-A 384 0 60 2.54 6.25 7.78 186 NL-0-B 384 0 60 2.54 6.25 7.78 131 NL-0-C 384 0 60 2.54 6.25 7.78 197 LL-1-A 384 0.4 60 2.54 6.25 11.1 71 422 LL-1-B 384 0.4 60 2.54 6.25 11.1 134 147 LL-1-C 384 0.4 60 2.54 6.25 11.1 121 188 NN-1-A 384 0.4 60 2.54 6.15 7.78 7 123 NN-1-B 384 0.4 60 2.54 6.15 7.78 5 141 NN-1-C 384 0.4 60 2.54 6.15 7.78 5 173 NL-1-A 384 0.4 60 2.54 6.25 7.78 205 169 NL-1-B 384 0.4 60 2.54 6.25 7.78 * 175 NL-1-C 384 0.4 60 2.54 6.25 7.78 10 183 LL-3-A 384 1.84 60 2.54 5.73 11.1 66 201 32 LL-3-B 384 1.84 60 2.54 5.73 11.1 95 223 32 LL-3-C 384 1.84 60 2.54 5.73 11.1 73 230 35 NN-3-A 384 1.84 60 2.54 6.15 7.78 * 195 12 NN-3-B 384 1.84 60 2.54 6.15 7.78 41 218 23 NN-3-C 384 1.84 60 2.54 6.15 7.78 * 229 * NL-3-A 384 1.84 60 2.54 5.73 7.78 68 242 25 NL-3-B 384 1.84 60 2.54 5.73 7.78 12 238 13 NL-3-C 384 1.84 60 2.54 5.73 7.78 88 183 82 *Strain gage damaged; no data available. Table 14. Results of push-off tests.

28 in the vertical leg of the deck side specimen and not at the interface. Typical Load Slip Behavior. All of the specimens exhib- ited similar pre-crack behavior in which little or no slip occurred as the load continuously increased until the bond between the concrete at the interface failed. The load that caused the interface crack was always the peak load. Once the interface bond failed, post-peak behavior varied with the amount of reinforcing steel across the interface. Figures 18 through 20 present typical load slip plots for specimens with no reinforcing, with one No. 4 stirrup, and with three No. 5 stirrups, respectively. In all cases, the load dropped consid- erably after the interface crack occurred. For the specimen with three No. 5 stirrups, the first drop in load was due to the interface crack, and the second drop was due to rupture of two of the bars. All specimens maintained a smaller load after considerable slip. The post-peak strength increased with increasing reinforcement across the interface. Failure Surfaces. The failure surface varied with the type of concrete constituting the specimens but generally was in Figure 18. Typical load versus slip behavior for specimens without shear reinforcement. Figure 19. Typical load versus slip behavior for specimens with one No. 4 stirrup.

29 or around the deck/girder interface. Figure 21 shows cracking in the interface region for the three concrete combinations of deck concrete (on the top) and girder concrete (on the bot- tom). The NW-NW crack was generally along the interface and relatively straight. The LW-NW failure surface was along the interface in some locations and through the lightweight concrete in others. The LW-LW failure surface was typically entirely through the deck-side concrete. Data Analysis Detailed information about each interface shear test, includ- ing load versus deflection plots, are provided in Attachment G. Comparison of Results to AASHTO LRFD. The current AASHTO LRFD (2010) requirements for nominal shear resis- tance at the interface of a precast concrete girder with a cast- in-place concrete deck are given in Section 5.8.4. Values for the cohesion factor (c) and friction factor (µ) are dependent on surface preparation and how the composite system is con- structed. AASHTO specifies for a cast-in-place concrete slab on a clean concrete surface raked to an amplitude of 0.25 in. A summary of the calculated and experimental horizon- tal shear resistance is provided in Table 15. The experimental strength divided by the calculated nominal strength is called the bias. A bias of 1.0 indicates the calculated value to be the same as the experimental value. As the value of the bias increases above 1.0, the level of conservativeness of the equation also increases. The coefficient of variation of the test data was also used to measure the dispersion of data. A lower coefficient of variation indicates a lower ratio of the standard deviation to the mean of the data. The bias was generally larger for the specimens with light- weight concrete (1.26 for LW-LW and 1.29 for LW-NW) than for normal weight (1.16). This is probably due to the light- weight aggregate absorption properties which enhance bond across the shear interface. However, there was greater scatter in the results with lightweight concrete as evidenced by the higher coefficient of variation (0.235 for LW-LW and 0.233 for LW-NW) than for normal weight (0.195). Also the bias decreased with increasing reinforcement. Test data for horizontal shear strength and those predicted by the AASHTO equation are presented in Figure 22. Data (for three NN specimens and one NL and LL specimen each) fell below those from the equation. The figure also shows the data using equations with reduction factors (f) of 0.9 and 0.7, which are the currently specified factors for shear for normal weight and lightweight concrete, respectively. With a f factor of 0.9, two specimens have strengths lower than predicted (one NN and one NL). A f factor of 0.7 provides much lower values than obtained from the tests for all specimens. Reliability Analysis. A reliability analysis was performed to examine the current resistance factor (f) used in AASHTO LRFD specifications for interface shear resistance in lightweight concrete. For this analysis, the shear resistance was considered a random variable and the uncertainty of the resistance esti- mate was separated into three categories: (1) a material factor that encompasses the uncertainty introduced due to variabil- ity in the materials used; (2) a fabrication factor representing Figure 20. Typical load versus slip behavior for specimens with three No. 5 stirrups.

30 variability in the fabrication of the specimens; and (3) a pro- fessional factor representing the uncertainty of the theoreti- cal model by representing variability of the ratio of tested values to calculated values. The material and fabrication fac- tors were taken from previous work (Nowak (1999) and Nowak and Rakoczy (2010)), but the professional factor was extracted from test data. The mean horizontal shear resistance could then be taken as the product of the nominal resistance multi- plied by these three factors. The current equation in the AASHTO LRFD specifications (2010) was used to calculate horizontal shear resistance at the interface. The professional factors, bias (lP) and coefficient of variation (CoVP) were calculated from the test results for each concrete combination listed in Table 14. The material and fabrication factors, lF, lM, CoVF, and CoVM were presented in previous research. Nowak and Szerszen (2003) provide information on the fabrication fac- tors for cast-in-place and plant-cast concrete. For horizontal shear strength, the width of the top flange and the area of the reinforcing steel are of greatest importance. A fabrication bias of lF = 1.0 and a coefficient of variation CoVF = 0.01 were selected. Nowak and Rakoczy (2010) presented compressive Normal weight – Normal weight Normal weight – Lightweight Lightweight-Lightweight Figure 21. Side view of typical failure cracks.

31 Specimen Designation LRFD Calculated V ni (kips) Maximum Test Load (kips) Mean Test Load (kips) Bias V test /V ni Mean Bias Overall Bias CoV Overall CoV LL-0-A 110 132 150 1.20 1.36 1.26 0.163 0.235 LL-0-B 1 40 1.28 LL-0-C 1 78 1.61 LL-1-A 134 242 192 1.81 1.43 0.249 LL-1-B 1 47 1.10 LL-1-C 1 88 1.40 LL-3-A 221 201 218 0.91 0.99 0 .070 LL-3-B 2 23 1.01 LL-3-C 2 30 1.04 NN-0-A 110 153 156 1.39 1.42 1.16 0.023 0.195 NN-0-B 160 1.46 NN-0-C 156 1.41 NN-1-A 134 123 146 0.92 1.09 0.172 NN-1-B 141 1.06 NN-1-C 173 1.29 NN-3-A 221 195 214 0.885 0.97 0.080 NN-3-B 218 0.99 NN-3-C 229 1.04 NL-0-A 110 186 172 1.69 1.56 1.29 0.206 0.233 NL-0-B 131 1.19 NL-0-C 197 1.79 NL-1-A 134 169 176 1.26 1.31 0.040 NL-1-B 175 1.30 NL-1-C 183 1.37 NL-3-A 221 242 221 1.10 1.00 0 .148 NL-3-B 238 1.08 NL-3-C 183 0.83 Table 15. Calculated horizontal shear resistance. 0 100 200 300 400 500 600 700 200150100500 300 350250 Sh ea r S tre ss (V / A cv ), p si Clamping Stress (Pc+Avffy), psi Lightweight-Lightweight Normal weight-Normal weight Normal weight-Lightweight (280Acv +1.0(Pc + Avffy))/Acv (0.9(280Acv +1.0(Pc + Avffy)))/Acv (0.7(280Acv +1.0(Pc + Avffy)))/Acv Figure 22. Shear stress using AASHTO equation with different F factors compared to test results.

32 strength data from over 8000 samples of lightweight and nor- mal weight concrete and determined the bias and coefficient of variation for a wide range of concrete strengths. For typi- cal deck concrete (4 ksi compressive strength), the factors for lightweight concrete are lM = 1.338 and CoVM = 0.123, and for normal weight the factors are lM = 1.213 and CoVM = 0.155. The information gathered from tests and the literature was used to calculate a resistance bias and coefficient of variation for lightweight/lightweight, normal weight/normal weight, and normal weight/lightweight combinations; a summary is given in Table 16. The limit state load combination for horizontal shear resis- tance at the interface of a bridge deck and bridge girder was considered to be the Strength I load combination. Using the information in Section 5.8.4.1 of the AASHTO LRFD Bridge Design Specifications and appropriate load factors, the hori- zontal shear resistance can be determined as follows: V DC DW LL IMnh ≥ + + +( )[ ]1 25 1 5 1 75. . . Φ where DC = dead load from weight of structural and non- structural components DW = dead load from the weight of the wearing surface LL + IM = live load and impact load from moving vehicles on the bridge Statistical parameters given by Nowak (1999) including a bias factor of 1.05 for both types of dead loads and 1.28 for the live load, and coefficients of variation of 0.1, 0.25, and 0.18 for DC, DW, and LL, respectively, were used. To analyze the limit state, different ratios of dead load to total load were considered. The mean load and standard deviation for each ratio was calculated using the load parameters. A reliability index (b) could then be calculated for each girder/deck concrete combination. Reliability indices describe a probability of failure (PF) as a function of the limit state function, the resistances, and the loads. The probability of failure is the probability that the resistance factor multiplied by the calculated resistance is less than the factored loads. A higher reliability index signifies a lower probability of failure (a b = 3.09 corresponds to a probability of failure of one in one thousand, and b = 3.71 corresponds to a probability of failure of one in ten thousand). Using the statistical parame- ters calculated for resistance and the statistical parameters for the load combination, a reliability index could be calculated using the following equation (Nowak and Szerszen, 2003): β µ µ σ σ= +( ) +( )R Q R2 Q2 1 2 where µR = mean of the resistance µQ = mean of the load combination sR = standard deviation of the resistance sQ = standard deviation of the load combination Reliability indices were calculated for an assumed resis- tance factor and multiple loading possibilities. The average b was then taken for multiple combinations of load varia- tions. Figure 23 shows b for normal weight and lightweight concrete. Note that for a given resistance factor, the light- weight-lightweight and normal weight-lightweight concrete combinations provide a higher reliability index than for the normal weight-normal weight concrete combination. The equations used in the analysis are further explained in Attachment G. Based on review of the results of nine tests for each con- crete combination, the lightweight resistance factors should be greater than 0.7 and thus closer to that for normal weight concrete. The tests also showed that although the coefficient of variation for lightweight was higher than for normal weight, the bias was also higher. The data from Nowak and Rakoczy (2010) also indicated that the bias of normal weight concrete compressive strength was lower than for lightweight concrete, and the coefficient of variation was larger. The combination of these factors suggests that an increased resistance factor for sections containing lightweight concrete is appropriate. Because there is no distinction between interface and beam shear resistance factors provided in the AASHTO LRFD specifications, changes to the shear resistance factors given in these specifications seem appropriate; recommendations will be given later in this report. Concrete Combination λF λM λP λR CoVF CoVM CoVP CoVR LL 1.00 1.338 1.26 1.69 0.01 0.123 0.25 0.28 NN 1.00 1.213 1.15 1.40 0.01 0.155 0.20 0.26 NL 1.00 1.338 1.29 1.73 0.01 0.123 0.23 0.26 Table 16. Bias and coefficients of variation.

33 Summary The interface shear tests revealed the following: • The bias of the measured shear strengths to the nominal shear strength computed with the AASHTO equation for a concrete deck placed on the top flange of a girder which has been intentionally roughened was 1.16 for N-N, 1.29 for L-N, and 1.26 for L-L. • The AASHTO LRFD equation for interface shear design is less conservative with increasing reinforcement ratios, which indicates the friction coefficient may be too high. • Based on a reliability analysis, lightweight concrete should have a higher strength reduction factor for interface shear than currently stipulated in specifications. Because AASHTO equations do not differentiate between different kinds of shear in reinforced concrete, changes to the shear strength reduction factors given in AASHTO need to be considered. 3.3 Laboratory Beam Test Results and Analysis Test beam specimens contained either 0.5 in. or 0.6 in. pre- stressing strands and had one of two T-shaped cross-sections, depending on the diameter of the prestressing strand being used. The cross-sections are shown in Figures 24 and 25 and section properties are listed in Table 17. The cross-sectional design provides (1) for large compressive stresses at the level of the prestressing strands and at the beam ends after release to allow easy measurement of transfer lengths and (2) for large tensile stresses in the prestressing strands at flexural failure. The beams were designed for stresses approximately equal to the allowable stress limits given in the AASHTO LRFD spec- ifications. As per AASHTO LRFD Article 5.11.4.2, the devel- opment length of prestressing strand depends on the tensile stress in the steel. The beams were designed for a tensile strain between 0.04 in./in. and 0.05 in./in. at flexural failure. This was accomplished by using the wide T-flanges, which result in a small neutral axis depth at flexural failure when compared to the overall depth of the beams, indicating large tensile strains in the steel. This also ensured that the beams would fail in a ductile manner. All beams were 24 feet in length. Concrete mix types and strand size used is given in Table 18 with mixture propor- tions given in Table 8. Overall, there were eight beams con- taining 0.5 in. strands and four beams containing 0.6 in. strands. Normal weight concrete beams were included to allow for comparison of results to those found for the light- weight concrete beams. Each beam specimen contained three straight, Grade 270, low-relaxation, fully bonded, pretensioned strands. These strands were spaced 2 inches on center and located at a distance of 2 inches on center from the bottom surface. The two outer strands were located at a distance of 2 inches on center from the vertical bulb surfaces. All strands contained an oily surface residue which was allowed to remain during fabrication so as to create a realistic worst-case scenario for prestress transfer and development length purposes. The strand was stored indoors to prevent corrosion. As shown in Figure 25, reinforcing steel was provided to meet anchorage requirements at transfer and shear requirements during flexural testing. All mild steel used in the beams was Grade 60. With this arrangement, a flexural or bond failure would occur prior to a shear failure. Figure 23. Reliability indices versus resistance factors.

34 b) Beams containing 0.6-in.-diameter strands. a) Beams containing 0.5-in.-diameter strands. Figure 24. Beam cross-sectionals. Figure 25. Reinforcement details.

35 The scheme used to identify the laboratory beams is illus- trated in Figure 26. The “Abutment” identifier specifies the closest adjacent abutment during casting. The live end of a beam refers to the end in which the strands were flame cut, while the dead end refers to that opposite the live end (closest to an abutment during casting). The beam side refers to the left or right side of the beam as seen when looking toward the beam dead end from the beam live end. The scheme can be used to identify a particular beam or a particular beam end by dropping the end components. For example, 1.LW1.5A refers to a single beam, while 1.LW1.5A.B refers to the dead end of that beam, and 1.LW1.5A.BL refers to the left-side of the dead end of that beam. Concrete and Strand Material Properties Tests were conducted on the concrete mixes and prestress- ing steel, but no material property tests were carried out on the mild reinforcing steel used. For analysis purposes, the mild reinforcing steel was assumed to have a modulus of elasticity of 29,000 ksi and a yield stress of 60 ksi. Test results for fresh concrete properties as well as compressive strength, modulus of elasticity, splitting tensile strength, and modulus of rupture are shown in Table 19. The 28-day compressive strengths for the LWHPC1, LWHPC2, and NWHPC1 mixes were less than the design f ′c of 8 ksi; however, the compressive strength for these mixes exceeded 8 ksi at 90 days. For the LWHPC3 mix, the design strength, f ′c of 10 ksi, was not obtained even after 90 days. This highlights the difficulty in achieving high compressive strengths of concrete containing lightweight aggregates in the field. Strand tension tests were performed on samples of the pre- stressing strand in accordance with ASTM A416. Six tests were performed for the 0.5 in. strand and four tension tests for the 0.6 in. strand. Average test results for modulus of elasticity (Eps), yield stress (fpy), and ultimate stress (fpu) were 27,500 ksi, 240 ksi, and 278 ksi, respectively. Also, direct pullout tests were performed on NASP Bond specimens to examine strand bond capability. All tested sam- ples for both strand sizes met the NASP requirements with measured slip values far less than expected at the required load values. Large Block Pullout Tests were also performed on strand samples according to procedures recommended by Logan (1997). More details of the NASP testing of strand samples are provided in Attachment H. Pullout blocks were fabricated using the NW1 concrete mix, and six 0.5 in. strand samples and four 0.6 in. strand samples were tested. All strand samples exceeded the minimum pullout load requirements proposed by Logan. Transfer Length The compressive strains at different locations along the beam length were measured at transfer and at 28 days after transfer using a digital DEMEC strain gage; a sample strain profile is shown in Figure 27. The measured strains on each side of the beam are shown in the figure along with the aver- age strain. The profile has been drawn assuming zero strains at the end of the beams. Transfer lengths were determined from strain profiles using a graphical procedure known as the 95% Average Maxi- mum Strain (AMS) method (Russell and Burns (1993)). The AMS method is the average concrete surface strain between the transfer lengths. The method was conceived as a way to prevent arbitrary interpretation of strain profile data when determining transfer lengths. Table 20 provides the measured transfer lengths for all beams. AMS significantly increased over time due to creep and shrinkage of the concrete. Article 5.11.4.1 of the AASHTO LRFD specifications requires that the transfer length of pretensioned concrete com- ponents be taken as 60 times the strand diameter. Ramirez and Property Beams containing 0.5-in.–dia. strands Beams containing 0.6-in.–dia. strands Ag (in2) 135 273 Ig (in4) 4537 15320 yb (in) 10.3 15.9 yt (in) 6.73 8.13 eg (in) 8.27 13.9 Sb (in3) 442 965 St (in3) 672 1884 Table 17. Beam cross-section properties. Pour Number Number of Beams Concrete Type Strand Size (in) 1 2 LWHPC1 0.5 2 NWHPC1 0.5 2 2 LWHPC3 0.5 2 LWHPC3 0.6 3 2 LWHPC2 0.5 2 NWHPC1 0.6 Table 18. Summary of beam specimens.

36 Russell (2008) have recently proposed changes to the AASHTO specifications based on research conducted in NCHRP Proj- ect 12-60 (2008) that showed a correlation between bond quality and the square root of the concrete strength. The pro- posed equation for transfer length, lt, is 1 120 40t b bd d= ′( ) ≥f ci where f ′ci is the concrete strength at transfer and db is the strand diameter. Figure 28 compares the transfer length mea- sured at transfer from this research with AASHTO require- ments and the proposed equation by Ramirez and Russell. In all cases the two methods provide an upper bound esti- mate of the transfer length in lightweight concrete beams. Figure 26. Beam identification diagram. Concrete Property Time after Casting Fresh 7 days 28 days 56 days 90 days NWHPC1 Slump (in) 5.5 Air (%) 5.0 Wt.(lb/ft3) 146 f c´ (ksi) 6.16 7.39 8.20 Ec (ksi) 4790 5720 5200 ft (ksi) 0.650 0.770 0.825 fr (ksi) 1.02 (2) LWHPC1 Slump (in.) 6.75 Air (ksi) 4.75 Wt. (ksi) 120 f c´ (ksi) 7.10 8.39 8.59 Ec (ksi) 3660 4120 3680 ft (ksi) 0.610 0.715 0.760 fr (ksi) 0.780 LWHPC3 Slump (in.) 10.0 Air (ksi) 8.0 Wt. (ksi) 115 f c´ (ksi) 5.02 7.74 8.07 8.31 Ec (ksi) 2920 3560 3580 3630 ft (ksi) 0.520 0.680 0.670 0.745 fr (ksi) 0.785 NWHPC1 Slump (in.) 6.0 Air (ksi) 4.5 Wt. (ksi) 147 f c´ (ksi) 6.20 7.97 8.54 8.43 8.29 Ec (ksi) 4570 5520 5290 5470 4360 ft (ksi) 0.670 0.735 0.820 0.780 0.796 fr (ksi) 0.885 LWHPC2 Slump (in.) 6.25 Air (ksi) 8.0 Wt. (ksi) 119 f c´ (ksi) 5.89 7.73 8.32 8.17 8.95 Ec (ksi) 3060 3420 3540 3740 3240 (2) ft (ksi) 0.620 0.665 0.715 0.660 0.766 (2) fr (ksi) 0.840 Note: (2) denotes a 2 specimen average. Table 19. Concrete material property test results.

37 Figure 27. Sample strain profile (2.LW3.5A at transfer). Beam ID Time Transfer Length (in.) AMS (µε) Dead End Live End 1.NW1.5A Transfer 11.5 18.5 453 28 Days 14.0 18.0 812 1.NW1.5B Transfer 12.0 20.0 457 28 Days 14.5 21.0 800 1.LW1.5A Transfer 12.0 20.5 502 28 Days 11.5 20.0 727 1.LW1.5B Transfer 12.0 21.0 503 28 days 12.0 19.0 733 2.LW3.5A Transfer 14.5 22.5 530 14 Days 14.0 23.0 667 28 Days 14.0 22.5 784 2.LW3.5B Transfer 14.5 21.0 537 14 Days 15.0 21.0 666 28 Days 16.5 21.5 797 2.LW3.6A Transfer 14.0 25.0 562 14 Days 15.5 23.5 687 28 Days 15.5 23.0 805 2.LW3.6B Transfer 16.0 23.5 551 14 Days 16.5 25.0 683 28 Days 16.0 24.5 790 3.NW1.6A Transfer 11.0 17.0 435 14 Days 12.0 18.0 543 28 Days 12.0 18.5 641 3.NW1.6B Transfer 12.0 17.0 417 14 Days 11.5 16.5 522 28 Days 12.0 16.5 610 3.LW2.5A Transfer 15.5 24.5 579 14 Days 15.0 21.5 665 28 Days 15.0 21.5 743 3.LW2.5B Transfer 15.0 22.5 565 14 Days 14.5 23.5 679 28 Days 14.0 21.5 754 Table 20. Measured transfer lengths.

38 Figure 29 compares these two calculation methods with the results from other research on lightweight concrete (Kolozs 2000; Meyer 2002; Nassar 2002; Zena 1996). All transfer lengths shown that are greater than either of the calculated values were measured on specimens containing strand with no test for bond quality. The research by Ramirez and Russell (2008) states that the difference in bond quality of strands produced by different manufacturers can be significant. Meyer conducted tests on strand quality using the large block pullout test suggested by Logan (1997). In addition, very scattered data was obtained from small rectangular cross- sections (Zena 1996). Regardless, comparison of the test results found in the literature with those from this research lead to the conclu- sion that these methods provide a reasonable upper bound for the transfer length of lightweight concrete strengths. When compared to the test for strand bond (based on the NASP test procedure) as proposed by Ramirez and Russell, Figure 28. Summary of transfer length results. Figure 29. Summary of reported HSLWC transfer length data.

39 the current specification provides conservative estimates of transfer length for HSLWC members. Development Length Development length of prestressing strand is the embed- ment length required to ensure a flexural failure mode. Arti- cle 5.11.4.2 of the AASHTO LRFD Specifications provides the following equation for the development length, ld, of pre- stressed components: l f f dd ps pe b≥ −( )( )κ 2 3 where fps = average stress in prestressing steel at nominal resis- tance (ksi) fpe = effective stress in the prestressing steel considering losses (ksi) db = the nominal strand diameter (in.) k = 1.6 for pretensioned members with depth > 24 in., 1.0 otherwise Ramirez and Russell (2008) proposed that the expression for development length be changed to l f d dd c b b= ′( ) ′( )  ≥120 100fci + 225 Development length was determined by testing each end of each beam in flexure using the test setup shown in Figure 30. An iterative test procedure was used with both flexural and bond failures resulting, and a range for development length was subsequently determined. Test results are shown in Table 21. The last column in Table 21 gives the observed type of fail- ure. Following are brief descriptions of each failure type: (1) Flexure—The beam failed in a flexural mode at a load greater than the AASHTO-predicted nominal moment capacity with no measured strand end slip greater than 0.01 in. (there were two types of flexural failure: concrete compression and strand rupture); (2) Bond Slip—The AASHTO moment capacity was not reached and at least one of the prestressing strands had an end slip greater than 0.01 in.; and (3) Hybrid—The AASHTO moment capacity was reached and at least one strand had an end slip greater than 0.1 in. Measured flexural bond length intervals and calculated development length intervals from flexural tests are shown in Table 22. The development length interval was conser- vatively calculated by adding the longest measured transfer length for a particular beam pair in question to the longest flexural bond length resulting in a bond failure (called Lower in Table 22) and the shortest flexural bond length, resulting in a flexural failure (called Upper in Table 22). Also shown are the development lengths calculated using the equations provided. For the AASHTO calculation, the effective pre- stress was calculated at the day of testing. It appears from the table that the lightweight concrete beams have slightly longer development lengths than normal weight beams. However, all development length upper bounds were substantially less than those calculated using the AASHTO Specifications as well as those calculated using the proposed Ramirez and Russell equation. In considering these results, it should also be recognized that the flexural test loading case is a worst- case scenario, as in general the loading results in a situation where the maximum moments occur at approximately the same location as the maximum shears. Despite this fact, many of the beams had to be loaded at extremely short flexural bond lengths to obtain bond slip failures. Summary The following observations can be made from the lab-cast beam tests: • The AASHTO transfer length requirement (found in Sec- tion 5.11.4.1) and the Ramirez and Russell (2008) equation for transfer length both provide a reasonable upper bound to the measured transfer lengths of lightweight and normal weight girders. • The AASHTO 5.11.4.2-1 equation and the Ramirez and Russell (2008) equation for development length both 6" 16'-0" 7'-6" LOAD CELL WIRE POT BEARING PAD LVDTs END SLIP LVDTs BEAM Figure 30. Development length test setup.

40 provide a reasonable upper bound to the measured devel- opment lengths of lightweight and normal weight girders. 3.4 Shear Performance of Full-Scale Beams Work Plan The investigators reviewed available literature to (1) iden- tify tests that had been conducted on shear strength of light- weight concrete beams and (2) develop a test program. Consideration was given to shear span-to-depth (a/dv) ratios, mild steel reinforcement, unit weight, and compres- sive strength. The testing program, presented in Figure 31, included full- scale prestressed girders with cast-in-place composite decks. All cast-in-place decks were made with lightweight concrete. One end of each beam was tested with a short shear span and the other a longer shear span. The results of these tests provided information on the differences in concrete contri- bution to shear strength between lightweight and normal weight concrete. Specimens T2.8.Typ and T2.8.Min were Beam Designation End Shear Span (in.) Embedment Length (in.) Transfer Length (in.) Flex. Bond Length (in.) Failure Mode 1.NW1.5A Dead 54 62 11.5 50.5 Flexure - Concrete Crushing Live 54 62 18.5 43.5 Flexure - Concrete Crushing 1.NW1.5B Dead 42 51 12.0 39.0 Flexure - Concrete Crushing Live 42 51 20.0 31.0 Flexure - Concrete Crushing 1.LW1.5B Dead 54 62 12.0 50.0 Flexure - Concrete Crushing Live 54 62 21.0 41.0 Flexure - Concrete Crushing 1.LW1.5A Live 42 51 20.5 30.5 Flexure - Concrete Crushing Dead 30 40 12.0 28.0 Hybrid 2.LW3.5B Dead 42 51 14.5 36.5 Flexure - Concrete Crushing Live 42 51 21.0 30.0 Bond Slip 2.LW3.5A Dead 36 45 14.5 30.5 Bond Slip Live 48 56 22.5 33.5 Flexure - Concrete Crushing 3.LW2.5A Dead 42 51 15.5 35.5 Flexure - Concrete Crushing Live 42 51 24.5 26.5 Flexure - Concrete Crushing 3.LW2.5B Dead 24 34 15.0 19.0 Bond Slip Live 36 45 22.5 22.5 Bond Slip 3.NW1.6B Live 72 79 17.0 62.0 Flexure - Concrete Crushing Dead 48 56 12.0 44.0 Flexure - Strand Rupture 3.NW1.6A Live 48 56 17.0 39.0 Hybrid Dead 36 45 11.0 34.0 Bond Slip 2.LW3.6B Live 72 79 23.5 55.5 Flexure - Concrete Crushing Dead 48 56 16.0 40.0 Bond Slip 2.LW3.6A Dead 54 62 14.0 48.0 Flexure - Concrete Crushing Live 60 68 25.0 43.0 Flexure - Concrete Crushing Table 21. Results of development length tests. Designation Longest Measured Transfer Length (in.) Measured Flexural Bond Length Range (in.) Development Interval (in.) Calculated Development Length (in.) AASHTO Ramirez & Russell Lower Upper Lower Upper 1.NW1.5 20 31.0 51 77 64 1.LW1.5 21 28.0 49 79 61 2.LW3.5 22.5 30.5 33.5 53 56 83 66 3.LW3.5 24.5 22.5 26.5 47 51 80 62 3.NW1.6 17 34.0 39.0 51 56 94 76 2.LW3.6 25 40.0 43.0 65 68 100 79 Table 22. Development length intervals.

9 spa. @ 9 in.5" Lspan BT.8.Typ ρ = 0.63%v5 11 spa. @ 10 in. ρ = 0.57%v 12 spa. @ 24 in. ρ = 0.24%v P 8 spa. @ 12 in. ρ = 0.48%v P 8'-6" 3'-0" 9 spa. @ 9 in. 5" ρ = 0.63%v 11 spa. @ 10 in. ρ = 0.57%v P 8 spa. @ 12 in. ρ = 0.48%v P 12'-9"3'-0" 5 spa. @ 2 / in.1 4 8 spa. @ 10 in.6" Lspan BT.10.Typ ρ = 0.57%v 7 9 spa. @ 12 in. ρ = 0.48%v 4 spa. @ 18 in. ρ = 0.32%v 7 spa. @ 15 in. ρ = 0.38%v 8 spa. @ 10 in. 6" ρ = 0.57%v 9 spa. @ 12 in. ρ = 0.48%v 7 spa. @ 15 in. ρ = 0.38%v 5 spa. @ 2 / in.1 4 PP 8'-2" 3'-0" P P 12'-2"3'-0" 5 spa. @ 2 / in.1 4 26 spa. @ 24 in.15 Lspan BT.10.Min ρ = 0.24%v 5 spa. @ 2 / in.1 4 PP 8'-9" 3'-0" P P 13'-6"3'-0" 5 spa. @ 2 / in.1 4 / "3 415" / "3 4 / "3 4 5 7 15 / "3 4 / "3 4 / "34 15" 10" Wide Bearing Pad5 spa. @ 2 / in.1 4 8" Wide Bearing Pad 8" Wide Bearing Pad 8" Wide Bearing Pad 10" Wide Bearing Pad 8" Wide Bearing Pad Figure 31. Loading parameters and stirrup spacing for large-scale girder testing. (continued on next page)

7 spa. @ 9 in. 7 spa. @ 10 in. 6 spa. @ 12 in. 15" 6"6" 4 spa. @ 2 / in.1 4 7 spa. @ 9 in.7 spa. @ 10 in.6 spa. @ 12 in. 15" 6"6" 4 spa. @ 2 / in.1 4 10" P P 14'-6"4'-9" P P 3'-0" 9'-11" ρ = 0.56%v ρ = 0.67%v ρ = 0.74%vρ = 0.56%vρ = 0.67%vρ = 0.74%v Lspan T2.8.Typ 5 spa. @ 15 in. 8 spa. @ 24 in. 7" Lspan BT.8N.Typ ρ = 0.38%v 4 spa. @ 2 / in.1 4 P P 3'-0"4'-10" 11 spa. @ 12 in. 9" ρ = 0.56%v PP 3'-1" 9'-3" 4 spa. @ 2 / in.1 4 11 spa. @ 12 in.9" ρ = 0.28%v Lspan T2.8.Min ρ = 0.56%v P P 3'-0"8'-8" PP 2'-11" 12'-9" 5 spa. @ 2 / in.1 4 8 5 spa. @ 2 / in.1 4 6 spa. @ 18 in. ρ = 0.32%v 5 spa. @ 15 in. 7" ρ = 0.38%v 6 spa. @ 18 in. ρ = 0.32%v 12 spa. @ 24 in. ρ = 0.24%v/ " 3 4 8 / "3 4 10" Wide Bearing Pad 10" Wide Bearing Pad 8" Wide Bearing Pad 8" Wide Bearing Pad 10" Wide Bearing Pad Figure 31. (Continued).

43 both AASHTO Type II beams whereas the other four gird- ers were the bulb tee shapes. The designation “Typ” identifies test specimen with a typical amount of shear reinforcement, and the designation “Min” refers to test specimens with the AASHTO LRFD minimum amount of shear reinforcement. The typical amount of reinforcement was determined for the two girder sizes by designing typical interior girders with a transverse girder spacing of 8 feet for each. The AASHTO Type II girder had a span length of 60 feet and the PCBT-45 girder had a span length of 85 feet. The minimum amount of reinforcement was determined as per AASHTO LRFD requirements. Full-Scale Girder Testing Material and Section Properties. The primary purpose of the full-scale girder tests was to compare shear strengths as calculated using the AASHTO LRFD Specifications to those obtained from the tests. Six girders were tested, with two tests conducted on each. Type and strength of concrete, the amount of shear reinforcement, and a/dv (where a is mea- sured from the center of bearing) were investigated in these tests. The compressive strength and modulus of elasticity were measured at the time of testing. Major properties for each of the twelve tests are listed in Table 23. Each test specimen was assigned a four-term designation. The first term indicated the girder size, either an AASHTO Type II beam (T2) or a PCBT-45 girder (BT), where the section properties are shown in Table 24 and basic geom- etries are shown in Figure 32. Detailed girder properties for these two classes of girders are provided in Attachment J. The second term gave the design compressive strength of the girder (either 8 ksi or 10 ksi). The third term described the amount of shear reinforcement (Typ for typical reinforce- ment or Min for minimum reinforcement, as prescribed by the AASHTO LRFD specifications). The last term indicated which end of the girder was being tested. All beams were constructed with lightweight concrete except for BT.8N.Typ, which was built with normal weight concrete (identified by the letter N). All six beams were designed for a typical span length for the given cross-section with the harping point moved closer to midspan than is common in practice. Since shear forces tend to dominate near the ends of the girders, the research- ers were able to take out the middle section of each girder, and hence, test shorter girders, without deviating from the research objectives. Fabrication drawings of the six full-size girders are given in Attachment J. The AASHTO Type II girders were 41 feet long, but con- tained shear reinforcement for a 60-ft-long beam. Six of the 24 (0.5-in.-diameter) strands in these beams were harped at midspan as shown in the girder plans found in Attach- ment J. Additionally, two of the straight strands ran along Test ID Concrete Type L (ft) a/dv Air Content (%) Slump (in.) wc (lb/ft3) f ′c (ksi) Ec (ksi) T2.8.Typ.1 LWHPC1 40.0 1.5 4.75 9.5 117 8.9 3610 T2.8.Typ.2 35.1 3.1 4.75 9.5 117 8.9 3610 T2.8.Min.1 40.0 1.5 4.75 9.5 117 8.9 3610 T2.8.Min.2 34.8 2.9 4.75 9.5 117 9.0 3610 BT.8.Typ.1 58.0 2.0 4.75 - 121 9.1 3650 BT.8.Typ.2 49.5 3.1 4.75 - 121 9.1 3660 BT.8N.Typ.1 NWHPC1 57.9 2.1 4.0 8 145 8.5 4820 BT.8N.Typ.2 49.0 3.0 5.0 7.75 145 8.6 4580 BT.10.Typ.1 LWHPC3 57.9 2.0 5.0 10.75 120 8.9 3910 BT.10.Typ.2 49.4 2.9 4.0 - 124 9.9 4060 BT.10.Min.1 58.0 2.1 4.0 - 124 9.7 4040 BT.10.Min.2 49.5 3.0 3.5 - 126 10.3 4140 Table 23. Design geometric and measured girder material properties. Property Girder Type AASHTO Type II PCBT-45 Depth (in.) 36 45 Web Width (in.) 6 7 Width of top Flange (in.) 12 47 Width of Bottom Flange (in.) 18 32 Area (in.2) 369 747 Centroid to Bottom (in.) 15.83 22.3 Moment of Inertia (in.4) 50,979 207,300 Table 24. Girder properties.

44 the top flange in order to reduce the stresses in the top and bottom flanges at release. Likewise, the PCBT-45 beams were 59 feet long, but had shear reinforcement based on an 85-ft- long span; six of the 34 (0.5-in.-diameter) strands were harped 2.5 ft away from the beam centerline. After steam curing at the precasting plant, the beams were transported to the laboratory. An 8-in.-deep cast-in-place deck was placed on top of each beam to simulate a deck-girder system in a bridge. Although the reinforcement was based on an 8-ft girder spacing, a 7-ft-wide deck was built due to labo- ratory constraints. Concrete properties for the six decks are listed in Table 25. Instrumentation. Prior to production at the precasting yard, electrical resistance strain gages were installed on the shear reinforcement in a pattern that minimized the number of gages needed for measuring the strains at the anticipated crack locations. A typical strain gage arrangement is shown in Figure 33. Strain versus applied load plots for each gage on each girder are given in Attachment I. 2" typ. typ. 2" 7'-0" 7'-0" 3'-9" 1" 8" 3'-0" 1" 8" AAS HTO Type II PCBT -45 Figure 32. Cross-sections at the ends of the two girder types being tested. Deck Designation Fresh Concrete Properties Hardened Concrete Properties (ksi) Property Result Property Days After Casting 1 7 28 Testing T2.8.Typ Air (%) — f 'c — — 5.59 Slump (in.) 6.5 Ec — — 3403 Unit Weight (lb/ft3) 118.4 fct — — 0.535 fr — — — — — — T2.8.Min Air (%) 4.5 f 'c 1.72 3.49 4.70 5.36 Slump (in.) 7.5 Ec 1980 2690 3080 3240 Unit Weight (lb/ft3) 120 fct 0.280 0.440 0.495 0.530 fr — — 0.640 — BT.8.Typ Air (%) 3.5 f 'c 2.64 4.07 5.45 6.14 Slump (in.) 6.25 Ec 3040 3175 3765 3825 Unit Weight (lb/ft3) 127.1 fct — 0.450 0.583 0.650 fr — — 0.780 — BT.8N.Typ Air (%) 2.63 f 'c 2.61 3.51 4.57 4.94 Slump (in.) 7.5 Ec 2640 2635 3000 3025 Unit Weight (lb/ft3) 122 fct — 0.430 0.483 0.535 fr — — 0.760 — BT.10.Typ Air (%) 2.38 f 'c 1.66 3.76 — 4.53 Slump (in.) 6.25 Ec 2300 2903 — 215 Unit Weight (lb/ft3) 122.8 fct — 0.490 — 0.470 fr — — 0.795 — BT.10.Min Air (%) 3.25 f 'c 1.40 3.55 4.93 5.52 Slump (in.) 7 Ec 1870 2610 3143 3305 Unit Weight (lb/ft3) 122.2 fct — 0.525 0.498 fr — — — 0.620 — Table 25. Properties of deck concrete.

45 Also, two vibrating wire gages (VWGs) were placed at the centroid of the bottom layer of flexural reinforcement at the beam centerline. These gages were primarily used to measure prestress losses and monitor the strain in the strand during testing. Various instruments were also deployed on the girder exte- riors. In most cases, sets of three LVDTs were placed on both sides of the web in a rosette pattern to measure the princi- pal strains in the concrete during testing. The intersection of the three LVDTs was located horizontally at the critical section as indicated by the AASHTO LRFD Specifications. Vertically, the rosettes were located 0.5dv from the bottom of the girder and horizontally dv is measured from the center of the support. For the AASHTO Type II girders, one of the LVDT rosettes was substituted with a rosette of resistance strain gages. How- ever, because the strain gage rosettes produced questionable post-cracking data, LVDT rosettes were used on both sides of the web for the remaining tests. Four LVDTs were placed on strands extending from the end of the girder to determine strand slippage as load was applied. The LVDTs were placed on two exterior and two cen- ter strands on the bottom row. The LVDTs recorded measure- ments to the nearest 0.0002 in. Four wire potentiometers (wire pots) were placed below the beam to measure the vertical deflection during testing. One wire pot was placed underneath one of the actuators, and a second wire pot was placed at the point of maximum deflec- tion. The other two devices measured the vertical deflection occurring at the supports. Measurements were recorded to the nearest 0.01 in. Each of the actuators used to apply a load to the girder had a load cell on top to determine shear force applied at the end of the girder. The load cells measured load to the nearest 500 lb. Test Setup and Procedure. All of the girders were sim- ply supported, with the loaded end of the girder being pin- supported to minimize actuator tilt. Figure 31 gives actuator locations for each test. Initially consideration was given to placing the two con- centrated loads 14 feet apart in order to simulate an AASHTO truck axle loading. However, analysis indicated that such a load spacing and short beam length could result in failure at the end of the girder not being tested. The 8-ft load spacing used on girder T2.8.Typ resulted in flexural failures (instead of the desired shear failure). Therefore, the two concentrated loads were placed 3 feet apart during subsequent tests. Figure 33 shows that the first actuator was located at a dis- tance “a” from the nearest support. Initial configurations had been selected to fill in any gaps in available data regarding this a/dv ratio. However, preliminary analysis indicated the shear strengths of the girders in the early tests were substantially greater than predicted by the AASHTO LRFD specifications. Consequently, the a/dv ratio for the second end of all of the PCBT girders was decreased from the planned 3.5 to about 3.0, as indicated in Table 23. The new a/dv ratios were selected with the goal of achieving a shear failure prior to flexural fail- ure yet keeping the a/dv ratio large. The load was applied to the girder using an electric hydrau- lic pump in stroke control in 20-kip load increments. The girder was examined at each load increment and cracks in the concrete were marked. However, no examination was made as the girder neared failure. On two occasions, the two primary actuators were observed to be too far out of plumb. At that point, a secondary actua- tor at midspan was engaged to maintain the deflection in the girder to reset the two primary actuators. After the two pri- mary actuators were realigned, load in those actuators was increased gradually to the point prior to unloading, and then the experiment progressed as planned. Loading continued at 20-kip increments until either a shear failure occurred or the girder was no longer able to withstand additional load without large increases in deflection, thus indicating that the beam was approaching a flexural failure. Test Results. Table 26 shows a comparison of the experi- mental results and the values predicted by the two design methods given in AASHTO LRFD specifications (the Sectional Design Method (5.8.3 and Appendix B5 with interpolation) a3'-0" P a 3'-0" P6" dv 6" = Strain gage location = Vibrating wire gauge location LC LC P P = LVDT rosette for measuring web strains = LVDT for measuring strand slip = Wire pot for measuring deflections (a) (b) Figure 33. Instrumentation on (a) the concrete surface and (b) the reinforcement.

46 and the Simplified Procedure for Prestressed and Nonpre- stressed Sections (5.8.3.4.3)). Subsequently, these two meth- ods will be called the sectional method and simplified methods (Sec and Sim in Table 26), respectively. The calculations using the AASHTO specifications set all load factors and strength reduction factors equal to 1.0 and were done with and without the 0.85 modifier for sand lightweight concrete (lv) given in 5.8.2.2 of the AASHTO LRFD specifications. The yield stress used for all mild reinforcement included in shear design cal- culations was 67.3 ksi (which was determined from testing samples taken from the reinforcement used in the girders). The effective strand stress (fpe) was determined from the vibrating wire gages located in each girder and f ′c used in the calculations was estimated from a strength gain curve. The load needed for determining shear strength was the sum of the dead load at the middle of the shear span and the applied load. For several beam ends, the stirrup spacing varied within the shear span; this change was incorporated in Vn by using the lowest calcu- lated Vn value within the shear span. Table 26 also lists the type of failure that occurred in each experiment. Each beam was designed according to the requirements of the AASHTO LRFD specification such that the flexural strength exceeded the design shear strength. Dur- ing the testing of T2.8.Min.2, an operational failure occurred that led to termination of the test before a flexure or shear failure occurred. Generally, a flexural failure is characterized by yielding of the longitudinal reinforcement, followed by crushing of the deck concrete near the point of maximum moment. How- ever, the tests were discontinued when very little additional force resulted in a relatively large increase in deflection, and crushing of the concrete at the top of the girder was immi- nent. A typical crack pattern for a flexural failure is shown in Figure 34. The flexural cracks extended from the bottom flange into the cast-in-place deck and the significant diagonal shear cracks within the shear span. Figure 35 indicates the level of stirrup strain in a typical test at failure. Approximately half of the gages were at or above the yield strain at failure, and there was concrete powdering or light flaking in the web, indicating that the girder was nearing its shear capacity. Shear failures generally occur in the form of web crush- ing, diagonal tension, shear compression, shear-bond failure, localized crushing above the support, horizontal shear along the flange-deck interface, and strand slip. Web crushing type failures, with or without strand slip, were observed in 8 of the 12 large-scale girder tests. In these cases, the concrete on the surface of the web spalled or crushed near the support. Figure 36 shows the end of a beam that has undergone a web- shear failure with the typical diagonal tension cracking and web crushing. Another characteristic of web crushing failures was significant yielding of the stirrups within the shear span. Figures 37 and 38 show the level of strains in the stirrups for two web compression failures: a short shear span test of a lightweight girder with f ′c = 8 ksi and a long shear span test of a lightweight girder with f ′c = 10 ksi, respectively. In both cases, all stirrups yielded before or at failure. A strand slip failure was defined as when the strands protruding from the Test ID Failure Mode(s) VExp(kips) Sectional Method Experimental Results Simplified Method Vn.Sec (kips) Vn.Sec No λv (kips) Exp/Sec Exp/Sec No λv Vn.Sim (kips) Vn.Sim No λv (kips) Exp/Sim Exp/Sim No λv T2.8.Typ.1 Flexural / Slip 361 275 279 1.31 1.29 298 293 1.21 1.23 T2.8.Typ.2 Flexural 292 230 233 1.27 1.25 293 289 1.00 1.01 T2.8.Min.1 Web Shear / Slip 382 243 248 1.57 1.54 246 244 1.55 1.57 T2.8.Min.2 * 308 219 223 1.41 1.38 241 239 1.28 1.29 BT.8.Typ.1 Web Shear / Slip 500 329 329 1.52 1.52 313 313 1.60 1.60 BT.8.Typ.2 Flexural 407 298 303 1.37 1.34 305 307 1.34 1.33 BT.8N.Typ.1 Web Shear / Slip 431 274 — 1.58 — 246 — 1.75 — BT.8N.Typ.2 Web Shear / Slip 382 254 — .50 — 239 — 1.60 — BT.10.Typ.1 Web Shear / Slip 518 293 299 1.77 1.73 296 296 1.75 1.75 BT.10.Typ.2 Web Shear / Flex. / Slip 428 286 292 1.49 1.47 293 295 1.46 1.45 BT.10.Min.1 Web Shear / Slip 475 253 260 1.88 1.83 202 207 2.35 2.30 BT.10.Min.2 Web Shear 371 240 246 1.55 1.51 199 204 1.87 1.82 Average Exp/Calc ratio for Lightweight Concrete 1.51 1.49 1.54 1.54 Average Exp/Calc ratio for Normal Weight Concrete 1.54 — 1.68 — *Equipment breakdown, shear strength based on load applied at breakdown. Table 26. Experimental versus calculated shear strengths. Figure 34. Flexural failure observed in Test BT.8.Typ.2.

47 5 P P 14'-6"4'-9" 1 2 6 7 8 9 10 1 1 11 2 3 3 Stirrup Number = strain gauge = malfunctioning gauge = stirrup yielded at gauge at cracking = stirrup yielded at gauge Figure 35. Strain gage locations for Test T2.8.Typ.1. Figure 36. Web-shear failure observed in Test T2.8.Min. P P 3'-0"4'-10" 5 6 7 8 Stirrup Number 1 1 1 2 2 2 2 3 3 4 3 = strain gauge = malfunctioning gauge = stirrup yielded at gauge at cracking = stirrup yielded at gauge Figure 37. Strain gage locations for Test T2.8.Min. P P 12'-9"3'-0" 6810 Stirrup Number 121416181921 11 1 1 1 11 222 2 2 2 33 3 3 4 4 = strain gauge = malfunctioning gauge = stirrup yielded at gauge at cracking = stirrup yielded at gauge Figure 38. Strain gage locations for test BT.10.Typ.2.

48 beam end were pulled into the concrete more than 0.01 in. In seven of the eight tests that resulted in a web-shear failure, strand slip in excess of 0.01 in. accompanied web crushing. Table 26 presents the calculated strengths for all of the light- weight and normal weight girders. The differences between the calculated strengths with and without lv = 0.85 for the general and simplified equations for shear strength are less than 3%. Also, the average ratio of experimental-to-calculated shear strength for lightweight concrete using either method was only minimally affected by the absence or presence of the lv factor. Therefore, including lv in calculating shear strength of lightweight concrete girders appears unnecessary. This is similar to the findings from the material property investiga- tion where it was found that the tensile strength of lightweight concrete was adequately predicted by the AASHTO provisions for tensile strength of normal weight concrete. AASHTO LRFD specifications also require the use of a lower resistance factor (f) for shear design of lightweight concrete girder (0.7 as compared to 0.9 for normal weight concrete). Table 27 presents average data from previous research on prestressed beams constructed with lightweight concrete by Malone (1999), Kahn et al. (2004), and Dymond et al. (2009). This previous research reported data for concrete compres- sive strengths ranging from 6.5 ksi to 11.0 ksi. Cross-sectional area for the beams ranged from 468 in2 to 1629 in2, and shear span-to-depth ratio, a/dv, ranged from 1.3 to 3.0. Some girders had no shear reinforcement, while others had minimal (qv ≈ 0.002) or typical (qv ≈ 0.011) shear reinforcement ratios. As listed in Table 27, the ratio of measured to calculated strengths averaged 1.52 for the present study, whereas for the previous research, this ratio is at least 1.70. Also, the coefficients of varia- tion indicated a greater variability with the Simplified method than with the Sectional method. In Table 28, the test results for lightweight prestressed con- crete girders are compared to results for normal weight pre- stressed concrete girders reported by Hawkins et al. (2005). The ratio of tested to calculated strengths is higher for light- weight than for normal weight prestressed concrete gird- ers; the coefficient of variation is also slightly higher for the lightweight girders. The bottom two rows show percentages of cases in which the measured strength would be less than 0.9 or 0.85 (the f factor) times the nominal strength for a normal distribution of data. For f equal to 0.9, the Sectional method for shear strength calculation of lightweight concrete prestressed girders gives a smaller percentage of cases with measured strengths less than design strengths than for nor- mal weight prestressed concrete girders. However, the reverse is true for the Simplified method. The ratio of measured to calculated strength for lightweight concrete girders without using lv is 3.5 as compared to a ratio of 1.4 for normal weight girders. The bottom row in the table evaluates the test results using a f of 0.85. For f of 0.85, all lightweight concrete results are less than their normal weight concrete companions cal- culated with a f of 0.9. For instance, the largest percentage for lightweight prestressed concrete girders is 2.6% which is less than the largest percentage for normal weight prestressed concrete girders (3.2%). Using f as 0.85 for the shear design of prestressed girders containing sand lightweight concrete yields a maximum percentage of 2.6. This is less than the larg- est percentage for normal weight prestressed concrete girders using a f of 0.9 (3.2%). The largest percentage for lightweight girders is for strengths calculated with the Sectional design method, whereas the largest percentage for normal weight girders is calculated using the simplified method. This com- parison was made since both methods are allowed in the AASHTO LRFD specifications. Statistic Previous Research Current Study Vexp_____ Vn sec Vexp_____ Vn sim Vexp_____ Vn sec Vexp______ Vn sec No λv Vexp_____ Vn sim Vexp_____ Vn sim No λv Average 1.70 1.79 1.51 1.49 1.54 1.54 Std. Dev. 0.32 0.42 0.19 0.18 0.38 0.36 COV 0.19 0.24 0.13 0.12 0.25 0.24 Table 27. Results from previous research and current study. Statistic Lightweight Beams Normal Weight Beams V exp _____ V n sec V exp _____ V n sec No λ v V exp _____ V n sim V exp _____ V n sim No λ v V exp _____ V n sec V exp _____ V n sim No λ v Average 1.61 1 .56 1.67 1 .58 1.23 1 .54 Std. Dev. 0.28 0 .27 0.42 0 .38 0.18 0 .29 COV 0.17 0 .17 0.25 0 .24 0.15 0 .19 % < 0.9V n 0.52 0 .71 3.1 3.5 3.2 1.4 % < 0.85 V n 0.31 0 .44 2.4 2.6 Table 28. Comparison of lightweight and normal weight concrete results.

49 Summary Observations regarding shear design of lightweight pre- stressed concrete girders follow: • Applying a modification factor (lv) to ′f c term in shear strength calculations is not needed for sand lightweight concrete prestressed concrete girders. This is supported by results from the material property portion of this project as well as the test results from the full-size girder tests. • AASHTO does not give different strength reduction factors for the types of shear in concrete bridge applications (i.e., interface shear, beam shear, and two-way shear); therefore, a common strength reduction factor for shear in members containing lightweight concrete is recommended. The AASHTO LRFD specifications require that a f of 0.7 be used for calculation of the shear strength of lightweight concrete girders, as opposed to 0.9 for normal weight con- crete girders. Based on analysis of the full-scale girder test results, a value of 0.85 is more appropriate for lightweight prestressed concrete girders. This is consistent with, but more conservative than, the findings of the interface shear portion of this study. 3.5 Time-Dependent Behavior of Lab-Cast and Full-Scale Beams The objective of this part of the research was to compare measured camber and strain data over time to calculated camber and prestress loss estimations to determine if the methods presented in the AASHTO LRFD specifications are applicable to lightweight concrete. Tests were conducted on 6 full-scale prestressed girders and 12 lab-cast beams. Creep and Shrinkage The following methods for predicting creep and shrinkage of concrete are identified in Section 5.4.2.3 in the AASHTO LRFD specifications: • CEB MC90 • ACI 209 model • AASHTO LRFD model All models consider relative humidity, volume-to-surface area, and time of loading. The AASHTO LRFD model has a factor for concrete strength, and the ACI model has factors for specifics of the mix design. None of the models includes a factor for lightweight concrete. The refined method of prestress loss calculation presented in AASHTO LRFD (Section 5.9.5.4) requires the calculation of creep coefficients and shrinkage strains. For the full-scale beams, the refined calculation was performed using each of the three creep and shrinkage models listed above. For the lab-cast beams, only the AASHTO LRFD method with AASHTO creep and shrinkage functions was compared to the measurements. Prestress Losses in Lab-Cast Beams Anchorage losses were compensated for by placing shims between the seating chuck and the stressing abutment dur- ing fabrication. Concrete was placed less than 24 hours after strand stressing for all placements and strand forces were continuously monitored from the beginning of stressing to release of prestressing force. Due to the heat of hydration and restraint of shrinkage and thermal effects prior to release, forces in the strands measured in the load cells at the non- stressing end of the prestressing beds varied during the cur- ing process. Therefore, the initial and effective prestress forces were approximated based on overall loss data as measured by the vibrating wire gages and crack initiation tests, which were used to determine the effective prestress force at the time of development length testing. Vibrating Wire Gages. Internal vibrating wire gages (VWGs) were used to measure the change in concrete strain at the level of the strands from the time of casting to the time of beam testing. All test beams were instrumented with a VWG except beams 1.LW1.5A and 1.LW1.5B. VWGs were placed 6 feet from the beam live ends, except beam 1.NW1.5A where the gage was placed 6 feet from the beam dead end. Gages were secured between the center strand and one of the outer strands using plastic zip ties. After casting, each VWG was connected to a datalogger, which recorded the measured strain at the level of the strands and temperature at specified time inter- vals. A sample plot of total change in strain versus time after casting is shown in Figure 39. Prestress transfer occurred at about 15 hours after concrete casting. Plots of prestress versus time for each lab-cast beam are given in Attachment L. Figure 40 shows a typical plot of prestress loss after transfer and the corresponding AASHTO estimate. In all cases, the AASHTO refined method of prestress loss overestimated the prestress loss for the lightweight concrete beams. Table 29 lists the experimentally determined elastic short- ening prestress loss together with the loss calculated using the measured modulus of elasticity and the modulus of elasticity calculated using the measured compressive strength at time of release. As can be seen, the measured losses are close to those calculated using either the measured or the calculated modulus. Table 30 presents the effective prestress force determined from the VWG data and as calculated using the AASHTO refined method. As can be seen, the measured effective prestress is 5 to 15% greater than that predicted by the AASHTO LRFD method.

50 Crack Initiation Tests. Crack initiation tests were con- ducted to determine the load at which the first flexural crack occurred in order to back-calculate the effective prestress force. In these tests, load was applied to the beam in 2-kip increments. The concrete surface strains were measured directly under the load and close to the bottom beam sur- face using surface-mounted gage points and a DEMEC digital strain gage. Expected flexural cracking loads were calculated assuming a modulus of rupture of 0.20 ′f c . Once the applied load was within 10 kips of the expected flexural cracking load, the beams were visually examined for cracking at each load increment. The surface strain measure- ments helped pinpoint the initiation and location of the flex- ural cracks. The load versus strain plot shown in Figure 41 indicates that the measured strains increased linearly with respect to load until cracking occurred. Then the load ver- sus strain plot became non-linear. The cracking load was taken as the lower of either the load at which a crack was visually observed or the load at which the load versus strain plot became non-linear. The beam was subsequently loaded until a crack could be visually detected across the bottom of the beam. Figure 39. Change in strain versus time for 2.LW3.5A. measured AASHTO estimated Figure 40. Prestress loss for beam 3.LW2.5A.

51 With the initial cracking load known, the effective prestress was calculated using the 28-day modulus of rupture. Table 31 lists the measured and effective prestress at the time of test- ing. The results show good agreement with AASHTO calcula- tions of prestress loss. Summary. Based on the VWG data, it appears the AASHTO LRFD specifications overestimate prestress loss in lightweight concrete beams after transfer, specifically at early ages. On the other hand, it seems possible, based on the crack initiation results, that the long-term estimate of effec- tive prestress loss is more reliable. This could be due to losses occurring prior to transfer. Further analysis is necessary to determine the root cause. Prestress Losses in Full-Scale Beams For the full-scale beams, two methods were used to esti- mate prestress losses. The first method was the AASHTO refined method (Section 5.9.5.4) and the second accounted for change in modulus of elasticity over time. For each method, the previously mentioned creep and shrinkage mod- els were used. Vibrating wire gage strain at the centroid of the prestressing strand at midspan of each girder was multiplied by the strand modulus of elasticity to determine a change in stress. The zero reading was taken immediately prior to release of prestress. Plots of prestress versus time for each girder are given in Attachment L. Elastic Shortening Losses. Elastic shortening losses were calculated by applying the full jacking force to the trans- formed section, and the resulting stress in the concrete at the level of the strand due to prestress plus self-weight was deter- mined. The stress in the concrete was multiplied by the strand to concrete modular ratio to arrive at the stress in the steel, which is the elastic shortening loss. Long-Term Losses. Two methods were used to estimate long-term loss of prestress: (1) AASHTO refined method (Sec- tion 5.9.5.4) and (2) an age-adjusted effective modulus (AAEM) method. Both methods use an age-adjusted effective modulus approach to deal with the effects of the slowly changing force in the prestress. In the AASHTO approach, the effects of creep, shrinkage, and relaxation are estimated separately. In the AAEM method, they are all considered in a single formulation. Inter- nal equilibrium, compatibility, and constitutive equations were written and solved simultaneously to determine prestress loss. This method allowed for the determination of the effect of deck Beam Prestress Loss (ksi) Calculated with Meas. Eci Calculated with Calc. Eci Measured 1.NW1.5A 10.4 10.9 11.5 1.NW1.5B 10.4 10.9 12.1 1.LW1.5A 13.5 13.4 NA 1.LW1.5B 13.5 13.4 NA 2.LW3.5A 16.6 16.7 14.2 2.LW3.5B 16.6 16.7 14.2 3.LW2.5A 16.3 15.1 15.4 3.LW2.5B 16.3 15.1 15.4 3.NW1.6A 11.0 10.8 11.2 3.NW1.6B 11.0 10.8 11.6 2.LW3.6A 17.0 17.1 15.2 2.LW3.6B 17.0 17.1 14.8 Table 29. Prestress loss due to elastic shortening. Beam Initial Stress (ksi) Time After Transfer (days) Prestress (ksi) Measured/ CalculatedMeasured fpe Calculated fpe 1.NW1.5A 200 207 175 167 1.05 1.NW1.5B 200 207 174 167 1.05 1.LW1.5A 199 NA NA NA NA 1.LW1.5B 199 NA NA NA NA 2.LW3.5A 199 139 172 152 1.13 2.LW3.5B 199 139 172 152 1.13 3.LW2.5A 203 83 177 163 1.09 3.LW2.5A 203 83 177 163 1.09 3.NW1.6A 198 83 177 168 1.05 3.NW1.6B 198 83 176 168 1.05 2.LW3.6A 200 139 174 153 1.14 2.LW3.6B 200 139 176 153 1.14 Table 30. Measured and estimated prestress losses.

52 Figure 41. Sample load versus strain (beam end 1.NW1.5B.A). Beam End Applied Load (kips) Measured fpe (ksi) Average Measured fpe (ksi) Calculated fpe (ksi) Measured/Calculated 1.NW1.5A Dead 32 152 157 166 0.95 Live 33 159 1.NW1.5B Dead 39 158 Live 39 158 1.LW1.5B Dead 29 149 159 161 0.99 Live 29 149 1.LW1.5A Dead 39 177 Live 48 163 2.lW3.5B Dead 34 148 149 149 1.00 Live 36 158 2.LW3.5A Dead 38 145 Live 31 146 3.LW1.5A Dead 40 178 165 158 1.04 Live 38 165 3.LW1.5B Dead 57 155 Live 42 162 3.NW1.6A Dead 65 185 177 164 1.08 Live 80 180 3.NW1.6B Dead 80 180 Live 93 165 2.LW3.6B Dead 55 149 153 149 1.03 Live 72 158 2.LW3.6A Dead 66 156 Live 60 149 Table 31. Measured and calculated effective prestress.

53 reinforcement and deck shrinkage on overall time-dependent changes in prestress. Instantaneous changes, such as elastic shortening losses and increases in strain at the time of deck placement, were calculated using the transformed cross-sectional properties of the girder. Measured Losses. The prestress loss in each girder was measured using a VWG placed at the centroid of the pre- stressing strands at the midspan of the beam. The initial elas- tic shortening losses were determined from the differences in strain before and after release. For these calculations, an ini- tial prestressing stress of 0.75fpu was used, which is the maxi- mum allowed by the AASHTO LRFD specifications. Results. Figure 42 presents the measured and calculated change in prestress force over time. The calculated prestress was estimated using the AAEM method. Figure 43 presents the change in prestress force over time as predicted by the AASHTO refined method; the changes at release and at the time of deck placement are identical for all methods. In both cases, the time of deck placement was about 50 days after release when there is a sharp increase in strand stress. These figures show that use of the ACI and CEB MC90 models over- estimate prestress loss. Most of the difference occurs in the calculation of time-dependent loss between release and deck placement. Table 32 presents the elastic changes in prestress at release and at the time of deck placement. Strains were calculated using the transformed section properties using both the mea- sured modulus of elasticity and the modulus of elasticity cal- culated with the measured compressive strength at the time of load application. As can be seen in Table 32, the measured elastic shortening losses are higher than calculated for all beams. Also, the measured modulus provides a better esti- mate of elastic shortening loss than the calculated modulus. The change in tendon stress at deck placement was very small. Measured strains were similar to the calculated values. Prestress Losses Using the AASHTO Refined Method Table 33 presents the time-dependent changes in prestress from immediately after release to the time of deck placement Figure 42. Measured and calculated strand stress for T2.8.Typ (AAEM method). Figure 43. Strand stress for T2.8.Typ (AASHTO refined method). Beam At Elastic Shortening (ksi) At Deck Placement (ksi) Calculated with Meas. Ec Calculated with Calc. Ec Measured Calculated with Meas. Ec Calculated with Calc. Ec Measured T2.8.Min -22.8 -22.1 -25.8 1.8 1.7 1.8 T2.8.Typ -22.8 -22.1 -25.9 1.8 1.7 1.8 BT.8.Typ -19.5 -17.4 -20.5 1.7 1.5 1.7 BT.8N.Typ -15.8 -13.1 -19.6 1.6 1.3 1.6 BT.10.Typ -17.6 -17.3 -20.5 1.6 1.5 1.6 BT.10.Min -17.6 -17.3 -18.3 1.6 1.4 1.6 Table 32. Early change in prestress.

54 using each calculation method. Table 34 presents the time- dependent changes in prestress from just after deck place- ment to destructive testing. Table 35 presents the total change in prestress from just prior to release to test. Figure 44 presents the bias, or measured value, divided by calculated value for the prestress change from release to test- ing. The data show the following: • For the lightweight concrete girders, all methods predict higher losses than those measured. • The magnitude of the error (prediction minus measured) was greater for the interval from just after release to deck placement, than that from deck placement to testing. • Predicted losses for the Type II beams were slightly better than for the PCBT beams. • The AAEM method predictions are closer to the mea- sured values than those predicted by the AASHTO refined method for the same creep and shrinkage model. • The AASHTO creep and shrinkage model provides slightly better predictions than the other two models. Camber of Lab-Cast Beams Initial Camber at Prestress Release. Initial camber val- ues were measured using the taut-wire method. Deflections were also calculated using basic beam mechanics together with the fresh concrete unit weights and initial measured cylinder compressive strengths. The gross section moment of inertia was used in calculating displacements. The initial prestress was assumed constant along the length of the beam and was calculated at midspan as the initial jacking force minus the calculated elastic shortening losses. Beam self-weight was cal- culated by accounting for all steel reinforcement in the beam and using the fresh concrete unit weights. Attachment M con- tains camber versus time plots for each lab-cast beam. Table 36 shows a comparison of the measured and calculated cambers. Beam Change in Prestress (ksi) AASHTO Refined AAEM AASHTO ACI 209 CEB MC-90 AASHTO ACI 209 CEB MC-90 AASHTO w/meas Ec Measured T2.8.Min -17.0 (0.64)* -22.7 (0.48) -21.6 (0.50) -17.0 (0.64) -22.7 (0.48) -21.6 (0.50) -19.4 (0.56) -10.9 T2.8.Typ -17.8 (0.48) -21.7 (0.40) -21.1 (0.41) -16.5 (0.52) -20.3 (0.42) -19.7 (0.44) -17.4 (0.49) -8.6 BT.8.Typ -18.3 (0.31) -22.8 (0.25) -23.0 (0.25) -19.4 (0.29) -24.5 (0.23) -22.3 (0.25) -21.3 (0.26) -5.6 BT.8N.Typ -14.9 (0.70) -17.9 (0.58) -16.4 (0.64) -12.4 (0.84) -15.2 (0.69) -13.4 (0.78) -16.0 (0.65) -10.5 BT.10.Typ -23.5 (0.23) -29.8 (0.18) -29.4 (0.18) -20.4 (0.26) -26.4 (0.20) -25.3 (0.21) -21.4 (0.25) -5.4 BT.10.Min -22.2 (0.13) -27.5 (0.10) -24.9 (0.11) -19.2 (0.15) -24.2 (0.12) -21.2 (0.13) -20.1 (0.14) -2.8 *Number in parentheses in each cell is the ratio of measured loss to calculated loss. Table 33. Time-dependent change in prestress from release to deck placement. Beam Change in Prestress (ksi) AASHTO Refined AAEM AASHTO ACI 209 CEB MC- 90 AASHTO ACI 209 CEB MC- 90 AASHTO w/meas Ec Measured T2.8.Min -3.5 (-0.10)* -2.2 (-6.5) -2.1 (-6.3) -1.8 (-5.5) -0.6 (-2.5) -1.6 (-5.0) -1.4 (-4.5) 0.4 T2.8.Typ -3.3 (-0.03) -3.5 (-36) -3.1 (-32) -1.0 (-11) -0.4 (-5.0) -1.9 (-20) -1.3 (-14) 0.1 BT.8.Typ -0.2 (1.63) 0.2 (-1.7) 0.7 (-3.3) -0.7 (2.3) -0.7 (2.3) -0.7 (2.3) -1.0 (0.3) -0.3 BT.8N.Typ -2.2 (0.18) -2.9 (7.3) -2.2 (5.5) -0.5 (5.5) 0.4 (-2.0) -0.8 (2.0) -0.4 (1.0) -0.4 BT.10.Typ -0.2(-3.71) 0.4 (0.6) 0.1 (0.1) -0.6 (-1.9) -0.8 -2.1) -0.6 (1.9) -0.9 (-2.3) 0.7 BT.10.Min -0.2 (2.67) 0.4 (-2.0) -0.6 (1.5) -0.8 (2.0) -0.9 (2.3) -0.7 (1.8) -0.8 (2.0) -0.4 *Number in parentheses in each cell is the ratio of measured loss to calculated loss. Table 34. Change in prestress from deck placement to testing.

55 Beam Prestress Loss (ksi) AASHTO Refined AAEM AASHTO ACI 209 CEB MC-90 AASHTO ACI 209 CEB MC-90 AASHTO w/meas Ec Measured T2.8.Min -42.3 (0.82)* -45.4 (0.76) -45.6 (0.76) -39.2 (0.88) -43.7 (0.79) -43.7 (0.79) -41.9 (0.83) -34.6 T2.8.Typ -41.5 (0.77) -45.6 (0.70) -44.6 (0.72) -38.0 (0.84) -41.2 (0.78) -42.1 (0.76) -39.7 (0.81) -32.1 BT.8.Typ -38.5 (0.63) -43.6 (0.56) -42.3 (0.57) -36.1 (0.67) -41.2 (0.59) -38.9 (0.63) -40.2 (0.61) -24.3 BT.8N.Typ -28.9 (1.01) -32.6 (0.89) -30.4 (0.96) -24.7 (1.18) -26.5 (1.10) -25.9 (1.12) -30.6 (0.95) -29.1 BT.10.Typ -39.5 (0.59) -45.3 (0.52) -45.2 (0.52) -36.9 (0.63) -43.0 (0.54) -41.7 (0.56) -38.3 (0.61) -23.4 BT.10.Min -38.3 (0.52) -43.0 (0.46) -41.3 (0.48) -35.9 (0.55) -41.0 (0.49) -37.8 (0.53) -37.0 (0.54) -19.9 *Number in parentheses in each cell is the ratio of measured loss to calculated loss. Table 35. Total change in prestress from prior to release to testing. Figure 44. Measured loss/calculated loss from just prior to release to testing.

56 As expected, the initial measured normal weight beam cambers were less than or equal to the measured lightweight cambers due to the lower stiffness of the lightweight concrete and the higher self-weight. Also, there was a large difference between the measured and predicted values, in particular, between the lightweight concrete from pour 2 and the normal weight concrete from pour 1 (note that the first number in the beam identification indicates the concrete pour number). In all cases, the initial measured camber was higher than that predicted for normal weight beam specimens and lower than that predicted for lightweight beam specimens. Time-Dependent Camber Figures 45 and 46 present the average measured cambers for beam pairs versus time. There is an apparent difference in the camber of the beams with 0.5 in. and 0.6 in. strands. For the 0.5 in. strands, the total cambers for the normal weight and lightweight beams are approximately the same, despite the higher initial cambers of the lightweight beams. However, for the 0.6 in. strands, the normal weight beam pair showed a significantly smaller camber than the lightweight beam pair. Figure 47 shows a typical plot of the measured average cambers for a beam pair and the predicted camber using both the PCI Multiplier Method and the PCI Improved Multiplier Method (PCI Bridge Design Manual (1997)). These predic- tions were calculated using tested cylinder properties. The AASHTO LRFD refined prestress loss method was used to calculate the improved PCI multipliers. Table 37 presents a summary of the measured and calculated camber. For the 0.5 in. strand normal weight beams, the PCI predictions were comparable to the measurements. However, the PCI method Beam ID Camber (in.) Measured/ Calculated Average Measured/Calculated Calculated Measured 1.NW1.5A 0.310 0.359 1.16 1.11 1.NW1.5B 0.310 0.336 1.08 3.NW1.6A 0.215 0.234 1.09 3.NW1.6B 0.215 0.237 1.10 1.LW1.5A 0.387 0.359 0.93 0.90 1.LW1.5B 0.387 0.359 0.93 2.LW3.5A 0.481 0.414 0.86 2.LW3.5B 0.481 0.395 0.82 3.LW2.5A 0.433 0.422 0.97 3.LW2.5B 0.433 0.422 0.97 2.LW3.6A 0.345 0.297 0.86 2.LW3.6B 0.345 0.297 0.86 Table 36. Initial measured versus calculated cambers. Figure 45. Time-dependent camber for beams with 0.5 in. strand.

Figure 46. Time-dependent camber for beams with 0.6 in. strand. Figure 47. Camber for beam pair 1.NW1.5. Beam Camber (in.) At release Final Calculated Measured Days After Transfer PCI Multiplier, Erection PCI Multiplier, Final PCI Improved Multiplier Measured 1.NW1.5A 0.310 0.359 141 0.533 0.645 0.615 0.633 1.NW1.5B 0.310 0.336 161 0.533 0.645 0.621 0.633 1.LW1.5A 0.387 0.359 119 0.697 0.845 0.747 0.609 1.LW1.5B 0.387 0.359 161 0.697 0.845 0.763 0.602 2.LW3.5A 0.481 0.414 160 0.858 1.040 1.050 0.664 2.LW3.5B 0.481 0.395 160 0.858 1.040 1.050 0.645 3.LW2.5A 0.433 0.422 146 0.840 1.018 0.968 0.641 3.LW2.5B 0.433 0.422 104 0.840 1.018 0.941 0.641 3.NW1.6A 0.215 0.234 104 0.394 0.478 0.423 0.336 3.NW1.6B 0.215 0.237 160 0.394 0.478 0.438 0.375 2.LW3.6A 0.345 0.297 140 0.616 0.748 0.714 0.453 2.LW3.6B 0.345 0.297 140 0.616 0.748 0.714 0.453 Table 37. Calculated and measured camber.

58 greatly over predicted cambers for all lightweight concrete beams. This may be due to sensitivity to the elastic modulus values used. As the elastic modulus for a particular beam pair decreases, the predicted deflections increase. Therefore, the low moduli of the lightweight beams resulted in high predictions. The PCI predictions were also high for both the normal weight and lightweight concrete beam pairs with 0.6 in. strand. Full-Scale Beam Camber For the full-scale beams, two methods were used to calculate initial cambers and three methods were used to calculate cam- ber changes with time. Camber was calculated using the PCI Basic Multiplier method (PCI Bridge Design Manual (1997)), PCI Bridge Design Manual Improved Multiplier Method (PCI Bridge Design Manual (1997)), and the previously described AAEM method. Four camber calculations were made using the AAEM method. In one calculation, the measured values of f ′c, f ′ci, E ′ci, and Ec were used together with the AASHTO creep and shrinkage equations. The other three calculations used the measured values of f ′c and f ′ci, and the modulus, creep and shrinkage values calculated using AASHTO, ACI 209, and CEB MC90. The PCI Bridge Design Manual Improved Mul- tiplier method was also used with each of the three creep and shrinkage models. Attachment M contains camber versus time plots for each girder. Initial Camber. Two methods were used to find initial camber. For the multiplier methods, the traditional approach to account for elastic shortening losses was used. This involved determining the elastic shortening loss at midspan and assuming that loss is constant along the length of the beam (as specified in AASHTO Section 5.9.5.5). The upward camber due to prestress was then calculated based on the jacking force minus the elastic shortening loss. Gross cross- sectional properties were used with the initial prestress force. For the second method, the transformed moment of inertia was calculated at several points along the beam. Because of the harped strands, the moment of inertia varied somewhat along the span. The points that were considered were the transfer point, quarter point, and midspan. With the transformed area, the jacking force, the self-weight moment, and the modulus of elasticity at release, the instantaneous curvature was calculated at each point. With these curvatures, the moment-area method was used to calculate the deflection at midspan. By analyzing each section separately, the elastic shortening loss at each location is different. This method also establishes the initial conditions needed to calculate camber changes using the AAEM method. Changes in Camber with the AAEM Method. The AAEM method uses creep and shrinkage values calculated by using AASHTO LRFD Bridge Design Specifications, ACI 209, or CEB MC90 creep and shrinkage equations. The change in curvature at each location is then calculated using the moment-area method. Changes in Camber with the PCI Basic Multiplier Method. The PCI Basic Multiplier method provides guid- ance to calculate the camber after release, before deck place- ment, after deck placement, and at the end of service life. For calculation of the camber before deck placement, PCI recom- mends multipliers found in Table 4.8.4.1 in the PCI Design Handbook (6th ed.). For estimating the camber after deck placement, the deflection due to the deck weight was calculated. The 28-day modulus was used for this calculation together with the transformed moment of inertia. The deflection due to deck weight was subtracted from the camber before deck placement to determine the camber after deck placement. The camber at end of service is a value that assumes the girder is about 10,000 days old. However, because the girders were con- siderably younger, a percentage of the end of service camber was used to estimate the camber at the time of testing. Changes in Camber with the PCI Improved Multiplier Method. The PCI Bridge Design Manual (BDM) Improved Multiplier Method was used to provide three additional esti- mates of camber. The AASHTO Refined method was used to estimate the prestress loss, which was required to complete the analysis. The three camber estimates used the creep coef- ficients found in AASHTO, ACI, and CEB MC90, and the pre- stress loss was determined with the corresponding creep and shrinkage model. The initial camber was found with the same method as the initial camber for the PCI Basic Multiplier method. To estimate long-term deflections to initial cambers from prestress and self-weight and the camber due to time- dependent prestress loss must also be estimated. Once the individual components of displacement were cal- culated, the formulas, found in Table 8.7.2-1 of the PCI BDM, were used. The final camber is typically estimated using the creep coefficient at end of service or 10,000 days. For this research, the final time was that at testing to destruction; the creep coefficient associated with that day was used. The PCI BDM notes that this method is not completely applicable to cambers in members with composite decks. It does not include effects such as the differential shrinkage of the deck and girder concrete. The PCI BDM recommends more rigor- ous methods of analysis for such cases. Measured Camber. To measure the actual camber of the girders being tested, a taut-wire system was used. To deter- mine the camber, the reader matched the wire against its reflection in the mirror and then read the value on a high- precision scale to the nearest 1/64th of an inch. Results. Figure 48 presents the measured camber versus time and the calculated cambers using the AAEM method for a Type II girder with typical shear reinforcement. Figure 49

59 presents the measured camber and the PCI multiplier meth- ods for the same beam. The “Measured Calculated” values in Figure 48 include an allowance to account for the effects of thermal gradients that develop during deck hydration. Table 38 presents the measured initial camber and the pre- dictions using three methods. In the first method, the cur- vature was determined at several locations along the beam, and the moment-area method was used to calculate camber. The table shows estimates using (1) the measured modulus of elasticity, (2) the modulus calculated with the measured com- pressive strength, and (3) the traditional method of apply- ing equivalent forces to the cross-section. The results for all methods were very similar. Table 39 presents the measured and predicted cambers just prior to placing the deck concrete. Camber predictions using the PCI Improved Multiplier method and the AAEM method with the AASHTO creep and shrinkage model were closest to the measured values. Table 40 presents the measured and calculated downward deflection due to the placement of the deck concrete using Figure 48. Measured and calculated cambers with AAEM. Figure 49. Measured and calculated cambers using PCI multipliers.

60 Beam Camber (in.) PCI Mult PCI Improved Multipliers AAEM AASHTO ACI 209 CEB MC-90 AASHTO ACI 209 CEB MC-90 AASHTO w/meas Ec Measured T2.8.Min 1.34 1.11 1.27 1.39 1.08 1.21 1.34 1.16 0.95 T2.8.Typ 1.34 1.11 1.22 1.33 1.09 1.17 1.29 1.12 0.94 BT.8.Typ 1.65 1.49 1.62 1.81 1.55 1.67 1.88 1.80 1.50 BT.8N.Typ 1.20 1.00 1.10 1.21 1.04 1.12 1.25 1.29 1.44 BT.10.Typ 1.65 1.52 1.67 1.94 1.58 1.71 2.00 1.68 1.65 BT.10.Min 1.65 1.48 1.61 1.78 1.54 1.66 1.84 1.64 1.57 Table 39. Camber at deck placement. Beam Camber (in.) Moment-Area w/ Measured Ec Moment-Area w/ Measured Ec Traditional Method Measured Camber T2.8.Min -0.16 -0.15 -0.16 -0.14 T2.8.Typ -0.16 -0.15 -0.16 -0.14 BT.8.Typ -0.16 -0.15 -0.17 -0.20 BT.8N.Typ -0.15 -0.14 -0.13 -0.11 BT.10.Typ -0.16 -0.16 -0.17 -0.17 BT.10.Min -0.15 -0.16 -0.15 -0.15 Table 40. Measured and calculated downward deflection from deck concrete. Beam Camber (in.) Using Moment-Area w/ Measured Ec Using Moment-Area w/ Calculated Ec Using Traditional Method Measured Camber T2.8.Min 0.74 0.71 0.75 0.72 T2.8.Typ 0.74 0.71 0.75 0.69 BT.8.Typ 1.05 0.90 0.92 0.97 BT.8N.Typ 0.84 0.66 0.67 0.99 BT.10.Typ 0.96 0.90 0.92 - BT.10.Min 0.96 0.90 0.92 - Table 38. Measured and calculated initial cambers. three calculation methods. The data presented in the table indicate that the methods predict camber change similar to the measured values. Table 41 shows the measured and calculated cambers from immediately after the deck was cast to testing of the girders. The methods based on the PCI multipliers all predicted a net upward camber during this time period while the methods based on the AAEM method predicted a downward camber. A net downward deflection was measured during the time interval and was closely matched by three of the four meth- ods using the AAEM method. Table 42 presents the measured and calculated total camber at the time each girder was tested to destruction. The multi- plier methods overestimated the camber growth after casting the girder for the majority of beams, while the AAEM method with the AASHTO creep and shrinkage model predicted cam- bers close to measured values for the majority of the beams. The AAEM method accounted for the downward displacement caused by differential shrinkage of the deck close to measured values for the majority of beams, but not the multiplier methods. Figure 50 presents the bias (measured value divided by cal- culated value) of the camber calculation methods. The closest

61 Figure 50. Bias of camber calculation methods. Beam Camber Change (in.) PCI Mult. PCI Improved Multipliers AAEM AASHTO ACI 209 CEB MC-90 AASHTO ACI 209 CEB MC-90 AASHTO w/meas Ec Measured T2.8.Min 0.07 0.18 0.12 0.14 -0.14 -0.14 -0.15 -0.26 -0.13 T2.8.Typ 0.07 0.16 0.15 0.17 -0.14 -0.14 -0.09 -0.23 -0.12 BT.8.Typ 0.06 0.07 0.07 0.08 -0.16 -0.10 -0.16 -0.27 -0.08 BT.8N.Typ 0.06 0.11 0.11 0.12 -0.14 -0.14 -014 -0.20 -0.09 BT.10.Typ 0.06 0.06 0.05 0.06 -0.12 -0.11 -0.12 -0.18 -0.07 BT.10.Min 0.05 0.07 0.06 0.08 -0.17 -0.18 -0.17 -0.24 -0.01 Table 41. Change in camber from casting of deck concrete to testing. Beam Camber (in.) PCI Mult. PCI Improved Multipliers AAEM AASHTO ACI 209 CEB MC-90 AASHTO ACI 209 CEB MC-90 AASHTO w/meas Ec Measured T2.8.Min 1.18 1.13 1.23 1.37 0.78 0.91 1.03 0.74 0.68 T2.8.Typ 1.18 1.11 1.21 1.34 0.79 0.87 1.04 0.73 0.69 BT.8.Typ 1.48 1.40 1.53 1.73 1.23 1.35 1.56 1.37 1.22 BT.8N.Typ 1.05 0.96 1.06 1.18 0.75 0.83 0.96 0.94 1.24 BT.10.Typ 1.47 1.42 1.56 1.84 1.30 1.44 1.72 1.34 1.41 BT.10.Min 1.48 1.40 1.52 1.71 1.22 1.33 1.52 1.25 1.41 Table 42. Camber at testing.

62 predictions were those made using the AAEM method with the measured modulus of elasticity. The AAEM method with the calculated modulus and AASHTO creep and shrinkage models also predicted cambers relatively close to measured. Summary The following observations can be made from the camber studies: • For the lab-cast beams with no cast-in-place deck, the PCI Improved Multiplier method predicted the camber of the normal weight beams relatively well, but over-predicted the cambers of the lightweight girders. • For the full-scale beams, the PCI Improved Multiplier method used with the AASHTO model for creep and shrinkage predicted cambers of the lightweight girders at the time the deck was placed close to those measured. The measured camber of the normal weight girder was signifi- cantly higher than predicted with this method. • For the full-scale beams and for calculations of camber after the composite deck was placed, the AAEM method with the AASHTO creep and shrinkage functions resulted in the closest predictions of camber to those measured. • High camber predictions for the full-scale beams were due to an overestimation of the camber growth from time of prestress transfer to deck placement. 3.6 Design Examples One simple composite bridge beam example was selected to investigate differences in resulting designs between light- weight and normal weight concrete. The effects on design of the elimination of the modification factor to the ′f c term in shear design and a change in the strength reduction factor for shear design were also investigated. The configuration presented as Example 9.4 in the PCI Bridge Design Manual (1997) was chosen. The following are the key parameters of the bridge: Overall deck width: 51.0 ft Design span length: 120.0 ft Number of beams: 6 Beam spacing: 9.0 ft Deck slab thickness: 8.0 in. Relative humidity: 70% Deck concrete f ′c: 4000 psi Girder concrete f ′c: 6500 psi Girder type: AASHTO-PCI BT-72 A typical interior girder was designed. The design was per- formed for the three different scenarios presented in Table 43. Sample design calculations for the three scenarios are pro- vided in Attachments N, O, and P. The results of the flexural designs are presented in Table 44. The designs with lightweight concrete need fewer total strands and fewer harped strands. Also, the prestress losses in the lightweight girders are somewhat higher, due to the lower modulus of elasticity. The shear design was performed using the sectional design model presented in AASHTO Section 5.8.3. The results of the shear design are presented in Table 45. Run B represents the current approach to shear design for lightweight concrete girders that uses both a lower strength reduction factor and a reduction to the ′f c term. For this run, the amount of shear reinforcement is almost twice that required for the normal weight girder. However, for Run C, with no reductions to the ′f c term and a strength reduc- Run Concrete φ (shear) v Ec,λ (ksi) Total Strands Harped Strands Losses at Transfer (ksi) Total Prestress Loss φMn (k-ft) A NWC 0.9 1.0 4890 44 8 17.4 40.4 10,640 B LWC 0.7 0.85 3720 40 6 21.5 46.4 9840 C LWC 0.85 1.0 3720 40 6 21.5 46.4 9840 Table 44. Results of flexural designs. Run Slab and Girder Concrete Strength Reduction Factor for Shear Modification Factor for cf ′ Term for Shear Design A Normal Weight 0.90 1.00 B Sand Lightweight 0.70 0.85 C Sand Lightweight 0.85 1.00 Table 43. Scenarios for design example.

63 Run Concrete (shear) v Vc (kips) Stirrup Bar Size Spacing (in.) Add’l Flexural Reinforcement at face of Bearing (in2) A NWC 0.9 1.0 200.4 #4 21 2.2 B LWC 0.7 0.85 179.2 #4 12 3.8 C LWC 0.85 1.0 210.8 #4 24 3.4 φ λ Table 45. Results of shear designs at critical section. tion factor of 0.85, the lightweight beam requires less rein- forcement, due to the lower dead load shear. Run C requires somewhat more longitudinal reinforcement at the face of the bearing because of the fewer strands. This design example illustrates typical differences between designs with normal weight concrete and lightweight con- crete. With the current specifications, for a given span length and girder spacing, the lightweight beam will require less pre- stressing but more shear reinforcement than a normal weight beam with the same cross section. However, with the elimi- nation of the modification to ′f c and changing the strength reduction factor for shear of lightweight concrete to 0.85, lightweight girders would require less shear reinforcement than a similar normal weight beam.