Ultra-high performance concrete for precast prestressed bridge girders : final report / by Victor Garas, C. Kennan Crane, Lawrence F. Kahn, and Kimberly E. Kurtis

School of Civil and Environmental Engineering
Structural Engineering, Mechanics and Materials Research Report No. 09-8
Ultra-High Performance Concrete for Precast Prestressed Bridge Girders
Final Report
Prepared for Office of Materials and Research Georgia Department of Transportation GDOT Research Project No. 2043
Task Order No. 02-08
by Victor Garas, C. Kennan Crane, Lawrence F. Kahn, and Kimberly E. Kurtis
June 2009

Contract Research GDOT Research Project No. 2043
Ultra-High Performance Concrete for Precast Prestressed Bridge Girders
Final Report
Prepared for Office of Materials and Research Georgia Department of Transportation
By Victor Garas, C. Kennan Crane, Lawrence F. Kahn, and Kimberly E. Kurtis
June 2009
The contents of this report reflect the views of the authors who are responsible for the facts and the accuracy of the data presented herein. The contents do not necessarily reflect the official
views or policies of the Georgia Department of Transportation. This report does not constitute a standard, specification or regulation.
i

Executive Summary
Ultra-high performance concretes (UHPCs) studied in this research developed compressive strengths greater than 25,000 psi, direct tensile strengths greater than 1,500 psi, modulus of rupture in excess of 4,000 psi, and modulus of elasticity greater than 6.9 x 106 psi. Tensile creep tests showed that the long-term behavior of the fiber-reinforced material is excellent and that even cracked specimens do not creep to failure. The tension test findings lead the conclusion that shear reinforcement is not needed for shear strength in bridge girders constructed with UHPC. Nevertheless, it was found in the laboratory tests that thermal treatment (curing) of UHPC above 170F and preferably at 195F is required if the concrete is to develop these excellent compressive and tensile properties.
The mixing and fabrication of quality control specimens and girders at a precast concrete plant demonstrated that UHPC could be commercially produced in Georgia and that multiple 4yd3 batches could be produced with good consistency and quality. Of concern in this field production was development of sufficient heat for adequate thermal treatment; adequate steam curing must be assured if UHPC is to be used for construction of bridge girders.
The transfer length of 9/16-in. prestressing strands was 55% of the transfer length given by the AASHTO LRFD specifications; therefore, the UHPC provided excellent bond and confinement for the strands.
Overall, the research found that UHPC may be effectively used for precast prestressed bridge girders in Georgia. Girders with a bulb-tee shape and with a depth of 28-in. may be used for simple spans of 90-ft. While stirrup reinforcement is not required for shear strength of the girders, stirrup reinforcement appears to be necessary to provide interface shear resistance between the composite deck slab and the UHPC girder. AASHTO LRFD specifications should be used for the composite girder flexural design, but an elastic mechanics approach using an ultimate tensile strength of 1,200 psi is needed for shear design of the UHPC sections.
ii

Acknowledgements
The research reported herein was sponsored by the Georgia Department of Transportation through Research Project Number 2043, Task Order Number 02-08. Mr. Paul Liles, State Bridge Engineer, Mr. Myron Banks, Concrete Engineer, Ms. Lyn Clements, bridge engineer, and Ms. Supriya Kamatkar, Research Engineer, of GDOT provided many valuable suggestions throughout the study. The opinions and conclusions expressed herein are those of the authors and do not represent the opinions, conclusions, policies, standards or specifications of the Georgia Department of Transportation or of other cooperating organizations.
Tindall Corporation personnel, especially Mr. Kevin Kirkley, assisted in all aspects of the field study. Lafarge North-America donated Ductal to Georgia Institute of Technology for laboratory and field studies. Mr. Vic Perry and Mr. Peter Calcetas of Lafarge North-America gave many valuable suggestions and supported the field research. The support of Tindall Corporation and Lafarge North-America and of their personnel is gratefully acknowledged.
Dr. Ben Graybeal along with Dr. Joey Hartman of the Federal Highway Administration were very helpful in providing guidance and advice concerning mixing, placing, curing and testing of UHPC. The following Georgia Tech graduate research assistants aided in the field study: Jennifer Dunbeck, Murat Engideniz, Luis Fajardo, and Katherine Snedecker. Mr. Javier Silva performed much of the analytical investigation. Dr. James Lai helped to guide the research program.
iii

Table of Contents

Executive Summary

ii

Acknowledgments

iii

Table of Contents

iv

Chapter

1. Introduction

1-1

Purpose and objectives

1-1

Scope

1-1

2. Background

2-1

Shear Behavior of Beams Made of Fiber-Reinforced Concrete

2-1

Ultra-High Performance Concrete (UHPC)

2-6

Tensile Creep of Concrete

2-16

Use of UHPC Field Construction and Performance

2-47

3. Analytical Investigation

3-1

Introduction to Analytical Investigation

3-1

Flexural Analysis for Type I Modified using HPC

3-2

Flexural Analysis for Bulb-Tee 28 using HPC

3-6

Flexural Analysis for Type I Modified using UHPC

3-8

Flexural Analysis for Bulb-Tee 28 using UHPC

3-10

Shear Analysis using UHPC

3-13

Horizontal Interface Shear

3-15

Conclusions based on Analytical Investigation

3-17

4. Material Properties

4-1

Compressive Strength

4-1

Tensile Strength

4-2

Modulus of Elasticity

4-7

Modulus of Rupture

4-9

Autogenous Shrinkage

4-18

Short-term Tensile Creep and Free Shrinkage

4-20

Long-term Tensile Creep and Free Shrinkage

4-28

Conclusions Regarding UHPC Material Properties

4-38

5. Field Evaluation

5-1

Introduction

5-1

Pour 1

5-1

Pour 2

5-3

Pour 3

5-6

Field Evaluation Conclusions

5-9

iv

6. Girder Behavior

6-1

Introduction

6-1

Test Girder

6-2

Transfer Length

6-11

Instrumentation and Test Setup

6-12

Flexural Test Results and Discussion

6-12

Flexural Test Conclusions

6-20

7. Conclusions and Recommendations

7-1

Conclusions

7-1

Recommendations

7-2

References

R-1

v

1. Introduction
1.1 Purpose and Objectives
The purpose of the research reported herein was to determine if Ultra High Performance Concrete (UHPC) has practical, economic application for construction of precast prestressed bridge girders in Georgia. The research was divided into five specific tasks, where each task had its own objectives. The goal of Task 1, the analytical investigation, was to determine if the expected higher compressive and flexural strength of UHPC would actually allow span lengths of AASHTO and PCI Bulb-T sections to be lengthened. The goal of Task 2, mix designs, was to examine the Lafarge Ductal UHPC mix and to develop a generic UHPC mix which would be more economical. The goal of Task 3, material properties, was to quantify the short and longterm material properties of UHPC, especially the tensile strength of Ductal. The goal of Task 4, Field Production, was to evaluate the ability of a precast plant in Georgia to make multi-yard batches of UHPC and to determine the quality of the material made in these large batches. The goal of Task 5, girder behavior, was to determine if the actual flexural strength of a composite girder made with UHPC matched the analytical predictions.
1.2 Scope
The scope of the research was broad. The background literature investigation considered all aspects of UHPC. Of special note was the lack of literature concerning direct tensile strength, especially as it related to diagonal tensile cracking, shear, in reinforced and prestressed girders. Therefore, investigation into tension and tensile creep behavior of all concretes were studied because a girder without shear reinforcement would rely on the diagonal tensile strength of the concrete to provide its shear resistance.
The analytical investigation covered girder strengths ranging from 6,000 to 30,000 psi and a range of girder sizes. Both pretensioned and posttensioned strands, and their combination were considered. Composite and noncomposite action between the deck slab and UHPC girder was studied; and slab concrete compressive strengths between 3,500 psi and 7,000 psi were considered.
The material study investigated various concretes with compressive strengths greater than 20,000 psi. Yet, it would found that only the Lafarge Ductal UHPC developed sufficient
1-1

tensile strength to permit elimination of shear reinforcement. Therefore, the experimental effort concentrated on Ductal and its short and long-term properties: compressive and tensile strength, elastic modulus, modulus of rupture, creep and shrinkage behavior.
The field evaluation investigated construction of small quality assurance samples to a 54ft long girder at one precast prestressed concrete plant in Georgia. The expense of the Ductal premix and fibers and the high labor cost prohibited trials at a second plant.
The 54-ft long girder was tested to determine flexural behavior. Shear behavior of test girders and of push-off and small beam specimens will be investigated under GDOT research project No. 07-05.
1-2

2. Background
2.1 Shear Behavior of Beams Made of Fiber-Reinforced Concrete
Reinforced concrete beams resist loads by means of internal moments and shears. Typically, a beam is designed first to contain the amount of reinforcement necessary for flexural resistance within the limits provided by design codes in order to ensure a ductile failure. Once this is done, a beam would then be designed for shear resistance. In a beam, the design for shear must ensure that the shear strength equals or exceeds the flexural strength at all points in order to avoid any shear failure which is frequently sudden and brittle [Anderson, 1957 and MacGregor and Wight, 2005].
In an uncracked simple span beam, the surfaces (stress trajectories) in which the principle tensile stresses act are plotted in Figure 2.1. The expected initial cracking pattern resembles the family of dashed lines shown, since concrete cracks when the principal tensile stresses exceed the tensile strength of the concrete. The cracking pattern in a test beam with longitudinal flexural reinforcement, but no shear reinforcement is shown in Figure 2.2 [Nawy, 2006] where two types of cracks can be seen, (a) flexural cracks which are the vertical cracks that start at the bottom of the beam where the flexural stresses are the largest, and (b) inclined cracks which form at the ends of the beam due to combined shear and flexure. These latter cracks are commonly referred to as inclined cracks, shear cracks, or diagonal tension cracks, and they develop at approximately an angle of 45o to the normal at sections close to the support in non-prestressed beam. Such cracks must exist before a beam can fail in shear; and in order to prevent such cracks from opening, special transverse, shear reinforcement has to be provided. The main four functions of the shear reinforcement in flexural member are the following: (1) carrying a portion of external
2-1

factored shear force Vu, (Vs), (2) restricting growth of diagonal cracks, (3) holding longitudinal reinforcement in place to provide the dowel capacity, and (4) providing confinement to the concrete in the compression zone if the stirrups are closed and closely spaced (less than 100 mm (4-in)). However, in addition to the shear reinforcement, the other portion of the external shear forces, Vu, is resisted by the concrete itself (Vc) through the (1) dowel action of reinforcement bars (Vd), (2) shear in the compression zone (Vcz), and (3) interlock between coarse aggregate particles (Vay). The distribution of internal shear stresses in a flexure member is shown in Figure 2.3. First, upon loading and before the development of any cracks, all shear stresses are carried by uncracked concrete. Upon the onset of flexural cracks and before inclined cracks start to develop, external shear is carried by Vcz, Vd, and Vay. It is not until inclined cracks start to develop before shear reinforcement start carrying some of the shear forces. With inclined cracks getting wider, stirrups crossing an inclined crack yield and Vs remains constant for higher loads. Once stirrups yield, cracks open more rapidly and Vay decreases due to the loss of interlock between coarse aggregates, forcing Vcz and Vd to increase until dowel splitting occurs and the compression zone crushes.
In reinforced concrete beams with span-to-depth ratio (L/h) greater than 5, flexural cracks at midspan generally occur before the principal tensile stresses at mid-height near the supports become critical. Once a flexural crack occurs, the tensile stress perpendicular to the crack in the concrete drops to zero, and major stress redistribution is necessary to maintain equilibrium. As a result, the onset of inclined cracking in a beam cannot be predicted from the principal stresses, unless shear cracking precedes flexural cracking.
A beam's shear span-to-depth ratio influences its failure mode (Figure 2.2). For any concrete beam without diagonal tension reinforcement, whether reinforced or prestressed or
2-2

Figure 2.1: Principal stress trajectories for beams of rectangular cross sections [Gere, 2001]
Figure 2.2: Failure patterns as a function of beam slenderness [Nawy, 2006] 2-3

both, the three major modes failure are (1) flexural failure, (2) diagonal tension failure, and (3) shear compression failure, among which the last two are considered to be brittle failure modes and should be completely avoided in design through providing the appropriate amount of shear reinforcement [Lim and Oh, 1999 and Nawy 2006].
Now, as a result of the absence of both the shear reinforcement (as per the manufacturer's recommendation), and coarse aggregates (typical for UHPC matrices, will be discussed later); a significant portion of the shear cracks propagation resisting mechanism in an UHPC beam is missing (i.e. Vs and Vay). This, in turn, dictates that the long-term diagonal tensile performance of the material is important and must be characterized before UHPC girders can be used in construction.

Resisting shear

Compression zone Dowel action Aggregate interlock Stirrups

flexural cracking

inclined

yield of Failure

cracking

stirrups

Applied shear

Figure 2.3: Distribution of internal shears in a beam with web reinforcement [MacGregor and Wight, 2005]

The effects of use of short steel fibers as shear reinforcement instead of conventional stirrups have been investigated for both reinforced and prestressed beams. Furlan and Hanai

2-4

[1999] postulated that the elimination of conventional shear reinforcement (stirrups or bent bars) can make the reinforcement simpler and may increase the productivity in long line precasting beds. Narayanan and Darwish [1987] conducted some 36 shear tests on simply supported rectangular prestressed and non-prestressed concrete beams, containing steel fibers (0.3x30 mm (0.012x1.18-in)) as web reinforcement with volumetric fiber fractions varying between 0.91% and 4.47%, variable shear-span effective depth ratio (a/d), and type and extent of prestressing. The results showed that the patterns of collapse were similar for both prestressed and nonprestressed concrete beams. For prestressed beams, fiber addition caused improvements in the ultimate shear capacity, and changed the brittle nature of a shear failure to be more ductile.
Imam et al. [1997] studied the incorporation of steel fibers in singly reinforced high strength concrete (HSC) beams without stirrups failing under the combined effect of flexure and shear. This study showed that inclusion of steel fibers in HSC beams without stirrups provided significant improvement of shear resistance and tended to increase the ultimate flexural capacity. It also showed that steel fibers can successfully replace the shear reinforcement, while the use of steel fibers as a complementary reinforcement of longitudinal bars had little effect.
Furlan and Hanai [1997 and 1999] studied the influence of prestressing and fibers on the shear behavior of thin-walled I-section beams with reduced shear reinforcement ratio. Nine concrete beams were built (six with prestressing forces) with three different mixtures: without fibers, with 0.2x2.3x25.4 mm (0.0079x0.091x1.0-in) steel fibers at 1% volume fraction, and with 0.05x42 mm (0.002x1.65-in) polypropylene fibers at 0.5% volume fraction. Shear reinforcement ratios varied from 0 to 0.225% (geometric ratio). This study showed that addition of steel fibers did not increase the compressive strength, but increased the tensile strength by 16% in some cases and also the shear strength except in the beams without shear reinforcement. More
2-5

importantly, upon comparing fiber reinforced concrete beams to those with no fibers, the former were characterized by: (1) smaller spacing between cracks, (2) slower development of cracks, (3) larger number of inclined cracks prior to collapse, (4) delayed appearance of inclined cracks and consequently the stirrups were tensioned later, and (5) more ductile failure. Thus steel fibers were suggested to be considered as equivalent shear reinforcement to stirrups. In this aspect, the advantages provided by steel and polypropylene fibers were similar, but because of the higher modulus of elasticity of steel fibers as compared to the polypropylene fibers, the strain in the stirrups in steel fiber beams was smaller. The potential for use steel fibers as shear reinforcement instead of stirrups was also confirmed in other research [Lim and Oh, 1999 and Cucchiara et al. 2004]. However, none of these studies have considered the effects of tensile creep on the longterm performance of flexural members with fibers substituting stirrups as shear reinforcement.
2.2 Ultra-High Performance Concrete (UHPC)
2.2.1 Definition First, it is important to recognize that high strength concrete (HSC) is not necessarily
high performance concrete (HPC) and also that ultra-high strength concrete (UHSC) is not necessarily ultra high performance concrete (UHPC).
ACI Committee 363 defines HSC as "concrete with a cylinder compressive strength that exceeds 41.4 MPa (6000 psi)", while UHSC was defined by Shah and Weiss (1998) as "a concrete mixture with compressive strength greater than 22 ksi (150 MPa)".
The term high-performance concrete (HPC) was first used for concrete mixtures with high strength, workability and durability [Mehta and Aitcin, 1990]. This definition, thus, recognizes high durability to be a major requirement in HPC beside high strength. As a result,
2-6

HPC concrete mixtures ought to be designed for high dimensional stability in order to keep the structure free of cracks for a long period of time [Mehta, 1999].
As for UHPC, it was defined by Collepardi et al. [1997] as "an ultra high-strength and high-durability concrete with advanced mechanical properties". Yet, when it comes to assessing the overall performance of concrete, previous definitions were broadened by the ACI Committee 116. ACI 116 defines High Performance Concrete (HPC) as "concrete meeting special combinations of performance and uniformity requirements that cannot always be achieved routinely using conventional constituent materials and normal mixing, placing, and curing practices; the requirements may involve enhancements of placement, compaction without segregation, long-term mechanical properties, early-age strength, volume stability, or service life in severe environments". In addition, Goodspeed et al. [1996] went further and stated that HPC can be specified not only by the strength, but by any of the following: freeze-thaw durability, scaling resistance, abrasion resistance, chloride penetration, creep, shrinkage, and modulus of elasticity. 2.2.2 Principles of Developing UHPC
It is well known that the strength of brittle materials like concrete is related to the porosity of that material. As porosity decreases, an exponential increase in the strength is often observed [Mindess and Young, 1981]. The key for reducing the porosity and obtaining high strength is reducing the water-to-cement ratio and providing proper compaction [Powers and Brownyard, 1948] and for that reason water-reducing admixtures are used to alter attraction forces between the cement particles improving the fluidity of cementitious systems, better dispersing the cement particles, and reducing the size of voids [Dodson, 1990].
2-7

The main principles of development of UHPC matrices are improving the homogeneity, increasing the dry-compacted density, and enhancing the microstructure of regular concrete. These have been achieved by either: 1- Modifying cement with a polymer (macro-defect free, MDF), or 2- Densification with addition of micro-fine particles [Shah and Weiss, 1998].
Macro-defect free (MDF) materials are made using cement, a water-soluble polymer (Such as PVA, typically less than 5%), and a low w/c (typically less than 0.2). Originally it was thought that the very high tensile strength of MDF (200 MPa (29,000 psi)) approaching that of steel) was due to the reduction in pore size which occurs as a result of processing. However, recent work has shown that the significant increase in strength arises as a result of the crosslinking between cement and polymer [Poyola et al., 1990]. As a result, high shear mixing process is required for MDF mixtures to produce the mechano-chemical reaction between the mineral and polymer phases [McHugh and Tan, 1993].
Densification with micro-fine particles is based on the concept of particle packing and is the approach frequently used. As previously mentioned, superplasticizers allow the cement particles to pack more uniformly, reducing the porosity of conventional concrete, and thereby increasing strength. The particle-packing concept can be further utilized by adding submicron particles (e.g. silica fume) that fill remaining void space, resulting in a dense, strong material. If these particles are also pozzolanic, additional increase in strength may occur. In addition, the increased density of these materials reduces the connected porosity, decreasing the penetrability to water and corrosive agents and, thus, increasing long-term durability.
Graybeal [2005] stated that most of the UHPC matrices are generally composed of fine sand, between 150 and 600 m (0.0059 and 0.0236-in), cement with an average diameter of approximately 15 m (0.00059-in), crushed quartz with an average diameter of 10 m
2-8

(0.000394-in), and silica fume that has a diameter small enough to fill the interstitial voids between the cement and the crushed quartz particles. Depending on the time, temperature, and particle size, quartz particles can react with alkaline solutions. Metamorphic quartz has been found to be alkali reactive, generally in the decreasing order of reactivity [Mehta and Monteiro, 2005].
Richard and Cheyrezy [1995], and Feylessoufi et al. [1997] stated that further improvements of the cementitious matrix could be achieved via thermal treatment as the reaction of silica fume is activated and the average size of pores is decreased through the application of thermal treatment. Porosimetric analyses with mercury intrusion conducted by Cheyrezy et al. [1995] showed zero porosity in confined RPC specimens cured between 150 oC (302 oF) and 200 oC (392 oF). Also, in a study by Collepardi et al. [1997], it was concluded that both shrinkage and swelling in RPC reduced upon applying steam curing. Similar trends were also observed by Richard and Cheyrezy [1995] where heat curing at 90 oC (194 oF) was applied. Also, Monosi et al. [2000] concluded from their study that high pressure steam curing at 160 oC (320 oF) causes further strength increase with respect to the RPC's that were steam-cured at 90 oC (194 oF).
In summary, ultra-high strength matrix could be achieved by (1) low water-to-binder ratio (typically below 0.2), and using high dosage of high-range water reducing agents (HRWR) typically based on polycarboxylate (PC) chemistry meeting the ASTM C 494 requirements for Type F high-range water reducing admixtures [Graybeal, 2005; Habel et al., 2007; and Ferron et al., 2007] as they provide better dispersion of cement particles due to the steric hindrance dispersion mechanism compared to other types of HRWR.), (2) large quantity of fine particles
2-9

(typically silica fume), (3) aggregate containing only fine sand, and (4) thermal treatment (high temperature curing).
As concrete becomes more homogenous, it exhibits higher strength but becomes more brittle than normal strength concretes. Increased homogeneity can be observed by comparing a cross section of a typical HSC in comparison with UHSC (Figure 2.4). Unreacted cement and aggregate particles produce significant heterogenics in the standard high strength system, while the UHSC system is much more uniform at the same scale. Figure 2.5 provides insight into how the material performance changes with increased strength by showing the stress-strain failure envelope of conventional, high strength, and ultra-high strength concrete. It can be seen that an increase in compressive strength results in a material with high stiffness and ultimate failure strain. But on the other hand, concrete with higher strength exhibit a more sudden drop in load carrying capacity after the peak load is reached (post-peak) [Shah and Weiss, 1998]. This problem could be solved by incorporation of short fibers in the concrete mix.
Fiber reinforcement of cement-based matrices are known to (1) change the rheology or the flow characteristics of the material in the fresh state, (2) improve the tensile and flexural strength, (3) improve the impact resistance or toughness, (4) control cracking, and (5) alter the mode of failure by increasing post-cracking ductility. In most of the applications, fiber content ranges from about 0.3 to 2.0 % by volume. In the precast prestressed industry, an increase in tensile strength has a significant beneficial effect on the achievement of longer spans and the reduction of the amount of prestressing strands needed. However, fiber-reinforced concretes differ from conventional concretes in having lower coarse aggregates content (350 to 750 kg/m3 (600 to 1250 lb/yd3)), and a smaller size of coarse aggregate (10 mm (3/8-in) maximum size) [Lankard, 1975; Swamy and Barr, 1989; and Banthia et al. 1999].
2-10

In a study by Shah and Weiss [1998], fractured specimens of fiber-reinforced concrete showed that failure takes place primarily due to fiber pull-out or debonding, and, unlike plain concrete, a fiber reinforced concrete specimen does not fail immediately after initiation of the first crack. In explaining the toughening mechanism in fiber reinforced composites, Shah [1984] stated that the composite will carry increasing loads after the first cracking of the matrix if the pull-out resistance of the fibers at the first crack is greater than the load at first cracking. At the cracked section, the matrix does not resist any tension, and the fibers carry the entire load taken by the composite. With an increasing load on the composite, the fibers will tend to transfer the additional stress to the matrix through bond stresses. If these bond stresses do not exceed the bond strength, then there may be additional cracking in the matrix. This process of multiple cracking will continue until either fibers fail or the accumulated local debonding will lead to fiber pull-out.
Figure 2.4: Comparison between conventional high performance concrete and reactive powder concrete at the same scale [Shah and Weiss, 1998]
Figure 2.5: Stress-strain diagrams of concrete illustrating: increasing brittleness with increased strength (a) using external confinement; (b) using fiber reinforcement [Shah and Weiss, 1998]
2-11

Cracking in tension was further discussed in details by Rossi [2001]. Three different stages were defined from the beginning of loading until failure. In stage one, microcracks form randomly throughout the entire volume of concrete. During stage two, the microcracks join together to form localized macrocracks which affect the mechanical behavior. In stage three, one or more macrocracks become wider and propagate causing the final failure. Similar scenario was also observed by Namaan and Homrich [1989]. However, different fibers may be more efficient than others in resisting crack propagation according to their relative size with respect to the crack they are bridging. Thus, having a large number of closely spaced fibers seems to be the most effective way to bridge microcracks formed in the mixture during stage one. This can be achieved by using short fibers (<5.08 mm (0.2 in)) with small diameters that allows the placement of a large number of fibers without the presence of workability problems. On the other hand, fibers used must be long enough (>20 mm (0.79 in)) to bridge macrocracks formed during stages two and three. However, the use of long fibers has been always associated with reduced workability and thus typically limited to a volume fraction of 3%. This workability problem can be overcome by: (1) the use of two different types of fibers in the mixture: short fibers bridging microcracks, and long fibers bridging macrocracks, and/or (2) the use of increased amounts of superplasticizers (HRWRs)
Several studies in the last few decades have focused on experimentally studying the influence of incorporating different types of short fibers on concrete performance. Hannant [1978] divided fibers used to reinforce concrete into two main groups, those with modulus lower than the cement matrix, such as cellulose, nylon and polypropylene and those with higher modulus such as asbestos, glass, steel, carbon, and Kevlar (aramid). The low modulus organic fibers are generally subject to relatively high creep which means that if they are used to support
2-12

permanent high stresses in a cracked composite, considerable elongations or defections may occur over a period of time. Therefore, they are more likely to be used in situations where the matrix is expected to be uncracked, but where transitory maximum loads are short-term such as handling stresses, impacts, or wind loads. Another problem with the low modulus fibers is that they generally have large values of Poisson's ratio and this, combined with their low modulus, means that if stretched along their axis, they contract diametrically much more than other fibers. This contraction leads to a high lateral tensile stress at the fiber-matrix interface which is likely to cause a short aligned fiber to debond and pull out. On the other hand, the high modulus short fibers may require mechanical bonding to avoid pull out unless the specific surface area is very large. For structural and nonstructural purposes, steel fibers are still the most commonly used of all the fibers, and are commonly produced with various cross-sections and may have bent ends to provide anchorage with the cementitious matrix [Mehta and Monteiro, 2005 and Bissonnette et al., 2007].
However, as only steel fibers are considered to be used in the proposed study, only the effect of incorporating short steel fibers on the tensile performance of concrete is discussed here.
Data from the tests by Krenchel [1974] on both plain and steel fiber-reinforced mortars showed that incorporation of 0.9 and 2 % fiber by volume of concrete increased the flexural strength by approximately 15 and 30 %, respectively. In addition, in both cases the elongation at rupture was 9 to 10 times that of the unreinforced mortar. In a study by Potrzebowski [1983], 2% volume fraction of steel fibers, 0.4 mm (0.016-in) in diameter and 40 mm (1.57-in) in length were used in concrete of 10 mm (0.40-in) maximum size of aggregate (MSA) to study the effect of fiber reinforcement on the splitting tensile strength. Results from this study showed that the splitting tensile strength increased with the increase of the amount of fibers passing the crack
2-13

plane. Another study by Rossi et al. [1986] showed that incorporating about 1% volume fraction of 0.5x50 mm (0.020x2.0-in) steel fibers increased the load capacity by 200% when compared to non-reinforced concrete. Kormeling and Reinhardt [1987] studied the effect of strain rate on mechanical properties of fiber-reinforced concrete in uniaxial tension. In this study, straight steel fibers, 0.4 mm (0.016-in) in diameter and 25 mm (1.0-in) in length were used at 1.5 and 3% volume fractions. The MSA was 8 mm (0.31-in). Results from this study showed that fracture energy of fiber-reinforced concrete was up to hundred times more than non-reinforced concrete. In a study by Naaman and Hormich [1989], the use of 12-14 % volume fraction of hooked or deformed steel fibers, 0.5 mm (0.020-in) in diameter and 30 mm (1.18-in) in length in UHPC resulted in a multiple cracking pattern at low and intermediate load levels. However, failure still occurred through opening of a single large tensile crack. In addition, the tensile modulus of concretes with deformed fibers was substantially higher than that of concrete with hooked fibers. The difference between the behaviors of the two fibers was attributed mainly to the surface texture of deformed fibers which created at smaller strains a better matrix-to-fiber bond than hooked fibers [Bissonnette et al., 2007]. Krstulovic-Opara and Malak [1997] studied the effect of using high strength steel fibers on the tensile behavior of slurry infiltrated mat concrete (SIMCON). In this study stainless steel fibers with 0.334 mm (0.013-in) in equivalent diameter and 241.3 mm (9.5-in) in length were used at 2.16-5.39% volume fraction and direct tension tests were performed. Results from this study showed a monotonic increase in the tensile strength and toughness upon increasing the fiber content. The maximum tensile strength varied between 7 and 17 MPa (1015 and 2030 psi) for fiber volume fractions varying between 2.16 and 5.39 %, respectively. In addition, the toughness values varied between 0.124 and 0.29 MPa (17.98 and 42.05 psi) for 2.16 and 5.39 % fiber volume fractions, respectively. Also, the values of strength,
2-14

strains at maximum stress, and energy absorption capacity reported in this study were respectively about one order, two orders, and three orders of magnitude larger than standard unreinforced concrete. Similar observations were also reported previously by Hannant [1978]. Also, according to the ACI Committee 544, the total energy absorbed in fiber debonding as measured by the area under the load-deflection curve before complete separation of a beam might be about 10 to 40 times higher for fiber-reinforced concrete than for plain concrete.
The Portland Cement Association (PCA) investigated the changes in mix proportion upon incorporating 0.254 x 0.056 x 25.4 mm (0.01 x 0.022 x 1-in) steel fibers in fiber-reinforced concrete mixture designed for highways and airport pavements and overlays [Hanna, 1997]. Based on this study, a chart was also developed to determine the increase in the cement content and the decrease in aggregate proportions for the fiber additions in the range 0.5 to 2 % by volume. Using this chart, the mix proportions in Table 2.1 showed how at a given water-tocement ratio, the cement paste content had to be increased with a corresponding decrease in the proportion of aggregates to maintain adequate workability when 2 % steel fibers were added to the plain concrete mixture. In addition, the maximum particle size of the matrix was also important because it affected the fiber distribution and the quantity of fibers which could be included in the composite. Concrete which is intended to be used in conjunction with fibers should not have particles greater than 20 mm (0.787-in) and preferably not greater than 10 mm (0.394-in); otherwise uniform fiber distribution becomes difficult to achieve [Hannant, 1978].
On the other hand, inspite of the various improvements in concrete performance associated with incorporating steel fibers, recent studies have shown that using short steel fibers may have a negative effect on long-term tensile performance (tensile creep). This effect is discussed in detail later, but in general the effect was attributed to either the increase in the void
2-15

ratio fraction upon incorporating fibers [Bissonnette and Pigeon, 1995] or to the hypothesis that

fibers are likely to act like coarse aggregate in a concrete mix, having a surrounding porous zone

similar to the interface transition zone (ITZ) in the case of aggregates [Bissonnette et al., 2007].

Table 2.1: Comparison of mix proportions between Plain concrete and fiber-reinforced-

concrete (lb/yd3) [Hanna, 1997]

Material

Plain concrete Fiber-reinforced concrete a

kg/m3 (lb/yd3)

kg/m3 (lb/yd3)

Cement

446 (752)

519 (875)

Water (water/cement ratio = 0.45)

200 (338)

234 (394)

Fine aggregate

853 (1440)

760 (1282)

Coarse aggregate

682 (1150)

607 (1024)

Steel fibers (2% by volume)

-

157 (265)

a The 14-day flexural strength 7.93 MPa (1150 psi) of the fiber-reinforced concrete was

about 20% higher than that for plain concrete.

In summary, the main factors controlling the theoretical performance of a fiber-composite

material are the physical properties of the fibers and the matrix and the strength of the bond

between the two, some of which are hard to evaluate [Mehta and Monteiro, 2005].

2.3 Tensile Creep of Concrete
Findley et al. [1976] defined creep as the slow, continuous deformation of a material under constant load. Some materials like concrete exhibit an elastic action when rapidly loaded and then a slow and continuous increase of strains at a decreasing rate occurs. Upon unloading, initial elastic recovery was followed by a continuous decreasing strain. These materials are called "viscoelastic". Examples of viscoelastic materials are plastics, wood, natural and synthetic fibers, concrete, and metals at high temperatures.

2-16

In this section, a theoretical background of the phenomenon of tensile creep of concrete is presented followed by a literature review of the tensile creep test setups, key results, and possible mechanisms that have been used to explain it.
2.3.1 Theoretical Background Creep of concrete usually occurs simultaneously with shrinkage. Although these two
phenomena have been considered to be additive for many years; in fact, it is well known now that they are not, and that the principle of superposition cannot be applied [Kovler, 1999].
Pickett [1942] introduced the idea of ``drying creep'' to explain the observed excess of total creep at drying over basic creep (the component of concrete creep under conditions of no moisture movement to or from the ambient medium). In other words, the Pickett effect expresses the fact that, during simultaneous creep and shrinkage, the total time-dependent strain (tot) is not equal to the sum of the separate strain of load-free concrete exposed to shrinkage (fs) and that of loaded concrete prevented from drying (bc), but tot differs by an extra strain, called drying creep (dc), as follows:
tot= fs+ bc+ dc...........................................(2.1) Due to the nonuniformity of the moisture distribution in a drying specimen (wet core in compression, and drying and shrinking surface in tension), the drying creep component was explained by many models by the so called microcracking effect [Pickett, 1942]. Due to the nonlinear inelastic behavior of creep in concrete caused by the tensile stresses, these microcracks cannot fully close when the moisture distribution approaches a uniform state. However, there can be no effect of microcracking for specimens under compression, and thus studying the Pickett effect under tension will be more helpful in solving the problem of microcracking in concrete skin at drying creep [Kovler, 1995].
2-17

Kovler [1995] investigated the problem of interdependence of creep and shrinkage under tension. Figure 2.6 shows that the nature of the basic creep is different than that of the total creep. Total creep increases monotonically with time, while the basic creep reaches an asymptotic value within few hours. In addition, and unlike the creep under compression, during the first 36-48 hours of the tensile creep test, the total tensile creep was less than the tensile basic creep. This means that the difference between the total and the basic creep (i.e. the drying creep, dc) had a negative value initially, likely due to shrinkage which decreases with time and gradually transforms the dc value to positive later. As a result, the negative dc value at the initial stage of the test and the positive value later cannot be described by one term or one physical mechanism (i.e. creep or shrinkage) but rather combined mechanisms which are: (1) creep-induced shrinkage dominating at the beginning (cs), and (2) shrinkage-induced creep, dominating at the later stage (sc) [i.e. microcracking according to Bazant and Xi, 1994]. This approach can be represented by the following equation that reflects the interdependency between creep and shrinkage and can be used both in compression and tension:
dc = cs + sc ..............................................(2.2) According to Kovler [1995], the first term in Equation 2.2, (cs), represents the additional shrinkage component of concrete that depends on creep. That is, the greater the basic creep of the material, the greater the creep-induced shrinkage. The sign of cs will coincide with that of the free shrinkage. The second term in Equation 2.2, (sc), represents the additional creep component of concrete that depends on shrinkage. That is, the greater the free shrinkage of the material, the greater the shrinkage-induced creep. The sign of sc will coincide with that of the basic creep. In other words, Equation 2.2 implies that the deformations observed in drying creep result from
2-18

four mechanisms which are: (1) free shrinkage, (2) basic creep, (3) creep-induced shrinkage, and (4) shrinkage-induced creep.
Thus, Equation 2.2 can be rewritten in the following form: dc = sc + cs = X1fs + X2bc ..................................(2.3)
Where X1 and X2 are positive coefficients that can be obtained by linear regression (as first approximation), and = 1 for compression and = -1 for tension.
The analysis of Equation 2.3 (Figure 2.7) showed that X1 varies linearly with the stress applied, , unlike X2. However, as both X1 and X2 depend on the stress applied, both cs, sc can also be considered stress-dependant and thus Equation 2.3 can be written in the following form:
dc = sc + cs = X1 () fs + X2('fs)bc () ............................(2.4) where 'fs is the free shrinkage rate, X1 = coefficient linearly dependant on and X2 = coefficient dependant on insensitivity of the shrinkage process.
Kovler further investigated the phenomenon of tensile creep in concrete and presented the needed corrections to the previously suggested model (Equation 2.4) [Kovler, 1999]. This further investigation was motivated by: (1) the shortcomings found upon trying to apply the previously suggested model (Equation 2.4) to experimental work by Thelandersson [1988], (2) the observed mechanical behavior of drying concrete which was characterized by very insignificant hysteresis loops at loading and unloading cycles, constancy of elastic modulus, and growth of tensile strength (Figure 2.8) [Kovler, 1995], and most of all (3) the new experimental results that became available after the first study [Kovler, 1996].
Kovler [1999], suggested the existence of what is called the "abnormal" behavior of drying creep strain in the initial period of drying (1.5-2.0 days), when drying creep was contrary to the load direction (i.e. as shrinkage). This "abnormal" behavior was due to the fact that when
2-19

concrete elements are under simultaneous tensile loading and drying, these two factors produce deformations of opposite signs (i.e. expansion produced by tensile loading and contraction produced by drying), and thus the resulting drying creep strain will not necessarily coincide in direction with the load but will change with time (Figure 2.9) [Kovler, 1995, and Kovler, 1999].
This "abnormal" behavior was explained by attributing the excess basic creep strain over the total creep strain in the initial period to swelling (or reversible shrinkage) of the sealed concrete specimens if strains were considered from the moment of rewetting [Kovler, 1996]. Such a phenomenon was believed to be affected by three major factors: (1) capillary stress at RH > 40%, (2) disjoining pressure, (3) changes in surface free energy; in addition to (4) movement of interlayer water.
In Kovler [1996], a simplified model was laid out to explain the dependency of the "abnormal" behavior or drying creep in the initial period (discussed above) on the vapor pressure in the capillaries (Figure 2.10). Initially, a loss in the free water in capillaries begins as the relative humidity drops below 100 % of initial saturation level. The initial radius of the curvature R does not change until a spherical meniscus of a radius R is formed (Figure 2.10). This process of free water evaporation is not accompanied by any deformation (i.e. shrinkage).
Capillary water then continues to be removed only if the radius of the meniscus drops to a value, r, smaller than the initial radius R (Figure 2.10). This process of capillary water removal is accompanied by shrinkage in the material. When concrete is sealed after the initial drying period, the vapor pressure in the capillaries increases quickly, the radius of curvature of the meniscus increases and the level of liquid becomes flatter. As a result, the capillary surface tension is released and a swelling in the material occurs (Figure 2.10).
2-20

Total Creep Basic Creep Total Creep Basic Creep
Figure 2.6: Typical curves of creep and basic creep strains for different tensile stresses applied [Kovler, 1995]
X2
X1
Figure 2.7: Regression coefficients X1 and X2 versus stress-strength ratio [Kovler, 1995] 2-21

(a) (before creep)
(after creep)
(b) (before creep)
(after creep)
Figure 2.8: Stress-strain diagrams obtained before and after (a) basic creep test, and (b) drying creep test [Kovler, 1995] 2-22

(a) Basic creep
Total strain Free shrinkage
(b) Total strain Basic creep
Figure 2.9: Results of tensile creep tests of concrete (a) all results, and (b) total and basic creep [Kovler, 1999] 2-23

Figure 2.10: Simplified model of capillary water movement during evaporation and sealing, [Kovler, 1996] c = tot - fs
bc,corr = bc - sw Figure 2.11: Total creep(c), and corrected basic creep (bc,corr) [Kovler, 1999]
2-24

The same concept can then be applied in case of concrete rewetting. According to this model, the deformations of sealed concrete should be looked at completely differently than concrete allowed to dry, as the vapor pressure within a sealed specimen is higher than an unsealed specimen.
To investigate this approach, Kovler [1999] measured deformations of load-free sealed concrete at water-to-cement ratio of 0.7 after some initial period of drying. The swelling deformations (sw) observed did not exceed 40 45 and that the amount of swelling depended on both the duration of curing and the duration of drying before sealing. Upon considering swelling deformations in calculating drying creep, the following approximate relations were presented:
sh= -t/(0.0010t + 0.0033)......................................(2.5a) c= t/(0.0023t + 0.0087) .......................................(2.5b) bc= t/(0.0135t + 0.1636) ......................................(2.5c) sw= t/(0.0285t + 0.0012) ......................................(2.5d) where sh, c, bc, and sw are free shrinkage, total creep, basic creep, and swelling strains respectively, and t is the time in days. The total tensile creep and the corrected basic creep are plotted in Figure 2.11. Based on these results, the corrected drying creep curve under tension coincided in direction with the tensile load applied from the very beginning of loading. It follows then that the drying creep of concrete under tension actually represents creep strain not shrinkage (as in the case of compression) [Bissonnette and Pigeon, 1995]. This means that the previously proposed two-mechanism model (Equation 2.4) cannot be valid, and that the mechanism of drying creep cannot be creep-induced shrinkage [Kovler, 1999]. The author hypothesizes that UHPC might be significantly different than what was obtained by Kovler [1999]. This is mainly due to the significant difference in the water-to-cement ratios between both the cases (i.e. typically less than 0.2 for UHPC and 0.7 in the previous study [Kovler, 1999]). This will likely reduce the
2-25

swelling deformation component in the case of UHPC, and hence, it would be of specific interest to verify applicability of this model to the tensile creep of UHPC.
2.3.2 Experimental Methods Very few studies have focused on studying the tensile creep behavior of concrete
[Umehara et al., 1994; Gutsch and Rostasy, 1994; Kovler, 1994; Bissonnette and Pigeon, 1995; Kovler, 1995; Kovler et al. 1999; Kovler, 1999; Altoubat and Lange, 2001; and Tao and Wiezo, 2006] either because of the relative complexity of conducting tensile creep tests as compared to conducting standardized compressive creep tests, or because of a perception that the behavior of any concrete in tension can be completely ignored. The latter has been justified because steel or other reinforcement is provided to resist tension after cracking; therefore tensile behavior is not important.
Among those previously mentioned studies, only Bissonnette and Pigeon [1995] reported test results over more than a year, while all the other studies reported results for a maximum period of about 200 hours only. The use of short-duration tests is likely because of the complexity of the test setup used in these studies as discussed in the following section.
2.3.2.1 Jalal Vakili [1983 and 1984] One of the first test setups used to characterize the tensile creep behavior of concretes
was the one developed and used by Vakili [1983 and 1984] where an electro-hydraulic loading system was used for short term tensile creep tests on asphaltic concrete. No schematic diagram of the equipment was provided.
The time-dependent loads were controlled by the shape of a load-time curve mounted on the drum of a data track. A pair of LVDTs attached to the central portion of the specimen
2-26

measured the vertical deformation. The specimen size was 100 mm (4-in) in diameter and 230 mm (9.0-in) in height. Specimens were bonded to aluminum end caps using a structural epoxy resin. A jig was used to ensure that the end caps were properly aligned and the adhesive was allowed to cure for 12 h before specimens were removed from the jig. Uniaxial tensile creep tests, were performed at 20C. A constant stress was applied instantaneously to the specimen and the deformations were recorded for a period of 190 seconds. After unloading, each specimen was allowed to rest for 2 h and then the permanent deformation was recorded. The same setup also was used to observe the experimental recovery curve of asphaltic concrete where a tensile stress of 0.0307 MPa (4.45 psi) was applied to the specimen for 217 seconds. The load was then removed and the deformations were recorded for an additional period of 418 seconds. In both of these studies, no additional information about the accuracy of measurements was provided.
2.3.2.2 Kovler [1994] A general view of the experimental device developed by Kovler [1994] is shown in
Figure 2.22. The device allows two dog bone specimens with a net cross section of 40 x 40 mm and a working length of 1,000 mm (1.57 x 1.57 x 39-in) to be mounted horizontally on the laboratory table, one for restrained shrinkage testing, and the other for free shrinkage testing.
The right end of each specimen is fixed. The displacements of the left movable grips were measured by a linearly variable displacement transducer (LVDT). Each displacement measurement cycle consisted of 256 measurements during 0.5 seconds, and the result was averaged and recorded. Such a procedure permitted very high accuracy and reproducibility of linear displacement measurements not less than 0.1 m (0.0000039-in).
The stresses in the restrained specimen were measured by means of a load cell with an accuracy of 0.0003 MPa (0.0435 psi). The restrained specimen was loaded by a computer-
2-27

controlled stepper motor according to a special program. The compensation cycle began when the absolute value of the total restrained specimen strain exceeded 5x10-6. Load was applied to recover the shrinkage strain. Two different rates of loading were chosen. In the initial period of fast evaporation and resultant drastic growth of shrinkage strain (period A), the loading rate was accepted to be 0.003 MPa s-1 (0.435 psi s-1), and in the following stable stage (period B) 0.001 MPa s-1(0.145 psi s-1)
The frequency of stress and strain measurements in these stages of the shrinkage process was chosen differently as well. The time interval between measurements in period A was taken as 180 seconds, and in period B it was 600 seconds. The duration of period A was 1-2 h.
Loading rates associated with shrinkage rates were chosen to exclude premature specimen failure. For the same purpose the grips were of a special shape, characterized by gradual widening of the internal part to eliminate any stress concentrations inside.
The possible physical eccentricity of the specimen, caused by non-uniform strain development in the net cross section, was checked and excluded. The movable grips were joined with spherical hinges. During the test the uniformity of progressive displacements of the grip was controlled on the left and right grip sides by means of two mechanical strain gages of accuracy 1mm (0.039-in): when the difference in their readings was more than 5 mm (0.20-in), the center of the hinge was moved horizontally by means of precise screws to the necessary direction. The measured value of the friction force did not exceed 20N (4.5 lb), and was therefore neglected at data analysis.
The uniaxial tensile loading apparatus developed in the Israel Institute of Technology (Figure 2.22) allowed performing two tests:
2-28

Restrained shrinkage test: in this test, one gripped end was fixed and the other was connected to a motor through universal joint. This system is a computer controlled closed loop one. When shrinkage occurs and its level is approaching a strain of 5x10-6, the motor automatically starts the motion to pull the specimen back to the initial position, to keep the length of the specimen constant at 1,000 mm (39.4-in). The load cell records the load induced in this motion.
Total strain (tot) of a restrained shrinkage specimen consists of three strain components: elastic strain (e), autogenous shrinkage strain (sh) and creep strain (cr). The total strain in the restrained shrinkage test is zero.
tot = e+sh+cr = 0.........................................(2.6) sh is directly measured from the free autogenous shrinkage test. e is calculated by the accumulation of the elastic strain increments in each cycle of the loading by the motor movement. The creep strain is obtained as the difference between free autogenous shrinkage strain of a free specimen and the cumulative strain of the restrained specimen (Figure 2.23 (a)). Constant load creep test: this test is similar to conventional creep tests in compression. The cast concrete was also sealed, and twin specimens were cured for 24 hours. Thereafter, a tensile stress was applied to the movable grip. The closed loop computer-controlled system was programmed to move the grip to maintain the creep load constant. In this test, the measured total strain of the loaded specimen is the sum of the autogenous shrinkage strain and the creep strain.
tot = sh+cr .................................................(2.7) Thus, creep strain was obtained by subtracting free autogenous shrinkage strain from the measured strain of the loaded specimen (Figure 2.23 (b)). It is important to develop a high accuracy method for displacement measurement for this kind of test because the absolute values of the deformations will be close to zero. This means that
2-29

the total deformation of the specimen, equal to the sum of shrinkage and creep components, should be compensated by instantaneous deformations quite frequently during the test, in order to provide more gradual growth of tensile stress. If this condition is not met, larger and instantaneous compensations of deformation will be needed, and they may cause premature failure. Failure will occur when the tensile strength of the material is less than the value of the tensile stress induced in it at a given time. The physical eccentricity in the specimen caused by the non-uniformity in deformation distribution over the cross-section, or by stress concentration in the grips, can also lead to premature failure. Tensile test schemes are very sensitive to these factors [Kovler, 1994].
Thus, by means of a uniaxial restrained shrinkage test, simultaneously with a free shrinking companion specimen, it is quite possible to obtain a variety of mechanical characteristics of early age concrete in uniaxial tension.
Kovler [1994] designed this apparatus to meet the following requirements: (1) high accuracy measurement of linear displacement, (2) complete automation of the experiment, both in data registration and in governing the specimen loading, and (3) exclusion of any premature specimen failure, which may happen due to irregularity of loading, large load steps, or eccentricity of the specimen and stress concentration in the grips.
2-30

(a)
(b)
Figure 2.22: Experimental device to measure tensile deformation of concrete, (a) schematic description of the restrained shrinkage testing, and (b) specimen grip [Kovler, 1994]
However, no information was given about the method of calculating the friction between the specimens and the table. Also, the literature reviewed where the same or very similar setups was used [Kovler 1994, 1995; Kovler and Bentur, 1999; Kovler, 1999; Salah and Lange, 2001; and Tao and Weizu, 2006] showed that this setup was only used for short-term creep tests with a
2-31

maximum duration of 7 days. Such short duration tests may have been conducted because of the limited number of specimens that could be tested at a given time the high cost of the apparatus, and the space inefficiency related to the apparatus.
According to ASTM C 512 "Standard Test Method for Creep of Concrete in Compression", no fewer than two specimens should be tested for creep and another two for shrinkage from a given batch under each test condition. The author believes that achieving that with this setup would be impractical for long-term tests because of the space restrictions.
2.3.2.3 Bissonnette and Pigeon [1995] Figure 2.24 shows the dead-load lever arm device (4:1) that was built at Universite Laval
[1995] to study the tensile creep at early ages of ordinary, silica fume and fiber reinforced concrete. This device allowed 12 concrete specimens to be tested for creep at the same time. The load was applied to the prismatic specimens 50x50x700 mm (2.0 x 2.0 x 27.6-in) through precision-made steel plates anchored at the ends of the specimens with threaded rods, these plates were placed and aligned in the PVC molds prior to casting. At both ends of the specimens, the load is transmitted through a hinge (Figure 2.24). The test setup was placed in a temperature and relative humidity controlled room (232oC and 505% R.H.).
The strain measurement device, similar to those used to determine the static modulus of elasticity of concrete, is also shown in Figure 2.24. Two frames were attached to the concrete specimens, 508 mm (20-in) apart (gage length). The upper frame was fixed, and the lower one was hinged, rotating around the rear rod (hinged at both ends) as the specimen deforms. The frames were attached to the specimens by means of brass plugs embedded in the concrete at thetime of casting. For the rear and front rods, a special grade of stainless steel was used to
2-32

(a)
(b)
Figure 2:23: Creep strain calculated from the data of (a) restrained autogenous shrinkage tests, and (b) constant load creep tests [Kovler et al., 1999] 2-33

avoid differential thermal strains between the measurement device and the concrete specimen. The thermal expansion coefficient of this steel 9.9x l0-6/oC (5.5x l0-6/oF) is close to the usual values for concrete 6-to-12x 10-6 /oC (3.33-to-6.67x 10-6 /oF). The dial gage was mounted on a lever arm (4:l), which means that the deformation that was measured represented four times the real deformation. The precision of this device is 1 m/m. Its performance was assessed through comparative shrinkage tests. The free shrinkage of three concrete specimens cast from the same batch was measured during 2 months with this device, and a standard extensometer was used to measure the free shrinkage of three other specimens of the same batch. The results showed very good agreement.
According to ASTM C 512 "Standard Test Method for Creep of Concrete in Compression", a creep loading frame should be capable of applying and maintaining the required stress on the specimen, despite any change in the dimension of the specimen. This setup as described before, in fact adequately satisfies this requirement as well as other test requirements like load centricity with the specimens.
It is the author's opinion that as the loading capacity of this system was limited to 225 kg (500 lb.), and as maintaining a typical stress/strength ratio of 0.4 for ultra-high strength concretes with specimens of the same size requires a load of approximately 1,500 kg (3,300 lb), applying this system to UHPC becomes practically unfeasible.
2-34

Figure 2.24: Creep test apparatus and strain measurement device [Bissonnette and Pigeon, 1995]
2.3.2.4 Bissonnette [1996] Attempts were made in order to achieve higher loads by modifying the previously
discussed setup by Bissonnette and Pigeon [1995]. First Modification:
Figure 2.25 shows the first of these attempts where a water tank was used as the sustained load in order to control the loading rate, and a system of pulleys was utilized to achieve an amplification factor of 32:1 (4:1 x 8:1) (loading capacity of 3,600 kg (7,900 lb)) instead of 225 kg (500 lb) as in the previous setup. The load was applied to the prismatic specimens 70 x 70 x 400 mm (2.76 x 2.76 x 15.75-in). The major problem associated with this setup was the frictional losses created in the system of pulleys. These losses resulted in inconsistent loads delivered from the loading system to the specimens. Sometimes these losses decreased the load measured at the load cell at the bottom of the specimens to about 50% of the theoretical applied load.
2-35

Second Modification: Previous problems were overcome in the modified setup shown in Figure 2.26. The setup
allowed 18 specimens for creep and another 18 specimens for shrinkage to be tested at the same time. Figure 2.26 shows a detailed description of the setup.
Loading frame: The device was made of six independent loading units, each of them having a capacity of three specimens in series mounted on a rigid steel frame (Figure 2.26). The load, generated by a pneumatic jack and amplified with a 6:1 lever arm, was transmitted to (and within) the specimen series through hinged rods. Each loading unit is controlled individually with a pressure regulator and the actual load imposed to the specimens is measured with a loadcell inserted between the lower specimen of the series and the lever arm. As air is a compressible fluid, air pressure was recommended to be checked once a day during the first week of the test and then twice a week thereafter.
Load transfer plate: Load transfer plates consisting of 25 mm (1.0-in) thick steel discs with a diameter of 75 mm (3.0-in) were installed at each end of the specimens during casting. The discs were anchored into the concrete with 6 threaded steel rods (d = 4.8 mm (0.189-in); 1024 NC) having different lengths (80, 105 and 130 mm (3.15, 4.13, 5.12-in)) so as to avoid the creation of a weakness plane. For the companion specimens, the configuration is similar, except for the discs that were made from plastic.
Specimen size: The specimens were cylinders having a diameter of 75 mm (3.0-in) and a length of 460 mm (18.11-in).
Extensometer: The extensometer used is shown in Figure 2.27. The extensometers used are resistive strain gage mounted on half-rings which work like a conventional load cell. The monolithic aluminum half-rings (D = 160 mm (6.30-in)) are composed of two stiff legs (w = 30
2-36

mm (1.18-in); t = 6.3 mm (0.25-in)) connected through a flexible part (l = 30 mm (1.18-in); w = 15 mm (0.59-in); t = 2.7 mm (0.0465-in)) instrumented with four resistive strain gages in fullbridge configuration. The bridge potential difference is linearly proportional to the diametrical displacement of the half-ring (k = 5 V/m).
The extensometers are installed on the specimens by means of two aluminum plugs glued with an epoxy binder on the concrete surface. With an axial adjustment screw located on one of the two plugs, the extensometers are lightly tightened in place. The relative precision of this device is 110-6 (absolute precision of 0.2 m (0.0000079-in) with a gage length of 210 mm (8.27-in)).
Procedure: Immediately before loading the creep specimens, the tensile strength of concrete was determined in accordance with ASTM Method C496, "Splitting Tensile Strength of Cylindrical Concrete Specimens". After determining the tensile strength, three creep specimens and three companion shrinkage specimens were taken out of the conditioning room and installed on the testing apparatus. The creep specimens were assembled in the loading unit, and a pair of half-ring extensometers was installed, adjusted and zeroed on all specimens. Then, the creep specimens were loaded up to a specified fraction of the tensile strength on a shelf within the loading frame, the companion specimens were put upright to avoid any restraint.
Strain reading sequence - Readings were recorded automatically throughout the experiment with a data acquisition system. During the loading process, data were recorded at a rate of 1 Hz (1 s-1). After the specified load was reached, the recording sequence evolved as follows: 0 < t < 1 hour: one reading every 5 minutes, 1 t < 6 hours: one reading every 15 minutes, 6 t < 24 hours: one reading every hour, and t 24 hours: one reading every four hours.
2-37

The author believes that among the previously discussed test setups, this setup seems to be the most suitable for UHPC applications due to high load capacity as well as the ability of gradual application of loads and real time acquisition of loads and deformations. However, a simpler strain measurement technique may be desired.
System of pulleys (amplification 8: 1)
Water tank (capacity of 75
L)
Figure 2.25: Initial modification setup where a water tank and a system of pulleys were used for loading, dimensions are in (mm)
2-38

Figure 2.26: Tensile creep apparatus loading unit
Figure 2.27: Tensile creep strain gage 2-39

2.3.3 Key Experimental Results Bissonnette and Pigeon [1995] have emphasized the importance of studying tensile creep
of concrete. They state that the only two strain components that can counteract any shrinkageinduced strains in a partially or fully restrained concrete member are the elastic tension strain and the tensile creep strain. The elastic tension strain capacity in the case of concrete is very small (i.e. 100 to 200 microstrain) and the tensile creep strain is not negligible, particularly when concrete is allowed to dry under load. In this study, six concrete mixes that varied in w/cm, silica fume content, curing period, and fiber content were investigated (Table 2.2). The applied stress was 0.77 MPa (112 psi), and 1MPa (145 psi) for specimens loaded after 1day and 7 days respectively. The results of the tensile creep tests showed that among the parameters tested (w/cm, cementitious materials characteristics, age of loading, and use of fibers), both w/cm and age of loading had the most significant effect of the tensile creep behavior (Table 2.3). However, the increase in tensile creep noticed upon incorporating silica fume and steel fibers in concrete mixes is of specific importance in the case of UHPC as it typically contains both. In this study, no explanation was provided to the effect of incorporating silica fume, but it was later attributed to the smooth and spherical shape of silica fume particles that may facilitate deformation at early age before they react [Kovler et al., 1999]. On the other hand, the increase in creep measured for mixes incorporating steel fibers was attributed to the higher air content measured for these mixes (i.e. 8.8% and 3.0% for mixes with micro and macro fibers respectively). These results need to be further investigated especially in the case of UHPC used in prestressed structural members subjected to high loadings at early ages.
2-40

While not examining UHPC, but rather a "microcrete", a study by Kovler [1999]

measured tensile creep over 5 days. The material used in this study was ordinary portland

cement concrete with crushed dolomite gravel with maximum particle size of 7mm (0.276 in)

Table 2.2: Testing program [Bissonnette and Pigeon, 1995]

Composition

Testing

Mix

w/cm 0.55 0.35

Cementitious Materials
Cement Silica fume Type I (7% by weight)

Steel fibers

Macro

Micro

(0.63 in long) (0.12 in long)

Age of loading
1d 7d

Sealed or
unsealed

1





both

2





unsealed

3





unsealed

4





unsealed

5







unsealed

6







unsealed

Table 2.3: Tensile creep results [Bissonnette and Pigeon, 1995]

Specific total

tensile creep

Summary of the Results

Mix

Sealed or unsealed

Age of loading

m/m/MPa (m/m/ksi)

@ age of 40

days

1

1 day 7 days

2

Unsealed

1 day 7 days

3

RH= 50%

1 day 7 days

4

1 day 7 days

149 (1,030) 100 (690) 100 (690) 59 (410) 165 (1,140) 131 (900) 165 (1,140) 69 (480)

The specific total creep increased upon increasing the w/cm (mixes 1 & 2; 3 & 4 (7-day only))
The specific total creep decreased upon delaying loading from 1 day to 7 days (all mixes)
The specific total creep increased upon replacing 7% of the cement mass with silica fume (mixes 1 & 3; 2 & 4)

unsealed RH= 50%

1 day

131 (900)

1

sealed unsealed RH= 50%

1 day 7 days

15 (100) 100 (690)

The specific total creep decreased upon sealing specimens despite the age of loading

sealed

7 days

41 (280)

1

100 (690) The specific total creep increased upon

5 6

unsealed RH= 50%

7 days

150 (1,035) 190 (1,311)

incorporating steel fibers in the mix, it also increased upon decreasing the size of fibers due to the increased void ratio that resulted from

incorporating fibers.

2-41

used as coarse aggregate, quartz sand of fineness modulus of 1.76 used as fine aggregate, and a water-to-cement ratio of 0.7. The typical compressive strength was 30 MPa (4,350 psi). One day old concrete was exposed to drying in a hot climate simulator that had temperature of 29 to 31oC (84.2 to 87.8 oF) and a relative humidity of 32 to 38% and RH. The tensile strength after one day was 1.20 MPa (174 psi). The applied tensile stress was between 0.45 MPa (65 psi) (0.375 stress/strength level) and 1.10 MPa (160 psi) (0.917 stress/strength level). Tensile strength test performed before and after the creep tests did not show any damage in the material, nor decrease but increase in the Young's tensile modulus of elasticity. It also showed an increase in the tensile strength with time despite the drying condition. These were all attributed to the continuous formation of hydration products while curing, especially with this high w/cm (i.e. w/cm = 0.7).
Altoubat and Lange [2001] studied tensile basic creep of concrete at early age (160 hours). In this study, mixes that varied mainly in their w/cm (0.4 and 0.5) and steel fiber content were investigated (Table 2.4). The constituent materials used were Type I cement, crushed lime stone with maximum size of 25.4 mm (1-in), and natural sand with fineness modulus of 2.2. Steel fibers with flared ends, 30.48 mm (1.2-in) long, and aspect ratio of 75 were used at 0.5% (volume fraction of the whole mixture). A moist curing technique was adopted where samples were covered with wet cloths throughout the test to suppress early-age autogenous shrinkage so that basic creep could be measured. This technique successfully suppressed the autogenous shrinkage as the measured relative humidity of the concrete under moist-cover condition was almost constant across the sample. This was also confirmed by zero shrinkage measured due to initial swelling [Kovler, 1996 and Altoubat and Lange, 2001]. The results summarized in Table 2.4 bring up two important problems that need further microstructural investigation. These are: (1) the effect of the w/cm on the tensile-creep behavior of concrete, as there is contradiction
2-42

between these results and the results reported previously [Bissonnette and Pigeon, 1995], (2) the effects of incorporating fibers in concrete on its tensile-creep behavior, and (3) whether drying is permitted or prevented. In addition, the same study showed that although the initial creep rate of plain concrete is higher than that of fiber-reinforced concrete, the creep function (solidification theory) of plain concrete stabilizes earlier than that of fiber-reinforced concrete, suggesting that steel fiber reinforcement provides stress relaxation for a longer period of time (Figure 2.28). However, this testing only lasted for a maximum of 160 hours, and thus there is a necessity for a more detailed and longer investigation to validate this concept over a longer period of time (i.e. more than 1 year).
Tao and Weizu [2006] studied the tensile basic creep of high strength concrete over a period of about 100 hours. In this study, concrete mixes with and without silica fume or fly ash were investigated. The water-to-cementitious ratio for all mixes was 0.35. The constituent materials used were Type I cement, crushed lime stone with maximum size of about 1 in, and natural river sand. For the silica fume binary blended mix, 6% of ordinary portland cement by weight in the control mix was replaced by silica fume, while for the fly ash binary blended mix, 30% of ordinary portland cement weight in the control mix was replaced by fly ash. Results from Tao and Weizu [2006] showed the silica fume concrete exhibited a greater tensile creep as compared to the control concrete, while the fly ash concrete showed the opposite (Figure 2.29). However, no explanation for these phenomena was provided.
Recently, Bissonnette et al. [2007] studied the effect of stress level, age of loading, fiber reinforcement, and the volume fraction of the cement paste on the tensile creep of concrete. In this study, tensile creep tests were carried out for a period up to 168 days. Tests were thus performed at stress levels ranging from 0.20 to 0.50 of the short-term strength under drying
2-43

conditions at 50% RH. Loading started at 7 or 28 days of age. Crimped or hooked steel fibers were used at 0.50 and 1.0% volume fraction. In addition, the cement paste of the matrix was 22%, 27%, and 32%. Results showed that fiber reinforcement did not have significant effect on drying shrinkage. On the other hand, as expected, tensile creep increased upon increasing the stress/strength ratio at loading and upon delaying the loading process (i.e. 7 vs. 28 days of age at loading). As for the effect of fiber reinforcement on tensile creep, using 0.50% volume fraction of hooked fibers resulted in an increase of about 30 to 35% in tensile creep when compared to the non-reinforced mixes, while the same dosage of crimped fibers lead to a reduction of 30 to 35%. However, doubling the dosage of crimped fibers resulted in increase in tensile creep. As a result, the authors suggested that an important part of the creep deformations could be taking place in the more porous paste-fiber interfacial areas. This means that the influence of fiber on tensile creep is a balance between their restraining effect and the viscoelastic properties of the surrounding cement paste. Results of the same study also showed a decrease in the tensile creep upon increasing the paste content, which is quite the opposite of what have been reported for years in the case of compressive creep. This result suggests that a significant part of tensile creep deformation could be taking place in the weak paste-aggregate interface.
However, according the author's knowledge, a microstructural study to confirm these recently proposed hypotheses has not been carried out yet and thus need further investigation.
2.3.4 Mechanisms According to the author's knowledge, the study by Bissonnette et al. [2007] was the first
comprehensive attempt to try explaining mechanisms of tensile creep of concrete. Bissonnette et al. [2007] recognized three main mechanisms: water seepage, viscous
shear, and microcracking. The assumptions of the seepage theory were consistent with the
2-44

occurrence of basic creep in tension. Under tension, the gel porosity increases and tends to adsorb some of the free water from the larger capillary pores, which in turns results in expansion. However, the seepage theory was not able to explain the additional tensile creep that took place under drying conditions that was measured in the same study. However, the same study proposed that drying creep in tension can be explained by the viscous shear theory. That is, when concrete

Table 2.4: Basic tensile creep results of moist-covered specimens [Altoubat and Lange, 2001]

Mix w/cm

Steel fibers (volume fraction)

Creep strain (m/m)
@ age of 100 hours

Summary of the Results

1

0.4

0

2

0.4

0.5

3

0.5

0

4

0.5

0.5

Effect of fibers

65

The incorporation of fibers in the wet condition

decreased the initial basic creep as they

controlled microcracking, and engaged more

volume of the matrix in stress transfer.

Under drying conditions, the previous effects of

50

using fibers are not evident likely because under

these conditions, more surface microcracking

occurs. This agrees with the results obtained by

Bissonnette and Pigeon (1995).

Effect of w/cm

The tensile basic creep increased upon

36

decreasing the w/cm. This may suggest that the

tensile creep behavior at early age is governed

by different factors than in mature concrete.

However, no mature concrete was investigated

in this study.

26

Tensile creep in fiber-reinforced concrete

seemed more sensitive to the w/cm.

2-45

Figure 2.28: Effect of fiber reinforcement and w/cm on creep [Altoubat and Lange, 2001]
Silica fume blend Control Fly ash blend
Figure 2.29: Specific total creep under isothermal conditions [Tao and Weizu, 2006] 2-46

is loaded in tension and allowed to dry, the adsorbed water layers are disturbed by both the applied load (adsorption of water in the pores) and the water loss process. This simultaneous action increases shear between gel particles and thus facilitates the sliding or flow of gel particles one against the other and results in an overall increase in tensile creep. In addition, the effect of microcracking was investigated qualitatively by measuring the modulus of elasticity before and after creep. However, the results from this study did not show significant reduction in the modulus of elasticity after creep. This again emphasizes the need for a more detailed microstructural investigation in order to quantify the effect of microcracking on creep in tension.
2.4 Use of UHPC Field Construction and Performance
2.2 Field Casting Up to this time, field use of UHPC in North America has been very limited. The first
highway bridge structures to be cast in the U.S. were two AASHTO Type II girders cast in fall of 2000 as part of the Federal Highway Administration's Ultra-High Performance Concrete Research Program (Graybeal, 2006 b). These girders were cast by Prestress Services of Kentucky, Inc. in Lexington, KY. The girders were 80 ft (24.4 m) and 30 ft (9.2 m) long and were cast on the same bed with the same strands running through both.
According to Ben Graybeal of FHWA, this first casting of UHPC was relatively simple in comparison with other castings of UHPC due to the simple shape of the beams, the low amount of surface area exposed to drying, and the lack of any shear reinforcement in the beams. The UHPC was mixed in multiple batches and transferred to a Ready-mix truck. The Ready-mix truck was used to combine the batches and allow for one continuous pour without cold joints. The material was poured from one end of the beam and allowed to flow toward the other. As the
2-47

material flowed down the beam, the pour location was gradually moved in the direction of the flow, always making sure to cast the material into itself and not directly into the empty formwork. As soon as the forms were filled, plastic was pressed into the surface of the plastic concrete to prevent dehydration.
The beams were allowed to set and reach at least 11ksi (76 MPa) in strength before the beams strands were cut. After the beams were cut down, they were successfully steamed at 200 F (95 C) for 48 hours as specified by the manufacturer. Post-curing cylinder strengths of 28 ksi (193 MPa) were observed in 3 x 6 in (76 x 130 mm) cylinders. Cores were also taken of the girders after they were tested at FHWA. The web core strength averaged 30 ksi (204 MPa) measured from 3 x 6 in (70 x 140 mm) cores, and the flange core strength averaged 29 ksi (203 MPa) measured from 4 x 8 in (101 x 202 mm) cores. While the section was not optimized for use of the high strengths achieved in UHPC, this first casting showed that it was possible to cast full-size highway girders with UHPC.
In late 2003 and early 2004, four 70-foot long test girders were cast by Prestress Services of Kentucky as the next phase of FHWA's work with UHPC. Unlike the previous girders, these girders used a highly-optimized bulb double tee section dubbed the "pi" section shown in Figure -30 (Graybeal 2004). The section was structurally optimized to take full advantage of the high compressive and tensile strengths of UHPC. Due to the unconventional shape of the girders, however, there was some difficulty in the casting of these sections.
2-48

Figure 2-30: The UHPC pi girder section was highly structurally optimized, but was complex to cast (Graybeal 2004).
The girders each required 4 batches of UHPC from the plant's mixer. The first and fourth batches were held in one Ready-mix truck while the second and third were held in another. This division allowed each truck to have the same average age of material. The two trucks then filled the two tees of the girders simultaneously using the method of placing UHPC into itself described above. When the tees were approximately 2/3 full and the trucks reached the far end of the beams, the trucks both began placing material into a trough that allowed for the deck concrete to be placed uniformly across the top of the beam. The trough was supported by and slid along the external vibration rails that were attached to the forms. This method of casting was used to reduce the preferential orientation of fibers caused by casting from a single point. While it was unconventional, this casting method was effective in placing the deck.
The larger exposed surface area in the pi girders required more immediate covering of the beams with plastic to prevent dehydration of the plastic UHPC. While no quantitative research has been done on how long the material can remain uncovered without deleterious
2-49

effect, 5 minutes is the rule of thumb for maximum time uncovered. Plastic was successfully placed on all of the beams after casting, but one beam experienced some drying shrinkage in an area where the plastic did not maintain contact with the UHPC.
According to Ben Graybeal, one of the biggest difficulties with the casting of the pi girders was timing the release of the center portion of the formwork (see Figure 2-31). The forms were designed to allow for the center portion to be released when the deck concrete had achieved sufficient strength to be self-supporting. This release also had to occur before the autogenous shrinkage in the deck caused restraint cracking longitudinally along the deck. The window between these times was not known precisely, but was thought to be less than three hours.
Figure 2-31: The formwork for UHPC pi girders was designed to release the internal restraint once deck was self-supporting (Graybeal 2004). The trough for casting of the deck can be seen above the formwork.
2-50

The time chosen for release was based on a qualitative measure of the deck penetrability. A 1/16 in. (1.6 mm) diameter flat head penetrometer was manually forced into the concrete. Penetrations of less than 1/8 in. (3.2 mm) were considered to be sufficient for form release. This method of determining form release time was obtained from trial and error evaluations of 5 pishaped sections cast prior to the casting of the full girders. While the test was not very precise, all four girders had the forms released without inducing any longitudinal shrinkage cracks in the decks.
Timing of the strand release was another difficulty observed with the casting of the pi girders. A cylinder compressive strength of at least 10 ksi (69 MPa) was required before strand release. Three of the girders were released at a compressive strength of 13 ksi (90 MPa) without incident. The second girder to be cast was released at a cylinder compressive strength of 11 ksi (76 MPa). The negative bending moment induced in the section by the prestressing at strand release caused a midspan flexure crack that went from the surface of the deck down to close to the level of the strands.
Two possible solutions were posed to this problem. For the last two girders cast, strands were added at the top of the tees as shown in Figure -30. This reduced the tensile stresses in the top of the beam at release. Ben Graybeal also suggested that release strength of 14 ksi (97 MPa) may be a better design release strength due to the more steady nature of the strength gain at this strength than at lower strengths. This would mean less variability between cylinder test strength and beam concrete strength. This higher design strength would also increase the bed time required for UHPC beams.
The first casting of UHPC pi girders showed the potential to create optimized precast prestressed bridge structures with this new material. This casting also showed some of the new
2-51

innovations that have to be used when casting this unique material. Lastly, the casting of the UHPC pi girders showed potential difficulties that must be addressed prior to casting the material.
One of the most highly publicized uses of UHPC was the roof canopy shells of the Shawnessy Light Rail Transit (LRT) Station in Calgary, Alberta, Canada (Vicenzino et al. 2005). These 16 x 20 ft (5 x 6 m) canopy shells were originally to be constructed from steel, but 0.75 inthick (20 mm) UHPC was used instead to provide a very open, architecturally pleasing design that still exhibited high resistance to Canada's aggressive climate. This innovative design received two PCI Design Awards in 2005 The Harry H. Edwards Award for innovation and carrying the industry to the next level of technology and Best Custom Solution.
From a field use perspective, this use of the material was atypical. First, it was done by the manufacturer of the UHPC, so greater familiarity with the material was expected. Second, the casting was done using very innovative, non-standard methods of precasting: injection and displacement molding. There were no screeded or hand-finished surfaces. All exposed surfaces were cast against formwork. Finally, the casting was performed at an indoor plant where greater environmental control would be expected than typical exterior construction of bridge girders.
Despite the many unusual aspects of this project, several important lessons were learned that are applicable to general use of this material. First, it was possible to mix the material in a full-size batch plant mixer in volumes greater than 1 yd3. Second, thermal treatment of the material at 140 F (60 C) was able to be achieved in the precasting plant. Finally, it was shown that precast UHPC could be used as a design alternative to steel when normal or high performance concretes would not suffice.
The first UHPC bridge in the United States was the Mars Hill Bridge in Wapello County, Iowa (Moore and Bierwagen 2006). This bridge consisted of three 110-foot-long UHPC girders
2-52

with a modified Iowa 45-inch bulb tee section. The two bids submitted for the project by local companies were higher than expected due to precaster concerns with material cost, batch time, long setting and curing times on beds, and placement difficulties. Because of these high costs, the girders were finally cast by Lafarge Canada Inc., Winnipeg Precast Division in Winnipeg, Canada on June 25, July 9, and July 16, 2005. The Winnipeg plant was an indoor plant, so some of the issues of temperature and humidity were more controlled than in an outdoor plant. Otherwise, the mixing and casting methods were the same as those detailed above.
The Mars Hill Bridge showed UHPC was a feasible bridge design solution. Specifically, this construction again showed that a precast plant could achieve the mixing volume of UHPC necessary for casting bulb-tee girders and that it could achieve the recommended 200 F (95 C) curing temperatures necessary for thermal treatment. The concerns of the local precasters showed one of the major difficulties with field use of UHPC gaining sufficient experience with the material to be able to price it competitively.
The Iowa Department of Transportation has also used a modified version of the pi girder for a new bridge in Buchanan County (Keierleber et al. 2008). This second generation pi girder (shown in Figure 2-32) was modified based on experience gained from the first generation pi girder. The deck and webs are thicker to allow for easier pouring and finishing. Holes for diaphragms and rounded corners were also added to the beams to enhance the transverse stiffness and load distribution of the structure.
The second generation pi girder was also cast at Lafarge Canada Inc., Winnipeg Precast Division in Winnipeg, Canada beginning in the first week of September, 2008. A drastically different method of mixing was used for these beams. Rather than using the plant's mixer to mix small batches that would then be mixed together in a Ready-mix truck, the material was initially
2-53

mixed in the Ready-mix truck. The mixing water was replaced with ice in order to keep the mix cool and to add additional friction to the mixing process. The biggest difficulty with this process was getting the constituent materials into the truck, but batching only took 45 minutes after the materials were added. This is a significant time savings when batch size is factored in.
The pi girders were cast from the ready-mix trucks in the same manner as the first generation pi girders. Casting went faster due to the increased web widths that allowed the material to fill the form more easily. The releasing of the internal form support was still not easy to time, but due to the increased deck and the addition of transverse rebar, it was not as critical as with the first generation pi girders. No restrained shrinkage cracking was observed on any of the beams. This casting showed significant improvements over the earlier pi girder casting and showed the ability of a plant to gain experience with the casting of this product.
Figure 2-32: The second generation pi girder was used for a UHPC bridge in Buchanan Co, Iowa.
The early full-scale field usages of UHPC in North America showed that while the material has the potential to be used in full-scale construction, it was very important for precasters to gain experience with the material. To date, only four known precasters in North-
2-54

America have any experience with casting of UHPC bridge components. Prior to the research program at Georgia Tech sponsored by the Georgia Department of Transportation, there had been no laboratory or field use of UHPC in Georgia.
2-55

3. Analytical Investigation
3.1 Introduction to Analytical Investigation
The purpose of the analytical investigation is to determine the effectiveness of using UHPC for precast prestressed bridge girders, specifically : (1) to determine the maximum span for AASHTO Type I Modified and PCI Bulb-Tee 28 (Figure 3.1) simply supported precast prestressed girders as limited by allowable stresses, ultimate strength and deflection requirements, when using UHPC; and (2) to determine the effect of the increased tensile strength of the concrete on the concrete shear capacity and to study the possibility of avoiding the use of stirrups.
The UHPC material is defined as a concrete having a compressive strength greater than 20,000 psi (138 MPa) and containing fibers in order to increase its tensile strength. The compressive strength for this material was varied from 20,000 psi to 29,000 psi (138 MPa to 200 MPa), and the flexural strength from 2,500 psi to 7,000 psi (17.2 MPa to 48.3 MPa). However, the flexural strength obtained using the 6 fc ' relationship was also used during the analysis of UHPC girders. The initial compressive strength of the concrete (at release) was considered 11,000 psi (75.9 MPa) for all cases. The creep and shrinkage losses were considered null for this material.
For all bridge analyses, the composite deck was taken as normal weight concrete with a compressive strength of 3,500 psi (24.1 MPa). The deck was 7-inch (178 mm) thick, and the girder spacing considered was 7 ft (2.13 m) on center. Pretensioning strands were 0.6-inch diameter 270 ksi low relaxation strands. They were assumed to be tensioned to 75% of the ultimate stress. Posttensioning strands were 0.5-inch diameter 270 ksi low relaxation strands which were tensioned to 70% of the ultimate stress. Reinforcement configurations containing only 0.6-inch diameter pretensioned strands, and configurations with both 0.6-inch diameter pretensioned strands as well as 0.5-inch diameter posttensioned strands in the same section were analyzed.
The GDOT bridge design program which is based on the 17th Edition of the AASHTO Standard Specification for Highway Bridges and a specially developed program were used for the bridge analyses.
3-1

For comparison, HPC is used for Type I and a BT 28 girders in sections 3.2 and 3.3 while UHPC is considered in subsequent sections. Various tensile strengths of UHPC are considered.
3.2 Flexural Analysis for Type I Modified Girders using HPC
3.2.1 Analysis Using HPC and Only Pretensioned Strands The first analysis considered HPC and the used only 0.6-inch diameter pretensioned strands.
The maximum span lengths for each compressive strength are presented in Figure 3-1. The span length limits as controlled by the deflection requirements of article 9.11.3.1 of the AASHTO specifications are shown in the graph for convenience (AASHTO, 1996). The maximum span length due to the deflection limit (hereafter termed Ld) increases as the compressive strength is increased because the modulus of elasticity follows the relationship 57,000 fc ' . The maximum span length for Type I Modified girder with an fc ' of 20,000 psi (137.9 MPa) satisfying the deflection limit of L/800 is 78 ft (33.8 m). A goal was to design prestressing reinforcement which would achieve this 78 ft (33.8 m) maximum span.
Girders solely using 0.6-inch diameter pretensioned strands were not able to reach the L/800 span length set by the deflection limit (Ld) . Moreover, increments in compressive strength beyond 18,000 psi (124.1 MPa) did not allow any increase in the length of the spans. The maximum span length achieved using concretes with a compressive strength of 20,000 psi (137.9 MPa) was 66 ft (20.1 m). The latter was about 85% of the span length based upon the deflection limits. The creep and shrinkage losses for this analysis were calculated using the equations for normal strength concrete.
The number of 0.6-inch diameter pretensioned strands for the maximum spans is presented in Figure 3.2. As expected, the longer the span, the greater the number of strands needed. For the 10,000-psi (69 MPa) compressive strength concrete, 16 strands were needed to reach a span length of 56 ft (17.1 m), whereas for the 20,000 psi (137.9 MPa) concrete a total of 28 strands were required to achieve the maximum span of 66 ft (20.1 m).
3-2

Max Span Length (ft)

80

75

70

65

60

Max. Span Length

55

lLd

50 10,000

12,000

14,000

16,000

18,000

20,000

Girder Compressive Strength (psi)
Figure 3-1 Maximum spans for a Type I Modified section using only 0.6-inch diameter pretensioned strands. Ld is the maximum span length set by the L/800 deflection limit.

3.2.2 Analysis Using HPC and Pretensioned and Posttensioned Strands According to the results obtained above, it was considered that by using the two-stage
prestressing process explained at the beginning of this chapter, it was possible to place more reinforcement and reach longer spans. This second group of analyses used the contribution of 0.6-inch diameter pretensioned strands and 0.5-inch diameter posttensioned strands. The placement of the pretensioned and posttensioned reinforcement is illustrated in Figure 3-2. In order to keep the web free to place a duct with the 0.5-inch diameter posttensioned strands, no more than 18 0.6-inch diameter pretensioned strands could be placed in the bottom flange of the Type I Modified section. Additionally, 2 0.6-inch diameter pretensioned strands were placed in the top flange. The diameter of the posttensioning duct as well as the eccentricity of the posttensioned strands varies according to the number of strands placed.
3-3

The maximum span lengths obtained from this analysis are presented in Figure 3-3; the Ld values are shown in the graph for convenience. As occurred in the previous analysis, the use of posttensioned strands did not provide a design that reached the maximum possible span length determined by the maximum deflection limit (Ld). The longest span for 20,000 psi (137.9 MPa) was 68 ft (20.7 m), 2 ft (0.61 m) longer than with pretensioning only. This small increase in length resulted because the final tensile stress governed the failure.

fc' slab = 3,500 psi
0.6-inch diameter pretensioned strands

Effective Slab Width: 84 in

Posttensioning duct. Diameter varies.

7 in
0.5-inch diameter posttensioned strands

18 0.6-inch diameter pretensioned strands (2 in. grid, 2 in. cover)

Varies

Figure 3-2 Reinforcement configuration at midspan for an AASHTO Type I Modified girder.

3-4

Max Span Length (ft)

80 75 70 65 60 55 50 10,000

12,000

14,000

16,000

Max Span Length
l Ld

18,000

20,000

Girder Compressive Strength (psi)

Figure 3-3 Maximum spans for a Type I Modified section using 0.6-inch diameter pretensioned strands and 0.5-inch diameter posttensioned strands.

From previous analysis, it was concluded that an increment in compressive strength beyond 18,000 psi (124.1 MPa) would not cause any important increase in span lengths. Yet, increasing the flexural tensile strength of the concrete would lead to longer spans.

3.2.3 Analysis Using HPC with Fibers and Pretensioned and Posttensioned Strands A third analysis was developed using concretes with a flexural tensile strength of 2,500 psi
(17.2 MPa). Flexural tensile strengths of this magnitude and up to 3,130 psi (21.6 MPa) were developed by Collepardi et al. using aggregates with a maximum size of 0.31 inches (8 mm) (Collepardi et al., 1997). The prestressing force for this analysis was considered to be applied in two stages (using pretensioned and posttensioned strands).
The controlling factors for the span lengths limited by the L/800 deflections are presented in Figure 3-4. As mentioned earlier, the compression at stage 2 (posttensioning takes place) no longer controlled the design. Instead, the tensile stress at service load as well as the ultimate strength conditions controlled due to the smaller number of strands used. In fact, for the 12,000 psi (82.8 MPa) concrete case, a smaller number of strands could be used to satisfy the allowable stresses and reach the 73-ft (22.3 m) span length, but due to the ultimate strength requirements

3-5

14 0.6-inch diameter pretensioned strands and 10 0.5-inch diameter posttensioned strands must be used.

Controlling Factor

7

6

5

4

1. Compression Stage 1 (pretensioning)

3

2. Compression Stage 2 (posttensioning)

3. Compression Stage 3 (deck is placed)

2

4. Compression Stage 4 (service)

5. Tension Stage 4 (service)

1

6. Ultimate Strength

0

72

73

74

75

76

77

78

79

Deflection Limit Span Length Ld (ft)

Figure 3-4 Controlling factor for deflection limit spans. Type I Modified section, using 0.6-inch diameter pretensioned strands and 0.5-inch diameter posttensioned strands. fr = 2,500 psi (17.2 MPa).

3.3 Flexural Analysis for Bulb-Tee 28 using HPC
The goal was to keep the height of the girder the same as that of the Type I Modified and to improve its deflection resistance. Mr. Craig Thompson of Standard Concrete, Atlanta, suggested that it would be easy and economical to modify a Bulb-Tee form. Therefore, the Bulb-Tee 28 section was considered as a possible alternative to the Type I Modified for use with HPC and UHPC. Note that this is a section very similar to that described in chapter 5 of this report.
3.3.1 Analysis Using HPC and Only Pretensioned Strands The Bulb-Tee 28 section is more effective in flexure due to its higher moment of inertia. An
initial analysis using HPC and only 0.6-inch diameter pretensioned strands was performed. The maximum span lengths obtained from these analyses are presented in Figure 3-5. The Ld values based on article 9.11.3.1 of the AASHTO specifications are shown in the graph for convenience (AASHTO, 1996). The Ld was 92 ft (28 m) for the 20,000-psi (137.9 MPa) concrete. This span is 17.9% (14 ft, 4.3 m) longer than the one attained using a Type I Modified section.
3-6

Max Span Length (ft)

95 90 85 80 75 70 65 60 10,000

12,000

14,000

16,000

Max. Span Length
l Ld

18,000

20,000

Girder Compressive Strength (psi)

Figure 3-5 Maximum spans for a Bulb-Tee 28 section using 0.6-inch diameter pretensioned strands only.

Girders using only 0.6-inch diameter pretensioned strands were not able to reach the Ld length based on the L/800 deflection limit. Once more, the results indicated that increments in compressive strength beyond 18,000 psi (124.1 MPa) did not produce any beneficial effect on the span lengths. The use of 0.6-inch diameter pretensioned strands only results in a maximum span length of 83 ft (25.3 m), which is 90% of the 92-ft (28 m) deflection limit span for the 20,000-psi (137.9 MPa) concrete. This value, compared to the one found for the Type I Modified section, is a 5% closer to the Ld.
The geometric characteristics of the Bulb-Tee 28 section allow for the placement of a higher number of strands, resulting in the achievement of longer spans. For the 10,000-psi (69 MPa) compressive strength concrete, 28 strands were needed to reach a span length of 74 ft (22.6 m), whereas for the 20,000-psi (137.9 MPa) concrete 42 strands were required to achieve the maximum span of 83 ft (25.3 m).

3.3.2 Analysis Using HPC with Fibers and Pretensioned and Posttensioned Strands As was done for the Type I Modified section, an analysis looking for the attainment of the Ld
span length using higher values of flexural tensile strength was performed for the Bulb-Tee 28 girders. However, for this section, a lower flexural tensile strength value of 1,850 psi (12.8 MPa) was found to be sufficient in order to reach the desired span lengths. Once more, the span length limits due to deflections were easily achieved and the number of strands was highly reduced. For

3-7

example, for the concrete having 20,000 psi (137.9 MPa) compressive strength, the 92-ft (28 m) span length was reached using 28 0.6-inch diameter pretensioned strands and only 12 0.5-inch diameter posttensioned strands, that is, 7.91 in2 (5105 mm2) of reinforcement area which is 73% of the area needed in the previous analysis to reach a span length of 87 ft (26.5 m) (using flexural tensile strength of 848 psi (5.85 MPa)).
The Bulb-Tee 28 section was more sensitive to the increase in flexural tensile strength not only because a smaller increase was needed to achieve the target span lengths, but also because the reduction in the area of reinforcement was greater than the one observed for the Type I Modified. In fact, when the Bulb-Tee 28 section was considered, the ultimate strength controlled the design for the cases having 12,000 psi, 14,000 psi and 16,000 psi (82.8 MPa, 96.6 MPa and 110.3 MPa) compressive strength, whereas the ultimate strength controlled only for the 12,000 psi (82.8 MPa) concrete when the Type I Modified section was used.
3.4 Flexural Analysis for Type I Modified using UHPC
The maximum span lengths found in the analysis are presented in Figures 3-6 and 3-7 for the cases when the allowable compressive stress due to effective prestress plus permanent (dead) loads is taken as 0.4 fc ' and 0.6 fc ' , respectively. The span length limit as controlled by deflections (Ld) as well as the span length limit as controlled by ultimate strength requirements (Lu) are shown in the graphs for convenience. The span length Ld increased as the compressive strength increased because of the increment in modulus of elasticity. A maximum span length limit of 81 ft (24.7 m) was reached for the 29,000 psi-concrete (200 MPa). On the other hand, the Lu, ultimate strength span limit, remains constant at a span length of 100 ft (30.5 m). This limit is reached using 18 0.6-inch diameter pretensioned strands and 31 0.5-inch diameter post-tensioned strands.
3-8

130

6*(fc')^0.5

120

ffr 2,500

Max. Span Length (ft)

110

ffr 4,000

100

fr 5,500

ffr 7,000 90

80
70 18,000

20,000

22,000

24,000

26,000

28,000

lLd lLu 30,000

Girder Compressive Strength (psi)
Figure 3-6 Maximum span length for Type I Modified girders made of UHPC and prestressed in two stages. Allowable compressive stress due to effective prestress and permanent load = 0.4 fc'.

150

6*(fc')^0.5

Max. Span Length (ft)

140

fr 2,500

130 ffrr 44,,000000
120

110

frr 5,500

100

frr 7,000

90
80
70 18,000

20,000

22,000

24,000

26,000

Girder Compressive Strength (psi)

28,000

LLd
LuUl 30,000

Figure 3-7 Maximum span length for Type I Modified girders made of UHPC and prestressed in two stages. Allowable compressive stress due to effective prestress and permanent load = 0.6 fc'

3-9

Figure 3-6 indicates that the effect of the flexural tensile strength of the concrete on the maximum span lengths attained was not important for values above 4,000 psi (27.6 MPa). For all of these cases, the compressive stress at stage 3 controlled the design and hence, the flexural tensile strength did not play a critical role. In fact, only very small increments (1 ft (0.3 m)) were noticeable for compressive strengths greater than 26,000 psi (179.3 MPa).
3.5 Flexural Analysis for Bulb-Tee 28 using UHPC
Figures 3-8 and 3-9 present the maximum span lengths found for the analysis of Bulb-Tee 28 sections, when the allowable compressive stress due to effective prestress plus permanent load was taken as 0.4 fc ' and 0.6 fc ' , respectively. The span length limits as controlled by deflections (Ld) as well as the span length limit as controlled by ultimate strength requirements (Lu) are presented in these graphs for comparison. The span length limit due to deflection increased as the compressive strength of the concrete increased because of the increase in modulus of elasticity following the relationship 50,000 fc ' . The maximum span based on the L/800 deflection limit was 94 ft (28.7 m) for the 29,000-psi (200 MPa) concrete. This value was 16% larger than the span length limit due to deflection determined for the Type I Modified section. However, the ultimate strength limit for this section was 96 ft (29.3 m), 4% shorter than the value found for the Type I Modified section. The reason for this drop in span length has to do with the fact that the two sections have the same depth (implying a similar moment arm), but the self-weight of the Bulb-Tee 28 section was greater due to its bigger cross section. The 94 ft (28.7 m) span was reached using 28 0.6-inch diameter pretensioned strands and 22 0.5-inch diameter posttensioned strands.
All the cases analyzed easily exceeded these limits, including the case when UHPC without fibers (flexural tensile strength = 6 fc ' ) was considered. These results suggest that due to the deflection limitations, girders constructed with UHPC might be over designed. The maximum span lengths achieved, when only allowable stresses were considered for the design, were 162 ft (49.4 m) and 165 ft (50.3 m) for the cases when the allowable compressive stress due to effective prestress and permanent load was taken as 0.4 fc ' and 0.6 fc ' , respectively.
3-10

Max. Span Length (ft)

170

6*(fc')^0.5

160

150

fr 2,500

140

ffr 5,000

130

120

ffr 4,500

110

ffr 7,000

100

90 80 70 18,000

20,000

22,000

24,000

26,000

28,000

Ldd Luu 30,000

Girder Compressive Strength (psi)

Figure 3-8 Maximum span length for Bulb-Tee 28 girders made of UHPC and prestressed in two stages. Allowable compressive stress due to effective prestress and permanent load = 0.4 fc'

170 160 150 140 130 120 110 100
90 80 70 18,000

20,000

22,000

24,000

26,000

28,000

6*(fc')^0.5 f fr 2,500 f fr 4,000 f fr 5,500 f fr 7,000 LdLd LuLu 30,000

Girder Compressive Strength (psi)
Figure 3-9 Maximum span length for Bulb-Tee 28 girders made of UHPC and prestressed in two stages. Allowable compressive stress due to effective prestress and permanent load = 0.6 fc'.

Max. Span Length (ft)

3-11

These values corresponded to the UHPC having a compressive strength of 29,000 psi (200 MPa) and a flexural strength of 7,000 psi (48.3 MPa). The maximum span length for the Bulb-Tee 28 section was 27.6% longer than the one found for the Type I Modified section when the allowable compressive stress was taken as 0.4 fc ' . For the case when this stress was considered as 0.6 fc ' , the span length increased by 17.9%. However, the deflection limit controlled the design.
Figure 3-8 indicates that the effect of the flexural strength on the achievement of longer spans was more significant for concretes with high compressive strengths. For these cases, the high values of compressive strength considered produced a change in the controlling condition from the predominant compression at stage 3 to tension at stage 4. However, as the flexural strength was increased, the compression at stage 3 became the controlling factor again, causing smaller span increments. In fact, the increase in span length when the flexural tensile strength was increased from 5,500 psi to 7,000 psi (37.9 MPa to 48.3 MPa) was only 50% of the span increase achieved when the flexural tensile strength was increased from 1,022 psi (value corresponding to 6 fc ' ) to 2,500 psi (7 MPa to 17.2 MPa).
Comparison of Figures 3-6 and 3-8 indicates that the Bulb-Tee 28 section was more affected by the changes in flexural tensile strength than the Type I Modified section. For the Bulb-Tee 28, the span length increased 19.1% when the flexural tensile strength was increased from 6 fc ' to 7,000 psi (48.3 MPa) for the 29,000-psi (200 MPa) concrete.
This increase was only 13.5% for the 20,000-psi (137.9 MPa) concrete. On the other hand, a span increase of 13.3% was attained for the concrete having a flexural tensile strength of 7,000 psi (48.3 MPa) when the compressive strength was increased from 20,000 psi to 29,000 psi (137.9 MPa to 200 MPa). These results indicated that for a Bulb-Tee 28 section, increases in flexural strength were more important than increases in compressive strength, when the compressive allowable stress due to effective prestress plus permanent load was taken as 0.4 fc ' .
The maximum span length for the Bulb-Tee 28 using 29,000 psi (200 MPa) UHPC was 97.6% longer than the span length attained using HSC (16,000 psi (110 MPa)) and only 0.6-inch diameter pretensioned strands. This difference was slightly larger than the 95% increase obtained for the Type I Modified section. Although longer spans are achieved using the Bulb-Tee 28
3-12

section (due to the larger moment of inertia), these results indicated that both sections were equally affected by the use of UHPC. As previously stated during the discussion for the Type I Modified section, the importance of using UHPC depends not only on the fact that the compressive and flexural tensile strengths are improved, but also on the extremely low creep and prestress losses of the UHPC.
Figure 3-9 indicates that the effect of flexural tensile strength increments on the achievement of longer spans also became significant at low values of compressive strengths when the allowable compressive stress due to effective prestress plus permanent loads is taken as 0.6 fc ' . The latter can be explained by the fact that the increment in the allowable stress from 0.4 fc ' to 0.6 fc ' changed the controlling parameter from compression at stage 2 to tension at stage 4.
3.6 Shear Analysis using UHPC
Figures 3-10 and 3-11 present the results of the flexural-shear analysis carried out using a Type I Modified section for UHPC with a compressive strength of 20,000 psi (137.9 MPa) and 29,000 psi (200 MPa), respectively. Figures 3-12 and 3-13 show the same for a Bulb-Tee 28. The flexural-shear capacity at the critical section decreased as the length of the girder increased. The capacity increased as the flexural tensile strength, fr, of the concrete increased. The effect of the fibers was more important for shorter beams.
The results of the Bulb-Tee 28 section analyses were similar to those obtained for the Type I Modified section. For the span length limits due to deflection (88 ft and 94ft (26.8 m and 28.7 m), respectively), the nominal flexural-shear capacity of the concrete (Vci) was greater than the ultimate shear (Vu) for concretes with flexural tensile strengths greater than 4,000 psi (27.6 MPa). The ratio Vci/Vu at the span length limit due to deflection (Ld) for the Bulb-Tee 28 section was found to be slightly greater than the one calculated for the Type I Modified at its corresponding Ld. The gain in flexural-shear strength when the compressive strength of the concrete was increased from 20,000 psi (137.9 MPa) to 29,000 psi (200 MPa) was found to be only about 2% (for a span length of 90 ft (27.4 m)).
3-13

190

170 Span length limited by deflections (76 ft)
150

Vci at Critical Section (Kips)

130

110

90

70

50 40

50 6*(fc')^0.5

60

70

80

S pan Length (ft)

fr 2,500

fr 4,000

90

100

110

120

fr 5,500

fr 7,000

Vu

Figure 3-10 Flexural-shear capacity at critical section vs. span length. Type I Modified. fc'=20,000 psi (137.9 MPa)

Vci at Critical Section (Kips)

180

160
Span length limited by deflections (81 ft) 140

120

100

80

60 40

50

60

6*(fc')^0.5

70

80

90

100

110

Span Length (ft)

fr 2,500

fr 4,000

fr 5,500

120

130

140

fr 7,000

Vu

Figure 3-11 Flexural-shear capacity at critical section vs. span length. Type I Modified. fc'=29,000 psi (200 MPa)

3-14

Vci at Critical Section (Kips)

250
Span length limited by deflections (88 ft) 200

150

100

50

40

50

60

70

6*(fc')^0.5

fr 2,500

80

90

100

110

S pan Length (ft)

fr 4,000

fr 5,500

120

130

140

fr 7,000

Vu

Figure 3-12 Flexural-shear capacity at critical section vs. span length. Bulb-Tee 28. fc'=20,000 psi (137.9 MPa)

260

Vci at Critical Section (Kips)

210
Span length limited by deflections (94 ft)
160

110

60 40

50

60

6*(fc')^0.5

70

80

90

100

110

Span Length (ft)

fr 2,500

fr 4,000

fr 5,500

120

130

140

fr 7,000

Vu

Figure 3-13 Flexural-shear capacity at critical section vs. span length. Bulb-Tee 28. fc'=29,000 psi (200 MPa)

3-15

Shear reinforcement is not required for Type I modified and for the BT 28 so long as the modulus of rupture equals about 3,000 psi.

3.7 Horizontal Interface Shear
The design for horizontal interface shear in composite sections was performed according to article 9.20.4 of the AASHTO specifications (AASHTO, 1996). For this design, the nominal horizontal concrete shear strength (Vnh) was calculated assuming contact surface clean, free of laitance and intentionally roughened (AASHTO 9.20.4.3 (a)). If the nominal horizontal concrete shear strength multiplied by (0.9 for shear) was found to be smaller than the ultimate horizontal shear required (factored composite dead load shear and live load plus impact shear), then the minimum area of steel reinforcement was provided (AASHTO 9.20.4.5); and the additional area of shear reinforcement needed was calculated according to articles 9.20.4.3 (c) and 9.20.4.3 (d) of the AASHTO specifications (AASHTO, 1996).
For the analysis of horizontal shear, the critical section was considered to be at a distance from the support equal to one half the total height of the section. The equation provided in article 9.20.4.3 (a) of the AASHTO specifications (AASHTO, 1996) does not include the effect of the compressive strength of the concrete on the capacity of the section against horizontal shear. For this reason, a single analysis for a 20,000-psi (137.9 MPa) concrete was developed for each one of the sections considered.

3.7.1 Analysis of Horizontal Shear for Type I Modified Figure 3-14 presents the results of the analysis performed for the Type I Modified section.
The nominal horizontal shear capacity (Vnh) of the composite section slightly decreased as the span length increased. This decrease occurred because the number of strands needed to reach longer spans increased, resulting in a small reduction of the eccentricity. Therefore, the value of d in the expression (Eq. 3.1) utilized for the calculation of Vnh decreased. On the other hand, the ultimate horizontal shear force (Vuh) at the critical section (h/2 from the support) increased as the span length of the girder increased.

Vnh = 0.8 bv d

(3.1)

3-16

The Vnh for the 50-ft (15.24 m) span was only 21% of Vuh. This percentage was smaller for all other spans. These results indicated that the horizontal shear controlled the stirrup design. Even if the use of UHPC showed that shear reinforcement was not required, the use of stirrups was necessary to satisfy the horizontal shear requirements. Although the results reported in Figure 3-14 only correspond to the critical section, it was observed that the use of stirrups or other shear transfer device for horizontal shear was required along the entire beam.

Horizontal Shear (Kips)

180

160

140

120

100

VVnh

80

VVuuh

60

40

20

0

40

50

60

70

80

90

100 110 120

Span Length (ft)

Figure 3-14 Horizontal shear capacity at critical section (h/2) vs. span length. Type I Modified. fc'=20,000 psi (137.9 MPa)

3.7.2 Analysis of Horizontal Shear for Bulb-Tee 28 Figure 3-15 presents the results of the analysis performed for the Bulb-Tee 28 section. The
nominal horizontal shear capacity (Vnh) of the composite section slightly decreased as the span length increased. The capacity remained constant for the last two span lengths analyzed due to the limit imposed for the value of d in Eq. (3.1). This limit restrains the value of d to a minimum of 0.8*h.
The Vnh reached for the 50-ft (15.24 m) span was 75% of Vuh. The increment in capacity of this section as compared to the one reached by the Type I Modified is the result of the wider top flange (bv) of the Bulb-Tee section.
For a span length of 88 ft (26.8 m), corresponding to the span length limit due to deflection, the use of stirrups for horizontal shear was required only at the end one-third length of the girder.

3-17

For the middle third region, the horizontal shear capacity was greater than the ultimate horizontal shear acting on the girder.
These results indicated that the horizontal shear controlled the stirrup design. Again, even if the use of UHPC showed that shear reinforcement was not required, the use of stirrups was necessary to satisfy the horizontal shear requirements at least at the regions close to the support.

Horizontal Shear (Kips)

180 160 140 120 100
80 60 40 20
0 40

PVhnh VHuh

60

80

100

120

140

Span Length (ft)

Figure 3-15 Horizontal shear capacity at critical section (h/2) vs. span length. Bulb-Tee 28. fc'=20,000 psi (137.9 MPa)

3.8 Conclusions based on Analytical Investigation
The maximum span length for HPC girders determined based on the L/800 live load deflection requirement of article 9.11.3 of the AASHTO specification, Ld, was 78 ft (23.8 m) for the AASHTO Type I Modified section. This value increased by 17.9% to 92 ft. (28 m) using a PCI Bulb-Tee section that kept the same total height as the Type I Modified (28 in, 711 mm). HPC girders with fc ' up to 20,000 psi (137.4 MPa) and using the contribution of 0.6-inch diameter pretensioned strands and 0.5-inch diameter posttensioned strands did not reach the maximum span length Ld. The maximum span length attained by the AASHTO Type I Modified section was 68 ft (20.7 m). This value corresponded to 87% of Ld. Similarly, the maximum span length for the PCI Bulb-Tee 28 was 87 ft (26.5 m), corresponding to 94.6% of the Ld value determined for this section.

3-18

The increase in flexural tensile strength of the UHPC showed a significant effect on the achievement of longer spans. The maximum span length, Ld, was attained using flexural tensile strengths of 2,500 psi (17.2 MPa) and 1,850 psi (12.8 MPa) for the AASHTO Type I Modified and the PCI Bulb-Tee 28 sections, respectively. The increment in flexural tensile strength resulted in a reduction of the number of strands.
The PCI Bulb-Tee 28 section was more sensitive to the increase in flexural tensile strength than the Type I Modified section not only because a smaller increase was needed to achieve the maximum span length Ld, but also because the reduction in the area of steel reinforcement was greater than the one observed for the AASHTO Type I Modified.
Girders analyzed using UHPC, including UHPC without fibers (fr = 6 fc ' ), generally could be designed to surpass the maximum span lengths as limited by deflection and ultimate strength requirements. The maximum span lengths for UHPC having 29,000 psi (200 MPa) compressive strength, 7,000 psi (48.3 MPa) flexural tensile strength and allowable compression of 0.6 fc ' , were 140 ft (42.7 m) and 165 ft (50.3 m) for the AASHTO Type I Modified and the PCI Bulb-Tee 28 sections, respectively. These results indicated that the high compressive and flexural tensile strength of UHPC could not be fully utilized in the AASHTO Type I Modified and PCI Bulb-Tee 28 sections. It appears that the same is likely for all other AASHTO and PCI Bulb-Tee sections.
The increased span lengths for girders made of UHPC results not only from the increased compressive and flexural tensile strengths but also from the property that UHPC has no creep and shrinkage prestressing losses. In general, it was observed that the increased flexural tensile strength of UHPC was much more important in the achievement of longer spans than increasing the compressive strength. Span increases of about 20% were achieved when the flexural tensile strength of the concrete was increased from 6 fc ' to 7,000 psi (48.3 MPa). Generally, span length increases were below 16% when the compressive strength of the concrete was increased from 20,000 psi to 29,000 psi (137.9 MPa to 200 Mpa).
Increases in flexural tensile strength of the concrete resulted in a reduction of the number of strands needed to reach a given span length. A maximum reduction of approximately 35% in the area of reinforcement was observed during the analysis of AASHTO Type I Modified girders when the flexural tensile strength was increased from 6 fc ' to 7,000 psi (48.3 MPa). The
3-19

maximum reduction was about 45% for the PCI Bulb-Tee 28 section when the flexural tensile strength was increased from 2,500 psi to 7,000 psi (17.2 MPa to 48.3 MPa).
Span length increases up to 10% were observed when the allowable compressive stress due to effective prestress plus permanent (dead) load was increased from 0.4 fc ' to 0.6 fc ' . This increase in the AASHTO specifications could be permitted based upon the lack of creep and increased ductility of UHPC.
The vertical shear design of UHPC girders was controlled by the flexural-shear cracking and not by the web shear strength. The effect of the compressive strength on the shear capacity of the concrete was found to be almost negligible when compared to that of the flexural tensile strength. For the maximum span length determined by deflection limit (Ld), the flexural-shear capacity of the concrete (Vc) at the critical section was greater than the ultimate applied shear (Vu) for concretes with flexural tensile strengths equal or greater than 4,000 psi (27.6 MPa). The latter was true for concretes with compressive strengths as low as 16,000 psi (110.3 MPa). This means that for the AASHTO Type I Modified and PCI Bulb-Tee 28 sections made of UHPC, vertical shear reinforcing bar stirrups are not required.
Analysis results of horizontal shear for composite sections showed that stirrup or some other type of reinforcement is required at the interface between the UHPC section and the cast-in-place deck slab.
3-20

4. Material Properties

4.1 Compressive Strength

Compressive strength was determined by testing 3 x 6-in (75x 50-mm) cylinders

according to ASTM C 39. All compressive testing was performed in a SATEC MKIII 800 RD

800 kip (3,558,580 kN) capacity compression testing machine. Prior to testing, the cylinder ends

were ground to provide a smooth surface. A minimum of three specimens were tested from each

batch for each measurement. All batches were tested at the ages of 7, 28, 90 and 365 days using

3x6-in (75 x 150 mm) cylinders.

Four different conditions were considered to examine the influence of fiber content and

thermal treatment. (Table 4.1). The nomenclature used in this study was based on the type of UHPC used (i.e., Ductal = D), fiber volume fraction (i.e., 2% volume fraction = 2f), and maximum treatment temperature reached while curing (i.e. 73 oF or 194oF (23oC or 90oC)). For

example, Mix "D-2f-90C" indicates that the premix was used with 2% steel fiber content, thermally treated at 194oF (90oC).

Table 4.1: Different UHPC mixes considered for compressive strength

Mixture ID D-2f-90C

Curing temperature1

Thermal treatment time

194oF (90oC)

48 hours

Fiber content 2% by vol.

D-2f-60C

140oF (60oC)

72 hours

2% by vol.

D-2f-23C

73oF (23oC)

N/A

2% by vol.

D-0f-90C

194oF (90oC))

1 All samples were cured at 100% RH.

48 hours

No fibers

Test results showed that the average 28-day compressive strength of mixtures D-2f-90C , D-2f-60C , D-2f-23C ,and D-0f-90C was 25.5 (176), 22 (152), 21.5 (148), and 15.8 (109) ksi (MPa) respectively (Figure 4.1). These results are high for concrete but lower than expected for UHPC. This may be due to surface imperfections or imperfect planeness. Results in Figure 4.1 also show that there was not noticeable increase in strength with time for all thermally treated mixtures while this was no the case for mixture D-2f-23C. In addition, the significant low strength of mixture D-0f-90C may be due to lack of optimum particle packing in the absence of fibers.

4-1

Compressive strength (psi)..

30000 25000 20000 15000 10000
5000 0

D-2f-90C D-2f-23C

D-2f-60C D-0f-90C

7

28

90

365

Age (days)

Figure 4.1 Compressive Strength of different UHPC

4.2 Tensile Strength
4.2.1 Splitting tensile test [ASTM C496] In the splitting tensile test, a compressive force is applied perpendicularly to the
longitudinal axis of the cylinder (Figure 4.2). This force creates an almost uniform tensile stress over the middle part of the vertical loading plane. However, high compressive stresses are generated close to the ends of the vertical plane as shown in Figure 4.2.b.
The tensile strength value calculated from the splitting tensile test is typically lower than the corresponding value determined from the modulus of rupture test, but it is higher than the result obtained from direct tensile testing [Graybeal, 2005]. This difference exists due to: (1) considering concrete a homogenous material, (2) neglecting the effect of the high compressive stresses generated at the ends of the vertical plane during the calculations; (3) neglecting lateral restraints provided by friction between the specimen and the thin plywood bearing strips. These three effects increase the apparent value of tensile strength. Splitting tension tests were performed at the same times as compression strength tests. All batches were tested at the ages of

4-2

7, 28, 90 and 365 days using 75 x 150 mm (3x6-in) cylinders. The four different conditions shown in Table 4.1 were considered for this test.
Test results showed that the average 28-day splitting tensile strength of mixtures D-2f90C , D-2f-60C , D-2f-23C ,and D-0f-90C was 3160 (21.8), 2790 (19.2), 2400 (16.6), and 2485 (17.1) psi (MPa), respectively (Figure 4.3). Results in Figure 4.3 also show that there was not noticeable increase in strength with time for all thermally treated mixtures while this was not the case for mixture D-2f-23C. In addition, the significant low strength of mixture D-0f-90C may be due to lack of optimum particle packing in the absence of fibers.
4.2.2 Direct tension test Based on the review lecture conducted, a new direct tension test specimen was developed
at Georgia Tech. (Figure 4.4). A test specimen is 9.25-in (235 mm) in height and has a reduced croos section of 2x2-in (50x50 mm). Tensile loads were applied to specimens through 3x2x1-in (75x50x25 mm) steel plates anchored with two tapered threaded bolts with an embedment length of 1.5-in (38 mm). This configuration was chosen not to create a stress concentration plane at the rods ends. These plates are to be placed and aligned in the steel molds prior to casting, highstrength / high modulus fast setting epoxy was applied to the loading plates prior to pouring UHPC in the molds to strengthen the connection and help enhancing the development of uniform stress field in the middle part of the specimen. Two 1.2-in (30 mm) strain gages were attached on two opposite sides of the specimen These two strain gages were connected in a half bridge configuration and calibrated prior to loading in order to measure longitudinal strains during the tension test. All batches were tested at the age of 7 days. The four different conditions shown in Table 4.1 were considered for this test.
Test results (Figure 4.5) showed that the average 28-day direct tensile strength of mixtures D-2f-90C , D-2f-60C , D-2f-23C ,and D-0f-90C was 1492 (10.3), 1413 (9.7), 1091 (7.5), and 1052 (7.3) psi (MPa), respectively. Results from both tension tested conducted were also compared. Results plotted in Figure 4.6 shows that the ratio between tensile strength obtained by the splitting tension test to tensile strength obtained by the direct tension test ranged between 2.0 and 2.3 with an average value of about 2.14. Obtaining such ratio was of specific importance as it allows conducting the relatively easy standardized splitting tension test and then predicting the corresponding direct tension strength of the material.
4-3

(a)
(b) Figure 4.2: (a) Splitting tension test and (b) Stress distribution across loaded diameter in the
splitting tensile test [Mehta and Monteiro, 2005]
4-4

Tensile strength (psi)

4000

D-2f-90C

D-2f-60C

D-2f-23C

D-0f-90C

3500

3000

2500

2000

1500

1000

500

0 7

28

90

365

Age (days)

Figure 4.3: Splitting tension test results

Figure 4.4: Tensile strength test specimen 4-5

Tensile strength (psi)

1800

1500

1200

900

600

300

0 D-2f-90C

D-2f-60C

D-2f-23C

UHPC Mix

D-0f-90C

Figure 4.5: Direct tension test results

4000 Direct tension test

3500

Split tension test

3000

2500

2000

1500

1000

500

0 D-2f-90C

D-2f-60C

D-2f-23C

UHPC Mix

D-0f-90C

Figure 4.6: Comparison between split tension test and direct tension test at 7 days

Tensile strength (psi)

4-6

4.3 Modulus of Elasticity
The chord modulus of elasticity was measured using 4x8-in (100x200-mm) cylinders loaded in compression. The tests were run in an SATEC Balwin 400 BTE 400 kip (1,800,000 kN) universal testing machine. Figure 4.7 shows a typical setup for elastic modulus test. All batches were tested at the age of 7 days. The four different conditions shown in Table 4.1 were considered for this test.
Test results showed that the average 28-day splitting tensile strength of mixtures D-2f90C , D-2f-60C , D-2f-23C ,and D-0f-90C was 6953 (47950), 7376 (50870), 6510 (44900), and 6570 (45310) ksi (MPa), respectively (Figure 4.8). These moduli are twice as much as normal strength concrete, they also agree to a great extent with the values obtained by Graybeal [2005] for UHPC. Figure 4.9 shows the Poisson's ratios obtained for the four mixes considered, Poisson's ratio values varied from 0.146 to 0.155. These values are lower than Poisson's ratios for normal strength concrete which may be due to lateral restraints applied by fibers.
Figure 4.7: Modulus of elasticity test setup
4-7

Modulus of Elastcitiy (ksi)

9000

8000

7000

6000

5000

4000

3000

2000

1000

0 D-2f-90C

D-2f-60C

D-2f-23C

UHPC Mix

Figure 4.8: Modulus of elasticity test results

0.20 0.18 0.16 0.14 0.12 0.10 0.08 0.06 0.04 0.02 0.00
D-2f-90C

D-2f-60C

D-2f-23C

UHPC Mix

Figure 4.9: Poisson's ratio of UHPC

D-0f-90C D-0f-90C

Poisson's ratio

4-8

4.4 Modulus of Rupture
A one-speed shear mixer of about 2.0 ft3 volumetric capacity was used for all mixes (Figure 4.10).

Figure 4.10 Batching Ductal premix in a 2.0 ft3 shear mixer
Once mixing was complete, 2x2x2-in (50x50x50 mm) cubes and 3x6-in (75x150 mm) cylinders for compressive strength and 2x2x9-in (50x50x225 mm), 2x2x16-in (50x50x400 mm), and 2x2x23-in (50x50x575 mm) beams for 3-point bending test were filled during vibration using a vibrating table. In this study, curing under standard ASTM conditions and using thermal treatment for 48 hours at 90 C two days after placement were considered.

4.4.1 Compressive strength

Cubes and cylinders were tested according to the ASTM C109, and ASTM C39

respectively except that the load rate used was increased to 145 psi (1 MPa)/min to be consistent with the Ductal Batching and Testing Procedures and the procedures found in the literature.

Five specimens were tested at ages of 2, 4, 7 and 28 days.

Table 4.2: Compressive strength results

3x6" Cylinders

2x2x2" Cubes

Age

fc(psi)

SD(psi)

Thermal/ Curing regime

fc(psi)

ASTM Thermal ASTM Thermal ASTM ASTM Thermal ASTM Thermal

2 10945.3

1667.8

13003.2

906.9

4 13112.8 22631.4 752.5 1821.0

1.7 17024.4 29809.1 1341.6 2598.4

7 14523.5 22643.8 800.0 3591.6

1.6 18389.0 30159.9 1860.0 1582.9

28 18356.8 22127.9 893.5 1598.0

1.2 23440.0 29986.9 2018.6 2015.3

Thermal/ ASTM
1.8 1.6 1.3

4-9

Compressive strength (psi)

30000 25000 20000 15000 10000
5000 0 0

a- 3x6 Cylinders

ASTM curing Thermal curing

5

10

15

20

25

30

Age (days)

35000

b- 2x2x2 Cubes

30000

25000

20000

15000

10000
5000
0 0

ASTM curing Thermal curing

5

10

15

20

25

30

Age (days)

Figure 4.11: Compressive strength results a- cylinders, b- cubes

Compressive strength (psi)

4-10

Compressive strength (psi)

30000 25000 20000 15000 10000
5000 0 0

a- Ambient curing

3x6 Cylinders 2x2x2 Cubes

5

10

15

20

25

30

Age (days)

40000

b- Thermal Curing

35000

Compressive strength (psi)

30000

25000

20000

15000

10000
5000
0 0

3x6 Cylinders 2x2x2 Cubes

5

10

15

20

25

30

Age (days)

Figure 4.12: Comparison between cylinders and cubes results a- Ambient curing, b- Thermal curing

4-11

4.4.2 Modulus of rupture Figure 4.13 shows the 3-point bending test setup used to determine the tensile properties
of UHPC. This test involves the three-point flexural loading of small-scale concrete prisms. During the test, the load on and the deflection of the prism are monitored. These data were then used to determine the modulus of rupture.
(a)
(b)
Figure 4.13: (a) Schematic diagram of 3-point bending test setup; (b) 3-point bending test with a deflectometer at mid-span
Midspan deflections were measured using either a digital dial gage with an accuracy of 0.0001" for the 2-days ambient cured series (Series A-2), or using a digital deflectometer connected to an electronic data acquisition system for all other series. This later setup is able to record and save loads and deflections through out automatically.
The nomenclature used in this study was based on the curing conditions (i.e. A = ambient curing and T = Thermal treatment) and age of loading (i.e. 2= tested at age of 2 days). For example, Series "T-28" indicates that specimens of this series were thermally treated at 194oF (90oC), and tested at 28 days.
4-12

In order to determine the aspect ratio (L/h) effect on the tensile behavior and properties on UHPC, also, as larger shear span to prism depth ratio would much more accurately represent the flexural response of UHPC, the span (L) was varied from 7", 14", and 21", while the cross section dimensions (2x2 in) were kept constant for all specimens. During casting of each prism, special care was taken to ensure that the UHPC flowed from one end of the prism to the other, thus ensuring a fiber distribution and alignment system that was similar to what would occur in the large-scale casting of a beam or plate type flexural member. At least three specimens of each span length were testes at each age for each curing regime using a Satec load frame. The deflection rate was set so that the expected first crack deflection would occur approximately 1 minute into the test. This rate varied depending on the prism cross-section and loading configuration. The test was stopped after a load dropped down at least to half of the maximum value.
The first cracking noticed for all specimens was tensile stress cracking on the bottom of the specimen. Thus, first cracking--recorded by the data acquisition system and physically observed on the specimen was usually quite clear. Figure 4.14 shows the load-deflection response curves for all series tested. Figures show that the prism response is linear until first cracking when a clearly defined decrease in load carrying capacity occurs. Soon thereafter the load again begins to increase transferring the load to adjacent fibers. The saw-tooth pattern visible in the response is indicative of additional individual cracks resulting from continuous fibers pull out on the tension face, which is thought to be the dominant failure mechanism of almost all the tested specimens. It was also noticed that in most of cases shorter spans, (i.e. 7 in spans and some 14 in spans) showed typical shear failure cracks prior to failure.
As mentioned before, prism flexure testing of three different loading configurations for each curing regime at each age was intended to identify the benefits and detriments of varying the beam span (L). From a qualitative standpoint, the results of both the A-2-14, and A-2-21 were more consistent than other series. While from a quantitative standpoint, as the span to depth ratio of the beam decreases, the modulus of rupture general decreased, the best configuration to predict the tensile strength obtained by splitting tension test was the 21 in configuration (Figure 6).
Figure 4.15 shows a summary of the modulus of rupture values obtained for all UHPC series tested. It can be noticed that the modulus of rupture kept increasing with age for all
4-13

ambient-cured series (i.e. A-2, A-4, A-7, A-28), while this was not the case of for T-4, T-7, and T-28 series. For all spans, the A-2 series had the minimum average modulus of rupture. The absolute average minimum and maximum modulus of rupture recorded was for series A-2 @ span of 14-in (i.e. 1198 psi (8.25 MPa)) and series T-4 @ span of 7-in (i.e. 4440 psi (30.60 MPa)), respectively.

Load (lb)

1800 (a) A-2 Series 1600 1400

7 in span 14 in span 21 in span

1200

1000

800

600

400

200

0 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70

Midspan deflection (in)

4-14

Load (lb)

2500 (b) A-4 Series 2250

2000

1750

1500

1250

1000

750

500

250

0

0.00

0.10

0.20

0.30

0.40

Midspan deflection (in)

7 in span 14 in span 21 in span

0.50

0.60

2500 (c) A-7 Series
2250

2000

1750

1500

1250

1000

750

500

250

0

0.00

0.05

0.10

0.15

0.20

Midspan deflection (in)

7 in span 14 in span 21 in span

0.25

0.30

Load (lb)

4-15

Load (lb)

3000 (d) A-28 Series 2750 2500

2250

2000 1750

1500

1250

1000

750

500

250

0

0.00

0.05

0.10

0.15

0.20

Midspan deflection (in)

7 in span 14 in span 21 in span

0.25

0.30

4000 3500

(e) T-4 Series

3000

2500

2000

1500

1000

500

0

0.00

0.05

0.10

0.15

0.20

Midspan deflection (in)

7 in span 14 in span 21 in span

0.25

0.30

Load (lb)

4-16

Load (lb)

3000 (f) T-7 Series 2500

2000

1500

1000

500

0

0.00

0.05

0.10

0.15

0.20

Midspan deflection (in)

7 in span 14 in span 21 in span

0.25

0.30

4000 3500

(g) T-28

3000

2500

2000

1500

1000

500

0

0.00

0.05

0.10

0.15

0.20

Midspan deflection (in)

7 in span 14 in span 21 in span

0.25

0.30

Figure 4.14: Average load-deflection behavior of different UHPC series

Load (lb)

4-17

Modulus of Rupture (psi).

6000 5000

A-2 A-4 A-7 A-28 T-4 T-7 T-28

4000

3000

2000

1000

0 7

14

21

Span (in)

Figure 4.15: Modulus of rupture of different UHPC series

4.5 Autogenous Shrinkage
4.5.1 Sample Matrix Three different conditions, designed to examine the influence of varying fiber content
and thermal treatment were considered (Table 4.3). The nomenclature used in this study was based on the type of UHPC used (i.e., Ductal = D), fiber volume fraction (i.e., 2% volume fraction = 2f), and the maximum treatment temperature reached while curing (i.e. 73 oF or 194oF (23oC or 90oC)). For example, Mix "D-2f-90C" indicates that the premix was used with 2% steel fiber content, and thermally treated at 90oC (194oF).

Table 4.3: Different UHPC mixes, curing and loading conditions Mixture ID Curing temperature1 Fiber content

D-2f-90C

194oF (90oC)

2% by vol.

D-2f-23C

73oF (23oC)

2% by vol.

D-0f-90C

194oF (90oC)

No fibers

1 All samples were cured at 100% RH.

4-18

4.5.2 Testing Methodology Autogenous shrinkage was measured on three replicate samples according to the method
described by Jensen (1995) using rigid corrugated polyethylene tubes capped on both ends to prevent any moisture loss to the environment. Linear deformation of these tubes was first measured at final setting using a digital comparator (Figure. 4.16). Autogenous shrinkage samples were kept in an environmental chamber at 20 C and 50% relative humidity over the test period.
Figure 4:16: Autogenous shrinkage test setup
4.5.3 Results and Discussion First, results in Figure 4.17 show that 87% of the total autogenous shrinkage occurred
during the thermal treatment. This is expected due to the accelerated rate of early hydration at higher temperature.
As for the effect of fibers, results in Figure 4.17 show that the use of 2% steel fibers by volume resulted in a 42% reduction in autogenous shrinkage at 14 days of age after batching. In addition, and by taking a closer look at Figure 4.17, it can be clearly seen that the effect of fibers in reducing autogenous shrinkage increased as the hydration of the cementitious paste progressed during the thermal treatment period. This result emphasizes that the properties of the fiber-matrix interface, which enhances, as hydration progresses, has a major influence on how efficiently the fibers can restrain deformations.
By comparing the drying shrinkage results shown in Figure 4.18 to the autogenous shrinkage shown in Figure 4.17, it can be seen that due to the low water-to-cementitious materials ratio used in the UHPC mixes, autogenous shrinkage represented about 93% of the
4-19

total shrinkage deformations for both mixes at age of 14 days after batching; autogenous shrinkage, then, must be considered when assessing performance of ultra-high performance concretes.
1200

1000

Shrinkage strain ()

800

600

400
200
0 0

D-2f-90C D-2f-23C D-0f-90C

48

96

144 192 240 288 336 384

Time after batching (Hrs)

Figure 4.17: Autogenous shrinkage of different UHPC mixes

4.6 Short-term Tensile Creep and Free Shrinkage
4.6.1 Sample Matrix Four different conditions, designed to examine the influence of varying stress/strength
ratio at the time of loading, fiber content, and thermal treatment were considered (Table 4.4). The nomenclature used in this study was based on the type of UHPC used (i.e., Ductal = D), fiber volume fraction (i.e., 2% volume fraction = 2f), maximum treatment temperature reached while curing (i.e. 73 oF or 194oF (23oC or 90oC)), and the stress level maintained during the creep test (i.e. 40% means that the tensile stress-to-tensile strength ratio at the time of loading was 40%). For example, Mix "D-2f-90C-40" indicates that the premix was used with 2% steel fiber content, thermally treated at 194oF (90oC), and loaded at 40% of its tensile strength at the time of loading.

4-20

Table 4.4: Different UHPC mixes, curing and loading conditions

Mixture ID D-2f-90C-40

Stress/strength at loading (%)
40

Curing temperature1 Fiber content

194oF (90oC)

2% by vol.

D-2f-23C-40

40

73oF (23oC)

2% by vol.

D-0f-90C-40

40

194oF (90oC)

No fibers

D-2f-90C-60

60

1 All samples were cured at 100% RH

194oF (90oC)

2% by vol.

4.6.2 Testing Methodology
Tensile creep and free shrinkage deformations were measured on 4x15-in (100x380 mm) cylinders, two loaded cylinders for tensile creep and two un-loaded companion specimens for free shrinkage (Figure 4.18). Each cylinder was fitted with four sets of steel inserts located diametrically opposite on the surface of the specimen. Each set had a gauge length of 10-in (250 mm). For each creep specimen, two 1.50 in (38 mm) thick steel plates were affixed, one on the top and one on the bottom using high modulus, high strength epoxy and left to cure for ~36 hours under a pressure of 400 psi (2.75 MPa) (no rods were embedded in the concrete for affixing the plates to the specimens in this study). At an age of 7 days, creep samples were loaded at a constant rate of 0.76 psi/second (0.0046 MPa/second) until the desired load was reached. For each specimen, deformations were measured on each set of inserts using mechanical DEMEC gage with an accuracy of 0.0001-in (0.00254 mm). The tensile creep test setup is shown in Figure 4.18. Test conditions were kept at 73oF (23 2 oC) and 50 3% RH for the whole testing period. Tensile creep and shrinkage deformations were measured initially at 1, 2, 4, 6, 12, 24 hours after loading. Subsequently, measurements were made daily for the rest of the 14-day testing period.

4-21

Loading frame Load cell
Creep specimen
Head plate
Shrinkage specimen Temperature and humidity reader
Figure 4.18: Short-term tensile creep test setup 4.6.3 Results and Discussion
As shown in Figure 4.19 and Table 4.5, the incorporation of short steel fibers limited free shrinkage. Results at 14 days of drying show that the free shrinkage of mix D-0f-90C-40, with no steel fibers, was about 135% more than that of mix D-2f-90C-40 in which 2% fibers by volume was incorporated. This suggests that the fibers offer some restraint to deformation, as would be expected when the fibers are well-dispersed in the matrix.
Figure 4.19 and Table 4.5 also show that after 14 days of drying, the free shrinkage of mix D-2f-23C-40 (i.e. where no thermal treatment was applied) was about six times that of the mix D-2f-90C-40 which was thermally treated at 194oF (90oC) for 48 hours. In addition, results in Figure 4.19 also show that mix D-2f-23C-40 continued to shrink for the whole testing period, while mix D-2f-90C-40 reached an asymptotic shrinkage strain value of about 45 after about 6 days of drying. Similar results for UHPC were reported by Graybeal [2005], where specimens
4-22

thermally treated at 194oF (90oC) for 48 hours did not show any measurable free shrinkage after thermal treatment, while air-cured specimens continued to shrink with time. Both microstructural refinement and consumption of most of the mix water due to the accelerated hydration associated with thermal treatment likely contribute to the reduced shrinkage in the thermally treated samples.
This section presents the effects of varying the stress-to-strength ratio, incorporating short and randomly-dispersed high strength steel fibers in mix, and thermal treatment prior to loading on the tensile creep of UHPC. Results in Figure 4.20 show tensile creep increasing steadily with time for each of the four mix types examined. First, both the tensile creep coefficient and the specific tensile creep increased upon increasing the stress-strength ratios. However, both parameters decreased in the presence of steel fibers and with the application of thermal treatment prior to loading. The effect of steel fibers in this study is of particular importance since the creep behavior is contrary to published reports [Bissonnette and Pigeon, 1995, and Bissonnette et al., 2007], which indicate an increase in tensile creep with fiber reinforcement in NSC and HPC. (This difference is discussed in further detail below.) Generally, these results show the tensile creep of UHPC to be lower than tensile creep results reported in the literature for normal strength and high performance concretes [Bissonnette and Pigeon, 1995, and Bissonnette et al., 2007]. However, the tensile creep measured here in UHPC is also significantly higher than the compressive creep previously reported by Graybeal [2005] for UHPC.
First, increasing the stress-to-strength ratio from 40% (D-2f-90C-40) to 60% (D-2f-90C60) increased the creep coefficient by 44% and increased the specific creep by 11% at 14 days of loading (Figure 4.20 and Table 4.5). These results in tension agree with the compression results of Graybeal [2005] who reported increasing compressive creep with increasing stress-to-strength ratio for the same steel fiber reinforced UHPC material. Similar results have been also reported by Bissonnette et al. [2007] where the tensile creep of normal and high strength concretes have shown a dependence on the stress-to-strength ratio.
As for the effect of steel fiber reinforcement, comparing the tensile creep results for mixes with (D-2f-90C-40) and without steel fibers (D-0f-90C-40) in Figure 4.20 and in Table 4.5 shows that the tensile creep coefficient decreased by 10% and the specific creep decreased by 40% with the addition of 2% fibers at 14 days. It is noted that the UHPC examined here was optimized based upon the use of 2% steel fibers by volume; hence, structure and performance
4-23

were designed to be enhanced by the presence of fibers. These results, however, are opposite to those reported by Bissonnette and Pigeon [1995] and by Bissonnette et al. [2007] which investigated the influence of steel fiber reinforcement in normal and high strength concretes undergoing tensile creep. In both of these previous studies, an increase in tensile creep was typically associated with the use of steel fibers. The variation between the prior and current results could be related to the differences in the material composition (e.g., particle packing, heterogeneity), material structure (e.g., lower strength, less microstructurally dense NSC vs. higher strength, lower micro/nano porosity UHPC), and processing (e.g., ordinary moist curing vs. thermal treatment). In particular, the application of thermal treatment to the fiber reinforced samples warrants further examination.
As previously noted, the application of thermal treatment was found to significantly affect tensile creep of UHPC. Thermally treating fiber reinforced UHPC at 194oF (90oC) for 48 hours prior to loading resulted in a 73% decrease in the creep coefficient and 77% decrease in specific creep when comparing the behavior of D-2f-23C-40 and D-2f-90C-40 at 7 days of loading (due to the premature failure of the D-2f-23C-40 mixture). It is proposed that this thermal treatment may result in refinement in the cementitious matrix nano- and microstructure, especially around the fibers. This proposed improvement in the structure and properties at the fiber/matrix interface may be viewed similarly although occurring likely at the nanoscale in UHPC to refinements by thermal curing at the microscale, interfacial transition zone (ITZ) around coarse aggregate [Mehta and Monteiro, 2005, and Bissonnette et al., 2007]. For nonthermal treated UHPC, the possibility of formation of water films around fibers during casting might have resulted in a higher local water-to-cementitious materials ratio. Mondal et al. [2008] monitored similar phenomenon around fine aggregates (0.046-0.093-in (1.18-2.36 mm) in size). The proposed porous interfacial zone around fibers might have been improved due to thermal treatment. Thermal treatment could have promoted further reaction of the cementitious paste, leading to a reduction in the porosity around fibers and minimization of the weakness in the fiber-matrix interfacial zone.
In addition, the D-2f-23C-40 specimens, which were not thermally treated, failed at the epoxy interface after one week of loading with the fracture occurring in the concrete. Substitute specimens were cast later but failed at the same location while loading. This failure may be related to the inadvertent introduction of defects into all the creep samples during the grinding of
4-24

the specimen ends before applying the epoxy adhesive to attach the steel loading plates. Premature failure occurred only in the non-thermally treated samples and never in the various thermally treated samples. This behavior suggests that some self-healing of any defects induced during grinding may have occurred during thermal treatment where water was available at the surface to aid the self-healing process.

300
D-2f-90C-40 & D-2f-90C-60

250

D-2f-23C-40

D-0f-90C-40

200

Strain ()

150

100

50

0

0

48

96

144 192 240 288 336 384

Time after drying (hrs)

Figure 4.19: Free shrinkage of different UHPC mixes [drying started at 7 days of age]

4-25

Strain ()

450 400 350 300 250 200 150 100
50 0 0

D-2f-90C-40 D-2f-23C-40 D-0f-90C-40 D-2f-90C-60

48

96

144 192 240 288 336 384

Time after loading (hrs)

Figure 4.20: Tensile total creep of different UHPC mixes loaded at 7 days of age

Mixture
D-2f-90C-40 D-2f-23C-40 D-0f-90C-40 D-2f-90C-60

Table 4.5: Summary of the results of the short-term study

Tensile strength

Tensile creep coefficient

Specific tensile Creep in /MPa
(/psi)

Value MPa (psi)

SD MPa (psi)

7-day

14-day

7-day

14-day

6.50 (943)

0.29 (42)

0.65

0.78

28.44 34.22 (0.20) (0.24)

6.09 (883)

0.40 (58)

2.38

N/A

124.66 (0.86)

N/A

6.04 (876)

0.30 (44)

0.78

0.87

51.78 57.44 (0.34) (0.40)

6.50 (943)

0.29 (42)

0.89

1.12

30.2 38.00 (0.21) (0.26)

Free shrinkage strain ()

7-day 14-day

43

44

216 247

108 103

43

44

Tensile creep results for mix D-2f-90C-40 in the current study were compared to compressive creep results for steel fiber reinforced (2% by vol.) UHPC specimens thermallytreated at 194oF (90oC) and loaded at 40% of the compressive strength by the time of loading

4-26

reported by Graybeal [2005]. At the age of 14 days of loading, this comparison shows that both the tensile creep coefficient and specific creep are about 12 times higher than the corresponding values during compressive creep. While the mechanisms of tensile creep behavior in UHPC have not been thoroughly examined, this significant difference in the magnitude and rate of development between compressive and tensile creep of UHPC, with the later being about an order of magnitude higher and faster at early age, suggests that tensile creep occurs phenomenologically quite differently than compressive creep in thermally treated, fiber reinforced UHPC. This may be, in part, due to the magnified effect of microcracking during tensile creep as compared to compressive creep, as well as the potential contributions of several other factors that are still to be investigated, including fiber restraint of compression creep, fiber/matrix debonding particularly during tensile creep, and deformation of the fiber reinforcement. This observation is of significant importance as it emphasizes that tensile creep should be fundamentally and fully investigated prior to specifying this material for use where tensile capacity is of primary importance.
Finally, another important result of this study is that both thermal treatment and fiber reinforcement affected tensile creep more significantly than these factors affect early (7-day) tensile strength. That is, while the increase in the 7-day tensile strength was about 8% upon incorporating short steel fibers, the corresponding decrease in the tensile creep coefficient and the specific tensile creep at 14 days of loading was about 10% and 40% respectively. Also, while the increase in the 7-day tensile strength was about 7% upon applying thermal treatment, the corresponding decrease in the tensile creep coefficient and the specific tensile creep at 7 days of loading was about 73% and 77% respectively. The observed differences in the results between tensile creep and tensile strength emphasizes the importance of conducting tensile creep testing to predict long-term tensile performance. It is proposed that, the more ability of cracks to coalesce and propagate during a tensile creep test rather than a tensile strength test may likely play an important role in describing the observed differences between the two sets of results.
4-27

4.7 Long-term Tensile Creep and Free Shrinkage
4.7.1 Sample Matrix
Four different conditions designed to examine the influence of fiber content and thermal treatment were considered (Table 4.6). The nomenclature used in this study was based on the type of UHPC used (i.e., Ductal = D), fiber volume fraction (i.e., 2% volume fraction = 2f), maximum treatment temperature reached while curing (i.e. 73 oF or 194oF (23oC or 90oC)), and the stress level maintained during the creep test (i.e. 40% means that the tensile stress-to-tensile strength ratio at the time of loading was 40%). For example, Mix "D-2f-90C-40" indicates that the premix was used with 2% steel fiber content, thermally treated at 194oF (90oC), and loaded at 40% of its tensile strength at the time of loading.

Table 4.6: Different UHPC mixes, curing and loading conditions

Mixture ID

Stress/strength at loading (%)

Curing temperature1 Fiber content

D-2f-90C-40

40

194oF (90oC)

2% by vol.

D-2f-60C-40

40

140oF (60oC)

2% by vol.

D-2f-23C-60

40

73oF (23oC)

2% by vol.

D-0f-90C-40

40

1 All samples were cured at 100% RH

194oF (90oC)

No fibers

4.7.2 Testing Methodology
Figure 4.21 shows the tensile creep testing frames designed at The Georgia Institute of Technology to carry out the tensile creep tests of UHPC. This design was inspired by an original tension creep frame designed by Bissonnette and Pigeon (1995) with some modifications that the researchers found advantageous. The Georgia Tech device consists of 6 testing frames that allow eighteen 3x3x19-in (75x75x483 mm) prism concrete specimens to be tested for creep and another 18 specimens for shrinkage at the same time. Test conditions were kept at 73oF (23 2 oC) and 50 3% RH for the whole testing period. More details of the tensile creep testing frames are provided below:
Loads: In this test setup, tensile loads were applied using 18x18x1-in (455x455x25 mm) plates weighting about 85-90 lbs (39-41 kg) each that are hanged at the end of the dead load

4-28

lever-arm via metal ball knob that kept the threaded rod connecting the dead loads to the loading always in a vertical position (Figure 4.22).
Dead load lever-arm: The dead load lever arm was designed with a magnification factor of (10:1). This high magnification factor was chosen to minimize the dead weights needed to apply high tensile stresses on UHPC. The dead load lever arm was connected from one end to the dead loads via metal ball knob as mentioned before and from the other end to the creep frame via ball-bearing (pinned) connection (Figure 4.23). In between, it was also pin-connected to the test specimens to ensure centric loading (Figure 4.23). During the test, the planer alignment of the dead load lever arm was guaranteed by providing two side rigid guides (Figure 4.23). Two track rollers were attached to each guide in order to eliminate friction between dead load lever-arm and the two guides.
Load cell: Two 1.2-in (30 mm) strain gages were attached on two opposite sides of the eye-bolt connecting the specimens from the top to the tensile creep frame. These two strain gages were connected in a half bridge configuration and calibrated prior to loading in order to be used as a load cell to ensure the transfer of the target load to the specimens and monitor the loads during the creep test (Figure 4.24). The tensile creep load capacity of each frame is 16,000 lb (71.2 kN) which is equivalent to 1,780 psi (12.30 MPa) tensile stress on a 3x3-in (75x75 mm) specimen. It is to be noted here that moments due to possible eccentricity of loads are eliminated through the two pin connections provided one at the top of the highest specimens and one at the bottom of the lowest specimen.
Load application: After dead loads were attached and prior to loading, a pallet jack was used to raise the dead loads to a position where the dead load lever-arm was in a horizontal position. At this point, the specimens-loading arm connection was tightened to keep the loading arm at an initial horizontal position before lowering the pallet jack (loading).
Specimens: Three tensile creep 3x3x19-in (75x75x483 mm) prism concrete specimens connected in series and three similar free shrinkage specimens were used for each case. These dimensions were chosen to give each specimen: (1) an aspect ratio (height/width) of more than 3 which eliminates the St. Venant's effect and gives uniform stress in the middle part of the specimen, and (2) width-to-fiber length ratio of 6 which produces more uniform tensile strength. The load was applied to specimens through 3x3x1-in (75x75x25 mm) steel plates anchored with four tapered threaded bolts with an embedment length of 1.5-in (138 mm) for two rods and 2.0-in
4-29

(50 mm) for the other two (Figure 4.25). This configuration was chosen not to create a stress concentration plane at the rods ends. These plates are to be placed and aligned in the steel molds prior to casting, high-strength / high modulus fast setting epoxy was applied to the loading plates prior to pouring UHPC in the molds to strengthen the connection and help enhancing the development of uniform stress field in the middle part of the specimen. The tensile creep specimens were connected through 1.0-in (25 mm) threaded rods that connect the end steel plates of each specimens to the following one.
Deformation measurements: Each specimen was fitted with four sets of steel inserts located on two opposite sides of the surface of the specimen. Each set had a gauge length of 10in (250 mm). Deformations were measured on each set of inserts using mechanical DEMEC gage with an accuracy of 0.00254 mm.
4-30

(a)
Figure 4.21 (a): Tensile creep setup developed at Georgia Tech, schematic diagram of a typical tensile creep frame 4-31

(b)
Figure 4.21 (b): Tensile creep setup developed at Georgia Tech, loaded specimens in tensile creep frames 4-32

Dead load lever-arm
Metal ball knob

Dead loads Figure 4.22: Tensile creep test setup: dead loads connection to the dead load lever-arm
Tensile creep frame

Dead load lever-arm

Loading arm frame connection
Loading arm specimens connection

Figure 4.23: Tensile creep test setup: dead load lever-arm connections to frame and specimens 4-33

Track rollers

Loading arm frame connection
Side guides

Figure 4.24(a): Tensile creep test setup: dead load lever-arm side guides
Tensile creep frame Eye bolt used as load cell Top UHPC specimen
Figure 4.24(b): Tensile creep test setup: specimen-frame top connection and load cell
4-34

Figure 4.25: Tensile creep test specimen 4.7.3 Results and Discussion
As shown in Figure 4.26, the incorporation of short steel fibers limited free shrinkage. Results at 330 days of drying showed that the free shrinkage of mix D-0f-90C-40, with no steel fibers, was about 2 times that of mix D-2f-90C-40 in which 2% fibers by volume were
4-35

incorporated. This suggests that the fibers offer some restraint to deformation, as would be expected when the fibers are well-dispersed in the matrix.
Figure 4.26 also shows that after 365 days of drying, the free shrinkage of mix D-2f-23C40 (i.e. where no thermal treatment was applied) was about 4 times that of the mix D-2f-90C-40 which was thermally treated at 194oF (90oC) for 48 hours. In addition, results in Figure 4.25 also show that mix D-2f-23C-40 continued to shrink for the whole testing period, while mix D-2f90C-40 reached an asymptotic shrinkage strain value of about 75 after about 5 months of drying. As mentioned before, this behavior can be attributed to microstructural refinement and consumption of most of the mix water due to the accelerated hydration associated with thermal treatment.
As for tensile creep, results in Figure 4.27 show that tensile creep decreased in the presence of steel fibers and with the application of thermal treatment prior to loading. The effect of steel fibers in this study is of particular importance since the creep behavior is contrary to published reports [Bissonnette and Pigeon, 1995, and Bissonnette et al., 2007], which indicate an increase in tensile creep with fiber reinforcement in NSC and HPC. (This difference is discussed in further detail below.) Generally, these results show the tensile creep of UHPC to be lower than tensile creep results reported in the literature for normal strength and high performance concretes [Bissonnette and Pigeon, 1995, and Bissonnette et al., 2007]. In both of these previous studies, an increase in tensile creep was typically associated with the use of steel fibers. The variation between the prior and current results could be related to the differences in the material composition (e.g., particle packing, heterogeneity), material structure (e.g., lower strength, less microstructurally dense NSC vs. higher strength, lower micro/nano porosity UHPC), and processing (e.g., ordinary moist curing vs. thermal treatment). In particular, the application of thermal treatment to the fiber reinforced samples warrants further examination.
4-36

Strain ()

350 300 250 200 150 100
50 0 0
700 600 500 400 300 200 100
0 0

D-2f-90C-40 D-2f-60C-40 D-2f-23C-40 D-0f-90C-40

50

100

150

200

250

300

350

Age (days)

Figure 4.26: Free shrinkage of different UHPC mixes

D-2f-90C-40 D-2f-60C-40 D-2f-23C-40 D-0f-90C-40

50

100

150

200

250

300

350

Age (days)

Figure 4.27: Tensile total creep of different UHPC mixes

Strain ()

4-37

4.8 Long-term Compressive Creep
4.8.1 Sample Matrix
Four different conditions designed to examine the influence of fiber content and thermal treatment were considered (Table 4.7). The nomenclature used in this study is similar to what was used in the tensile creep study except that the letter "C" was added to differentiate between tensile and compressive studies.

Table 4.7: Different UHPC mixes, curing and loading conditions

Mixture ID

Stress/strength at loading (%)

Curing temperature1 Fiber content

D-C-2f-90C-40

40

194oF (90oC)

2% by vol.

D-C-2f-60C-40

40

140oF (60oC)

2% by vol.

D-C-2f-23C-60

40

73oF (23oC)

2% by vol.

D-C-0f-90C-40

40

1 All samples were cured at 100% RH

194oF (90oC)

No fibers

4.8.2 Testing Methodology
Compressive creep was measured on twelve 4x15-in (100x380 mm) cylinders under sustained load from each UHPC mixture according to ASTM C 512 specifications. Three additional non-loaded companion specimens were used to measure shrinkage. All specimens were cured and and thermally treated as discussed above. The following tests were performed: (a) compressive strength, (b) modulus of elasticity, and (c) compressive creep and shrinkage. Compressive creep tests started at the same times as tensile creep tests. Figure 4.28 shows a picture of the creep frames used in this study. It is worth mentioning that the diameter of the cylinders used in this study was smaller than the standard from ASTM because the bearing capacity of some creep frames used was not enough for applying the required stress to 6x12-in (150x300 mm) UHPC cylinders recommended in ASTM C 512. After curing, specimens from each batch were placed in the environmentally controlled room at 50 3% RH and 73 3oF (23 2oC) where creep and shrinkage testing was conducted.
The cylinders were instrumented with four sets of steel inserts located diametrically opposite on the surface of the specimen. Each set was a 254-mm (10-in) long gauge line for

4-38

measuring deformation with a detachable mechanical gauge (DEMEC gauge) shown in Figure 4.29.
Figure 4.28: Compressive creep test setup
Figure 4.29: DEMEC reader and calibration bar used for creep and shrinkage 4.8.3 Results and Discussion
As shown in Figure 4.30, the incorporation of short steel fibers limited free shrinkage. Results at 330 days of drying showed that the free shrinkage of mix D-0f-90C-40, with no steel fibers, was about 1.5 times that of mix D-2f-90C-40 in which 2% fibers by volume were
4-39

incorporated. This suggests that the fibers offer some restraint to deformation, as would be expected when the fibers are well-dispersed in the matrix.
Figure 4.30 also show that after 365 days of drying, the free shrinkage of mix D-2f-23C40 (i.e. where no thermal treatment was applied) was about 4 times that of the mix D-2f-90C-40 which was thermally treated at 194oF (90oC) for 48 hours. In addition, results in Figure 4.30 also show that mix D-2f-23C-40 continued to shrink for the whole testing period, while mix D-2f90C-40 reached an asymptotic shrinkage strain value of about 60 after about 2 months of drying. As mentioned before, this behavior can be attributed to micro structural refinement and consumption of most of the mix water due to the accelerated hydration associated with thermal treatment. These results are very similar to the tensile creep study (Figure 4.30). It can be noticed from comparing results from both studies that free shrinkage from prism specimens (tensile creep study) was more than that of cylindrical specimens (compressive creep study). The high area-to-volume ratio of prisms used (1.8) versus 1.0 for cylinders may partially explain this behavior.
As for compressive creep, results in Figure 4.31 show that compressive creep decreased in the presence of steel fibers and with the application of thermal treatment prior to loading. Results from this study followed generally the same trends as tensile creep (Figure 4.27). However, tensile creep was found to be higher than compressive creep partially due to the magnified microcracking effect in the case of tension relative to compression (Figure 4.32). This observation requires further investigation.
4-40

Strain ()

300

D-C-2f-90C-40

D-C-2f-60C-40

250

D-C-2f-23C-40

D-C-0f-90C-40

200

150

100

50

0

0

50

100

150

200

250

300

350

Age (days)

500 450 400 350 300 250 200 150 100
50 0 0

Figure 4.30: Free shrinkage of different UHPC mixes

` D-C-2f-90C-40
D-C-2f-60C-40
D-C-2f-23C-40
D-C-0f-90C-40

50

100

150

200

250

300

350

Age (days)

Figure 4.31: Compressive total creep of different UHPC mixes

Strain ()

4-41

Strain ()

350
(a) 300

250

200

150

100
50
0 0

D-2f-90C-40 D-C-2f-90C-40

50

100

150

200

250

300

350

Age (days)

400 (b)
350

300

250

200

150

100 50 0 0

D-2f-60C-40 D-C-2f-60C-40

50

100

150

200

250

300

350

Age (days)

Strain ()

4-42

Strain ()

700 (c)
600

500

400

300

200
100
0 0

D-2f-23C-40 D-C-2f-23C-40

50

100

150

200

250

300

350

Age (days)

Strain ()

450 400 (d)

350

300

250

200

150

100

50

0

0

50

100

150

200

250

Age (days)

D-0f-90C-40 D-C-0f-90C-40

300

350

Figure 4.32: Comparison between tensile and compressive creep of UHPC, (a) 2f-90C-40, (b) 2f60C-40, (c) 2f-23C-40, and (d) 0f-90C-40

4-43

4.9 Conclusions Regarding UHPC Material Properties
Based on the materials study conducted to characterize the effect of thermal treatment and fiber reinforcement on the short and long-term performance of UHPC, the following conclusions were drawn: 1- Eliminating fiber reinforcement or thermal treatment significantly reduced the compressive
strength, tensile strength, and modulus of elasticity of UHPC. For use in bridge girders, both fibers and thermal treatment must be used to achieve the promise of UHPC. 2- Tensile strength tests showed that the splitting tension tests over-estimated the true tensile strength of UHPC. A conversion factor of can be used to predict true tensile strength of the material from splitting tension test results. 3- Short-term tests showed that the 7-day tensile strength of fiber reinforced specimens were about 8% greater than those without fibers; yet, tensile creep of specimens without fibers was up to 70% greater than the creep strain of specimens with fibers. These results emphasize the importance of conducting tensile creep testing rather than just tensile strength tests to predict long-term tensile performance of UHPC. 4- Autogenous shrinkage results showed that 87% of the total autogenous shrinkage occurred during the thermal treatment due to accelerated rate of early hydration at higher temperature. Also, the use of 2% steel fibers by volume resulted in a 42% reduction in autogenous shrinkage at 14 days of age after batching. 5- Short and long-term studies showed that the application of thermal treatment significantly decreased the tensile creep and free shrinkage of UHPC. 6- In contrast to tensile creep results provided in the literature for NSC and HPC, reinforcement with short straight steel fibers decreased the tensile creep of UHPC. It is proposed that this is, in part, due to the enhancements at the fiber/matrix interface during thermal treatment. 7- Results from this study suggest that the tensile creep phenomenon in UHPC occurs differently than compressive creep in UHPC. This result emphasizes the importance of further study of the tensile creep behavior of UHPC, particularly for applications where satisfactory long-term tensile performance is required. 8- Overall, this research suggests that the combined effect of thermal treatment and incorporation of short steel fibers act to enhance the short and long-term performance of UHPC, with thermal treatment having the most pronounced effect.
4-44

5. Field Evaluation of UHPC
5.1 Introduction
The purpose of the field evaluation was to determine if Ultra High-Performance Concrete (UHPC) technology could be successfully moved from laboratory to field use. To accomplish this, three field castings were performed at the Tindall Corporation, Georgia Division Plant in Conley, GA. For these castings, the UHPC used was Ductal manufactured by Lafarge Corp.
The first field casting occurred on July 28, 2007 and consisted of 1.5 yds3 of UHPC. The objective of this casting was to prepare 24 test cylinders and 21 modulus-of-rupture (MOR) beams. The second field casting occurred on August 18, 2007 and consisted of 3.25 yds3 of UHPC. The objective of this casting was to prepare 48 test cylinders, 39 shear-friction push-off specimens, and six 10-foot-long beams for the research on interface shear. These specimens were chosen for the second casting to further investigate batching of new Ductal premix before the construction of the bridge girders. The third and final casting occurred on September 15, 2007 and consisted of 12 yds3 of UHPC. The objective of this third casting was to prepare 60 test cylinders, two 34-ft long prestressed girders, and one 54-ft long prestressed girder.
During each batching and casting operation, the authors noted the following:
1) Observe how the precast plant's mixer would handle the material; 2) Adapt the mixing procedure to the plant's mixer; 3) Test the ability of the material to be cast directly from a concrete bucket; 4) Test the feasibility of casting large prestressed structural members at a precast plant; 5) Test the ability of the material to consolidate in full size bridge girders without internal or
external vibration; 6) Assess the ability of the plant to achieve specified thermal curing conditions for a full-
size member on an existing prestressing bed.
5.2 Pour 1
For the first casting of UHPC at Tindall was considered a trial batch; all 1.5 yds3 was batched simultaneously in Tindall's 4 yds3 shear mixer. The dry powders for this UHPC were provided as a premix in 80 lb (36 kg) bags. These bags were added one at a time by hand to the
5-1

mixer. The manufacturer's recommended procedure (Lafarge 2000c) was followed for the determination of the best mixing parameters for Tindall's equipment under the supervision of Mr. Peter Calcetas of Lafarge Corp.
The manufacturer recommended a maximum mixing temperature of 77 F (25 C). Because of the high ambient temperatures in Atlanta in July and the high amount of cement in the premix, this temperature was exceeded during batching of Pour 1. The maximum temperature experienced during mixing was around 85 F (29 C). This high temperature was ameliorated in future batches by substituting ice for some of the mixing water.
Due to the absence of coarse aggregate in UHPC, traditional slump cone measurements were not used. Instead a flow table test for mortars similar to that detailed in ASTM C1437 was used to gain an understanding of the material's workability as shown in Figure 5-1.
Figure 5-1: Flow table testing of UHPC. High fluidity was observed even prior to 20 drops. The concrete was placed in forms by means of large scoops; scooping the material was
difficult, but the constraints of the form size and the material volume needed made scooping the 5-2

material the best option. All beam-shaped specimens were cast from one end and the material was allowed to flow longitudinally down the beam.
After casting, the UHPC specimens were allowed to cure in their molds under ambient conditions for 48 hours. After 48 hours, all specimens were demolded in preparation for thermal treatment. Thermal treatment consisted of placing the specimens on either side of a steam line and then covering them with plastic. During this thermal curing, a high temperature of approximately 167 F (75 C) was achieved and held for 72 hours. Manufacturer specifications recommend a temperature of 200 F (95 C) for 48 hours, but allow for a lower temperature for a longer time. Graybeal observed that the lower temperature method of curing does not lead to as high compressive strengths (Graybeal 2005).
5.3 Pour 2
The purpose of the second pour at Tindall was to prepare 12 monolithic push-off specimens (Figure 5-2), 27 composite push-off specimens, six 10-ft.-long beams (Figure 5-3), and 48 cylinders.
Figure 5-2: Formwork for monolithic pushoff specimen prior to casting of UHPC.
5-3

Figure 5-3: Formwork for 10-ft long beams prior to casting of UHPC. New premix material was used for this pour and no large clumps of premix were observed in the plastic concrete. This increase in homogeneity made placement of the material easier by making it smoother to pour and scoop. Figure 5-4 shows the concrete from this batch flowing down a 10 foot long beam.
5-4

Figure 5-4: UHPC flowing down 10-ft. long beam
Flow table testing gave a diameter larger than 7.9 in (200mm) after 20 drops. This high value demonstrated that vibration was not necessary for placement of this batch of material. There was some difficulty in filling forms this small directly from the bucket. The bucket could not seal well enough to fully stop the flow of UHPC, which made it difficult to control the amount of material placed in the form. Several cubic feet of material were lost due to overfilling of molds and material being dropped on the ground as the bucket was transferred between forms.
Most specimens were finished by screeding the surface and immediately covering the specimen with 6 mil. plastic tight against the surface of the concrete. The screeding process was more difficult than with traditional concretes, largely due to the stickiness of the mix and the presence of steel fibers. These made excess concrete difficult to remove because it was difficult to isolate the amount to be removed. The most effective method of removing the concrete was to
5-5

use a screed board to push overfill to one side while using a scoop or trowel to remove this excess. Once an appropriate amount of concrete was in the forms, however, the self-leveling properties of the UHPC eliminated the need for any troweling.
Curing was similar to Pour 1 except that temperatures between 202 and 208 F (94 and 98 C) were maintained for the 48 hours of requisite thermal treatment as recommended because of the proximity of the specimens to the steam line.
5.4 Pour 3
Batching for pour 3 only differed noticeably from pour 2 in scale. Since 13 yds3 of UHPC were being cast, 4 batches were required. Ice was again substituted for 90% of the water to keep the mixing temperatures below 77 F (25 C). Each of these batches was mixed as previously described above and then transferred via concrete bucket to a Ready-mix truck. This Ready-mix truck was used because the 4.0 yd3 mixer at Tindall could not facilitate enough concrete for casting these specimens from one batch. The seals on the concrete bucket were still not tight enough, causing almost 1 yd3 of material to be lost before it could be transferred to the truck.
Flow table was performed on the concrete for pour 3. After the required 20 drops, the material flowed off of the table. This indicated that the final diameter was greater than 10 in. (254mm) and vibration was not necessary for placement of this batch of material.
In casting the UHPC girders, the most striking characteristics of the concrete were the very high flowability and the homogeneity of the material. These traits were noticeable in the casting of smaller specimens, but were much more so in the casting of the large girders Figure 5-5 shows the material as it pours down the chute of the Ready-mix truck. The material did not exhibit any visible segregation or bleeding whatsoever. Figure 5-6 shows the material as it is being poured into one end of the beam form and down the length of the beam.
5-6

Figure 5-5: Highly homogeneous UHPC pouring fluidly from Ready-mix truck.
Figure 5-6: Beam forms are filled by pouring UHPC in one end and allowing the material to flow along the axis of the beam. 5-7

The chute was slowly moved along the beam only after the material has filled the form at one end and has flowed more than 5 beam depths along the length of the bottom of the beam from the pour location. This placement method is important for several reasons. Ideally, the alignment of the fibers would be completely random leading to an isotropic material, but since some preferential alignment will occur in the direction of flow, this direction is best for both flexure and shear performance of the beam.
The curing process for the third pour was much more difficult than with any of the previous batches. Nineteen hours after casting, the first cylinder was broken to determine if the beams were strong enough to remove the forms. A compressive strength of 440 psi was measured, so the forms were not released at that time. At 42 hours after casting, the cylinder strength of the material was 15,460 psi, which was strong enough for the forms to be removed. The forms were removed between 44 and 47 hours after casting. The beams were covered with plastic whenever possible to prevent moisture loss.
The strands were cut beginning at 9:00am on 9/18/07, approximately 68 hours after casting. During the cutdown, one of the workers was injured causing a delay in beginning steam curing. The steam curing was begun at 6:00pm on 9/18/07 using the plant's steam system. After the initial 24 hours, Beam 1 averaged 144 F (62 C), and Beam 3 averaged 122 F (50 C). These values were 56 and 78 F (32 and 45 C) less respectively than the temperatures recommended by Lafarge. Due to the low steam temperatures, the beams were left to steam until they were moved to Georgia Tech on 9/28/07, 13 days after casting and 10 days after thermal treatment was begun. Figure 5-7 shows the UHPC strength gain with this curing.
5-8

Compressive Strength (psi)

30000 25000 20000 15000 10000
5000 0 0

steam curing

5

10

Days After Casting

Figure 5-7: Strength gain for UHPC steam cured in field. Low temperatures increased steaming time, yet, 30,000 psi strengths were not attained.

5.5 Field Evaluation Conclusions
Based on the background research and on the three batching and placements of UHPC at the Tindall precast concrete plant, the following six conclusions were drawn:
1) It is feasible to cast large structural UHPC members at precast plants in Georgia provided the plant personnel first gain experience using the material.
2) UHPC can be easily poured into large forms and placed without vibration. Scooping the material or placing it in smaller forms is very difficult because of the steel fibers.
3) In order to prevent plastic shrinkage cracking and blistering on the exposed surfaces of UHPC specimens, it is important to wet the surface and cover it with plastic immediately after pouring.
4) Sufficient thermal curing may be difficult to achieve at precasting plants. Portable steam generators may be used to supplement plant generators, and portable generators are commercially available at competitive prices.
5) The UHPC compressive strengths measured in field castings were 12% lower than those found from the laboratory batches. This reduced strength was probably due to the lower steam curing temperatures provided at the plant. Yet, the UHPC strengths of over 25,000

5-9

psi found from the field pours were significantly higher than the strength of high performance concrete currently used in Georgia bridges. 6) Combining multiple batches of UHPC in a Ready-mix truck is time consuming and equipment intensive. New methods of batching UHPC in Ready-mix trucks are being developed and should be evaluated to speed material placement.
5-10

6. Girder Behavior
6.1 Introduction
The goals of Task 5 were as follows: (1) to verify that a full size, high quality AASHTO type girder could be made with UHPC, (2) to determine the transfer length of 0.6-inch prestressing strand in UHPC girders, (3) to determine if UHPC girders can be constructed in Georgia, (4) to determine the shear capacity of the girders, (5) to determine the flexural capacity and ductility of a composite UHPC girder, and (6) to measure the creep and shrinkage deformations of the girders for determination of prestress loss.
Goals (1) and (3) were accomplished together with the field evaluation of UHPC as discussed in Chapter 5. Full-size UHPC girders can be constructed in Georgia, but special care must be taken to assure adequate thermal treatment.
Goal (4), shear capacity, will be accomplished as part of Georgia DOT Research Project No. 07-05, Task Order No. 02-37, Shear-friction and Interface Shear of UHPC. Two girders were specifically constructed to determine shear and interface shear of full-size precast prestressed composite girders. Due to the good results for analytical and laboratory results, it was determined that these two girders should be used for the 07-05 study. The reader is advised to review the final report for GDOT research project No. 07-05 after its completion in March 2009 for results of the shear tests on those two girders.
Goals (2), (5) and (6) were accomplished with the construction of a 54-ft long girder at Tindall Corporation, Conley, Georgia plant. Data from this single test girder provides the baseline information about UHPC bridge girder behavior and UHPC benefits which satisfy research project 2043 goals.
The original concept was to use an AASHTO Type I Modified girder as the test girder section. The analytical investigation showed in Chapter 3 that the Type I cross-section could only be used for girders up to 65-ft span lengths. A project goal was to develop a low-height girder with a span length greater than 80 ft. so that a two-span bridge where each span was 40 ft. could be replaced by a single 80-ft. span. Yet, the analytical investigation also showed that a 28in. deep bulb-tee section could be used for span lengths up to 92-ft. before being limited by deflection considerations. Therefore, the test girder modeled a small bulb-tee section 32 in. high
6-1

so that it more closely conformed to the findings of Chapter 3. This "modified" bulb-tee was 2 in. deeper so that the bottom flange could have an extra layer of strands in the bottom flange.
6.2 Test Girder
Figures 6-1 through 6-6 illustrate the three test girders, elevations and cross sections. The girders were reinforced with 28 bottom and 2 top 9/16-inch diameter grade 270 low relaxation prestressing strands as illustrated. Only Girder 1 is discussed in this report.
The 54-ft long test girder was designed primarily for testing its flexural capacity and to determine transfer length of the 9/16-in. strands and long-term prestress losses. Nevertheless, the girder will be retested as part of the shear-friction study to investigate the effect of shear reinforcement at the interface between the cast-in-place deck and the UHPC girder. Therefore, various spacings of the 2-#4 bar stirrups were used along the length of the girder as shown in Figure 6-1. Based on the material properties study, it was known that the shear capacity of the girder itself, without shear reinforcement, was sufficient to develop the girder's flexural strength. Therefore, about one-third of the girder was constructed with no shear reinforcement other than the fibers in the UHPC itself.
Each of the 9/16-in. prestressing strands was stressed to a load of 40.5 kips (210.9 ksi). Eight load cells mounted on the anchorage end of the prestressing bed verified that the jacking load was accurate.
Sixty-eight hours after the UHPC was placed, as described in section 5.4, the strands were cut-down. An accidental fire in one of the end forms caused the strands to heat irregularly, and many strands broke prematurely. The girders slid about 12-ft along the bed. Yet, there was no damage to the girder. The strength of the UHPC before cut-down was 15,460 psi. The girder was then covered and steamed cured for a week at about 144F. The 56-day, post thermal treatment, compressive strength of the girder UHPC was 24,890 psi. At test day, the compressive strength and modulus of elasticity were 24,425 psi and 7,366ksi respectively. The apparent decrease in strength from after 56 day is well within standard deviation and is not statistically significant.
The girder was then transported to the Georgia Tech Structural Engineering Laboratory. Formwork was erected and an 8-in. thick deck, 5-ft. wide deck was cast atop the girder as shown in Figures 6-7 and 6-8. As ordered, the deck concrete strength was designed to be 7,000 psi. Yet, at 90-days, the compressive strength of the deck concrete was 10,260 psi.
6-2

6-3

Figure 6-1. Girder 1 Elevation, 54-ft. long

Figure 6-2: Test Girder 1 Cross Section 6-4

4 Spaces at 3 in. # 4 Stirrups
2 in.

11 in.

11 in. 2 #5 lifting loops
12 in.

Beam Length = 34 ft.
6 Spaces at 24 in. # 4 Stirrups, With Formliner

2 3/4in.

5 Spaces at 12 in. # 3 Doghouse Bars

11 in. 2 #5 lifting loops
1 Spaces at 120 in. No Formliner

11 in.
4 Spaces at 3 in. # 4 Stirrups
2 in.

5 Spaces at 12 in. # 3 Doghouse Bars

2 3/4 in.

1 ft.

Bearing to Bearing Length = 32 ft.

1 ft.

ELEVATION VIEW

Figure 6-3. Girder 2 Elevation, 33-ft. 2 in. long, xx-in. wide top flange

6-5

4 Spaces at 3 in. # 4 Stirrups
2 in.

11 in.

11 in. 2 #5 lifting loops
12 in.

Beam Length = 34 ft.
17 Spaces at 12 in. # 4 Stirrups, No Formliner

2 3/4in.

5 Spaces at 12 in. # 3 Doghouse Bars

11 in. 2 #5 lifting loops
1 Spaces at 120 in. With Formliner

11 in.
4 Spaces at 3 in. # 4 Stirrups
2 in.

5 Spaces at 12 in. # 3 Doghouse Bars

2 3/4 in.

1 ft.

Bearing to Bearing Length = 32 ft.

1 ft.

ELEVATION VIEW

Figure 6-4. Girder 3 Elevation, 33-ft. 2 in. long, xx-in. wide top flange

6-6

2 13/32 in

18 in

8 in

No. 4

stirrups

2 in

3 in

7 in

No. 5 lifting loops No. 4 stirrups

24 in

17 13/32 in

1 1/4 in C.L.

4 in

7 in

2 in 2 in 2 in 2 in
2 in

3 in 7 in

7 spaces at 2 in

2 in

18 in

SECTION AT MIDPOINT

No. 3 doghouse bars
1 1/8 in C.L.
SECTION AT END

Figure 6-5: Test Girder 2 Cross Section

No. 4 U-bars

6-7

2 13/32 in

9 in 3 1/2 in

2 1/2 in

No. 4 stirrups
3 15/16 in 1 1/16 in

No. 5 lifting loops No. 4 stirrups

24 in

17 13/32 in

1 1/4 in C.L.

4 in

7 in

2 in 2 in 2 in 2 in
2 in

3 in 7 in

7 spaces at 2 in

2 in

18 in

SECTION AT MIDPOINT

No. 3 doghouse bars
1 1/8 in C.L.
SECTION AT END

Figure 6-6: Test Girder 3 Cross Section

No. 4 U-bars

6-8

Figure 6-7. Construction of the composite deck in the Structures Laboratory 6-9

6-10

Figure 6-8. Drawing of the composite deck slab showing instrumentation

6.3 Transfer Length
DEMEC gauge points, spaced 2-in. on-center, were embedded in the bottom flange at each end and on each side of the test girder over a distance of 36 in. The gauge points were located at the center of gravity of the prestressing strands. The distance between points was measured with a digital, mechanical extensometer 10-in. long. Readings were taken before cutdown and each day after cut-down for two-weeks and periodically thereafter. The technique recommended by Russell (1992) was used to average the strain readings and to find the transfer length (lt).
AASHTO LRFD gives the transfer length as lt = 50db where (db) is the diameter of the prestressing strand, which would give a 28-in. transfer length of the test girder. The mean transfer length determined for the four ends of the test girder at one year was 15.4 in., about 55% of the AASHTO prediction. Figure 6-9 illustrates a typical transfer length graph taken from the southeast end of the test girder.

Strain (microstrain)

1000 900 800 700 600 500 400 300 200 100 0 0

Transfer Length Beam 1 SE

Smoothed Data Points Plateau 95% Plateau

15.4

5

10

15

20

25

30

Distance from End (in)

Figure 6-9. Typical transfer length plot showing increasing strains to a constant plateau. Where the strain line crosses 95% of the constant strain is lt .

6-11

It was concluded that the transfer length of the strands was much less than the specified length; therefore, the strands are adequately developed in the UHPC. Further, the length over which shear reinforcement must be provided due to lower prestress forces is conservatively predicted by the AASHTO requirement.
5.4 Instrumentation and Test Setup
Figures 6-8 and 6-10 illustrate the instrumentation plan and test setup for the test girder. LVDTs are used to measure strains in order to develop a strain profile and to determine slip between the composite deck and the UHPC girder. The two point loading is designed to give a short section of constant moment so that maximum strains can be developed over a 4-ft. section. Figure 6-11 shows the girder setup prior to testing. The internal vibrating wire strain gages were monitored during the test.
5.5 Flexural Test Results and Discussion
The predicted load deflection response of the composite section is given in Figure 6-12. The tension-controlled behavior that was predicted would have allowed for large deflections prior to failure, so the test was to be terminated when the strain in the prestressing strands at midspan reached 2%. Each half of the girder could then have been reused for the shear-friction study.
The load was applied quasi-statically using a 1,000 kip capacity Enerpac hydraulic cylinder with a 12-in maximum stroke. The loading was paused every 15,000 lbs (or 0.4 in. once the load began to plateau) in order to record strain from the internal vibrating wire strain gages, verify deflections manually, and observe crack spacing and propagation.
6-12

6-13

Figure 6-6. Elevation showing instrumentation and loading of test girder.

Figure 6-11 Prestressed UHPC girder with composite deck prior to flexural testing.

Yielding Cracking Dead Load and Prestress

Ultimate

Figure 6-12 Predicted load-deflection response of test girder 6-14

The load-deflection curve was linear up to a load of 72,400 lb. The least squares fit of the slope was 127,032 lb/in with an adjusted R Square value of 0.997. This value is 29% lower than the calculated slope of 179,700 lb/in. One possible explanation for this lower stiffness is a reduced effective width of the deck due to shear lag effects. Another explanation could be microslipping between the girder and the deck causing a reduction in the stiffness of the composite system. Apparent nonlinearities in the measured strain profile of the structure may support either of these hypotheses. Figure 6-13 shows a strain profile taken during the linear region of the load deflection curve. At the interface between the girder and deck, there is an apparent strain discontinuity of approximately 0.0001.
Figure 6-13 Strain profile of girder and deck at applied load of 60,000 lbs. Notice the discontinuity in the profile between the girder and the deck.
6-15

The strain discontinuity could represent gradual slippage of the deck, but the LVDT slip gages did not measure any significant deflection until a load of 70,000 lbs as shown in Figure 614. Since the deck strains were measured along the outside edge of the deck, this could represent shear lag phenomena where the edges of the deck were not acting fully with the rest of the system. This theory is supported by the lack of slip at low load, but there were no gages attached to the underside of the deck with which to measure differences in longitudinal strain across the width of the deck.
At a load of approximately 70,000 lbs, an instantaneous slip of .008 in was observed on the east end of the deck. This was accompanied by a loud cracking sound, and an immediate loss of 3,400 lbs of load. A crack was observed to have formed along the interface between the girder and the deck extending from the eastern load point all the way to the east end of the girder (Figure 6-15). The load of 70,000 lbs corresponds to an interface shear force of 325 kips according to AASHTO LRFD 5.8.4.2. This is 25% greater than the interface shear capacity of 259.2 kips predicted by the specification.
From this point on in the test, the slope of the load deflection curve decreased significantly, and became much less regular. It is the opinion of the authors that this corresponded to partial loss of composite action as the concrete to concrete bond failed along the interface. From this point on, shear friction along with dowel action became the main mechanism for carrying shear across the interface.
The complete load deflection curve is shown in Figure 6-16 along with the theoretical load deflection curve for the composite girder system. Since composite action was not maintained throughout the test, the theoretical load deflection curve for the non-composite girder and deck is also shown. These predictive models are based on values obtained from materials testing and the as-built dimensions of the structure. As mentioned previously, the initial stiffness observed in the experimental data was lower than expected, but the data showed an initial trend along the theoretical composite curve with a transition toward the theoretical non-composite curve as the beam lost composite action. Another change in slope can be observed at around 120,000 lbs as the prestressing steel reached its proportional limit. This again corresponds closely with yield value from the non-composite model.
6-16

Figure 6-14 Slip between girder and deck at different loads. Notice the sharp change in behavior of the east end after 70,000 lbs.
Interface crack
Figure 6-15 Interface crack between UHPC girder and HPC bridge deck. 6-17

Figure 6-16 Theoretical and experimental load-deflection curves. With the exception of a low ultimate strength, the girder behaved as would be expected for a partially composite system.
The ultimate failure of the girder occurred unexpectedly following a peak load of 174,480 lbs. A loud cracking and slip of the deck indicated a complete loss of composite action. At this point, the crack along the interface became large enough to see through. This corresponded with a drop in load to approximately 150,000 lbs. At that time, ultimate failure occurred by crushing of the concrete at the top of the girder, and the applied load dropped to 7000 lbs. Even considering the reduction of performance due to non-composite action, the girder was expected to carry 35% more load and deflect almost twice as much before failing. Results of strain data obtained from LVDTs showed that the ultimate strain in the concrete prior to failure was approximately 0.0021. This is 30% less than the design value of .003 and 28% less than the value of 0.0034 reported for UHPC (Graybeal, 2005).
6-18

Close inspection of the failure location showed a plate-like morphology of the failed surfaces (Figure 6-17 and 6-18). This failure was similar in appearance to laminates of fiberreinforced polymer compressed parallel to the direction of lamination and may give an explanation of the low ultimate failure. The direction of flow of the UHPC described in Chapter 5 may have prevented a significant percentage of the fibers from orienting orthogonal to the web surface. When the web and top flange were stressed in compression, Poisson effects would have created significant tension in this direction. Without fibers to restrain deformation in this direction, a delamination occurred between parallel plates of concrete-fiber composite, significantly decreasing the failure load and strain at failure. To this date, much of the work on fiber orientation has been concerned with providing sufficient tensile reinforcement in areas of high tensile stress. The effect of fibers on Poisson effect tensile failures in high compression regions has not been evaluated. Further investigation is needed to quantify the effects of fiber orientation on compressive strain and stress capacity of UHPC.
Figure 6-17 Compression failure of UHPC girder web. Notice the plate-like pieces of spalled UHPC that make up the crack edges 6-19

Figure 6-18 Close-ups of plate-like cracking along compression failure surface.
5.6 Flexural Test Conclusions
UHPC was shown to be a very promising material for the creation of prestressed girders with cast-in-place composite decks. Before this material is used in field applications, however, more research should be done to correctly quantify the interface shear performance of the bond between UHPC girder and cast-in-place deck. This is particularly important given the very high prestressing forces and live loads these girders are expected to resist. This high prestressing force drastically increases the demand on the interface compared to previous, standard designs. Further understanding will be gained and design equations will be suggested as part of Georgia DOT Research Project No. 07-05, Task Order No. 02-37, Shear-friction and Interface Shear of UHPC.
Finally, before UHPC can be used confidently in field construction, the strain and stress capacity of the material must be quantified in relation to the fiber orientation experienced in field casting of the material. Particularly, the effect of longitudinal fiber orientation on transverse
6-20

Poisson effects must be studied to prevent brittle compression failures of highly stressed UHPC. After completion of the interface shear testing at Georgia Tech, forensic evaluation of different girder elements should be conducted to quantify in-situ material properties for use in further design.
6-21

7. Conclusions and Recommendations
7.1 Conclusions
The principal conclusion from this study is that UHPC may be effectively used for precast prestressed bridge girders in Georgia; yet, interface shear between the UHPC girder and cast-in-place deck must be specially designed. The analytical studies and the results from other researchers indicate that either a Bulb-Tee shape girder like Virginia's BT-29 or a pi-shaped girder like that developed by the FHWA is required to have a section less than 30-in. deep which will span over 90-ft. and will satisfy AASHTO deflection, strength and serviceability requirements. The experimental studies presented herein and by others show that the UHPC by itself will satisfy shear and diagonal tensile strength of the girders; shear reinforcement is not required for shear strength. Yet, shear reinforcement is necessary to adequately connect the girder to a composite deck for Bulb-Tee and for AASHTO shaped cross sections. The composite deck is required to provide stiffness to minimize deflection and for ultimate strength flexure requirements.
The materials study concluded that thermal curing at temperatures above 170F are required to develop the high compressive and tensile strength and to bond the fibers to the paste so that the excellent shrinkage and creep properties are attained. The very low tensile creep of adequately cured UHPC means that the diagonal tension, shear strength of girders can be maintained over the service life of a UHPC bridge. Lack of adequate thermal treatment could mean that long-term tensile creep could result in diagonal tension, shear concerns.
The field evaluation demonstrated that UHPC girders can be constructed in Georgia, but that special attention and special specifications are required to assure that adequate thermal treatment (curing) of the UHPC is assured. Plants using UHPC should be prequalified before being permitted to use UHPC for GDOT structures.
The transfer length of 9/16-in. diameter strands in UHPC girders is about 55% of the transfer length given by the AASHTO specifications. The UHPC provides improved force transfer between the strand and concrete as compared to normal strength concrete.
Current AASHTO LRFD specifications can be used for the design of UHPC girders, but the shear design must be considered separately using elastic mechanics concepts and an ultimate
7-1

tensile strength of 1,800 psi. Reinforcement should be provided to carry 100% of the interface shear.
7.2 Recommendations
It is recommended that the research on use of UHPC for bridge girder applications in Georgia be continued as follows:
1) The planned interface shear and shear friction tests must be conducted and design recommendations provided for UHPC girders.
2) Long-term durability of UHPC. While the FHWA and other tests have shown outstanding durability of UHPC, the authors have observed small rust spots on the surface of UHPC girders stored at an FHWA laboratory. Architecturally, such rust spots may be undesirable. A study should examine such rusting.
3) Demonstration bridge. A 60-ft to 90-ft simple span bridge should be constructed using UHPC to observe the bridge long-term behavior. The girders and deck should be instrumented so that prestress losses can be determined. Load and vibration tests would determine the sensitivity of persons to such a flexible structure. The girders should be constructed using a bulb-tee section, like the Virginia BT-29 section.
7-2

References
ACI Committee 116, "Cement and Concrete Terminology", in ACI Manual of Concrete Practice. American Concrete Institute: Farmington Hills, MI. 2000, pp. 116R.1-116R.73.
ACI Committee 363, "State-of-the-Art Report on High-Strength Concrete: Chapter 1 Introduction (Approved)". American Concrete Institute, 2005.
ACI Report 544.2R-78, ACI Journal and Proceedings, Vol. 75, No. 7, 1978, pp. 283-289.
Akita, H., Koide, H., Tomon, M., Sohn, D. (2003), "A practical method for uniaxial tension test of concrete", Materials and Structures Vol. 36, No. 260, July 2003, pp. 365-371.
Altoubat, S. A., and Lange, D. A. (2001), "Tensile Basic Creep: Measurements and Behavior at Early Age", ACI Materials Journal, Vol. 98, No. 5, September-October, 2001, pp. 386-393.
Anderson, B. G. (1957) "Rigid Frame Failures," ACI Journal, Proceedings, Vol. 53, No. 7, January 1957, pp. 625-636.
ASTM C190, "Standard Test Method for Tensile Strength of Hydraulic Cement Mortars," American Society for Testing and Materials Standard Practice C496, Philadelphia, Pennsylvania, 1990.
ASTM C293, "Standard Test Method for Flexural Strength of Concrete (Using Simple Beam With Center-Point Loading)," American Society for Testing and Materials Standard Practice C496, Philadelphia, Pennsylvania, 2008.
ASTM C496, "Standard Test Method for Splitting Tensile Strength of Cylindrical Concrete Specimens," American Society for Testing and Materials Standard Practice C496, Philadelphia, Pennsylvania, 1990.
ASTM C512, "Standard Test Method for Creep of Concrete in Compression," American Society for Testing and Materials Standard Practice C496, Philadelphia, Pennsylvania, 2002.
ASTM C78, "Standard Test Method for Flexural Strength of Concrete (Using Simple Beam with Third-Point Loading)," American Society for Testing and Materials Standard Practice C496, Philadelphia, Pennsylvania, 2008.
Banthia, N., MacDonald, C., and Tatnall, P. (1999), "Structural Applications of Fiber Reinforced Concrete", ACI International SP 182, 1999.
Bazant, Z. P., and Xi, Y. (1994) "Drying Creep of Concrete: Constitutive Model and New Experiments Separating its Mechanisms." Journal of Materials and Structures, Vol. 27, 1994, pp. 314.
Bissonnette, B. (1996), "Le Fluage en Traction: un Aspect Important de la Problematique des Reparations minces en Beton", DSc Thesis in Civil Engineering, Universit LAVAL, Qubec, Canada, December 1996.
Bissonnette, B., and Pigeon, M. (1995), "Tensile Creep at Early Ages of Ordinary, Silica Fume and Fiber Reinforced Concretes", Cement and Concrete Research, Vol. 25, N0. 5, 1995, pp. 1075-1085.
R-1

Bissonnette, B., Pigeon, M., and Vaysburd, A.M. (2007), "Tensile Creep of Concrete: Study of Its Sensitivity to Basic Parameters", ACI Materials Journal, Vol. 104, N0. 4, 2007, pp. 360368.
Cheyrezy, M., Maret, V., and Frouin, L. (1995), "Microstructural Analysis of RPC (Reactive Powder Concretes", Cement and Concrete Research, Vol. 24, No. 7, October 1995, pp. 14911500.
Collepardi, S., Coppola, L., Troli, R., and Collepardi M. (1997), "Mechanical properties of modified Reactive powder Concrete", ACI international Conference on "Superplasticizers and Other Chemical Admixtures in Concrete", SP 173, October 1997.
Cucchiara, C., Mendola, L.L., and Papia, M. (2004), "Effectiveness of Stirrups and Steel Fibers as Shear Reinforcement", Cement and Concrete Composites, Vol. 26, 2004, pp. 777-786.
De Larrard, F. and Sedran, T. (1994) "Optimization of Ultra-High Performance Concrete by The Use of a Packing Model," Cement and Concrete Research, Vol. 24, No 6, pp. 997-1009.
Dodson, V. H. (1990), "Concrete Admixtures", Van Nostrand Reinhold, New York, 1990. Ductal Manual of Batching and Testing Procedures.
Elkem,http://www.multigrout.elkem.com/hits/web_0006.nsf/lupstartpage/downloadenginter"(las t accessed in June 2008).
Ferron R.P., Gregori, A., Sun, Z., and Shah, S. P. (2007), "Rheological Method to Evaluate Structural Buildup in Self-Consolidating Concrete Cement Pastes", ACI Materials Journal, Vol. 104, No. 3, May-June 2007, pp. 242-250.
Feylessoufi, A., Crepin, M., Dion, P., Bergaya, F., Van Damme, H., and Richard, P. (1997), "Controlled Rate Thermal Treatment of Reactive Powder Concretes", Advanced Cement Based materials, Vol. 6, No. 1, 1997, pp. 21-27.
Findley, W.N., J.S. Lai, and K. Onaran (1989), "Creep and Relaxation of Nonlinear Viscoelastic Materials : With an Introduction to Linear Viscoelastic": Dover, 1989.
Furlan, Sydney Jr., and Hanai, J.B. (1999), "Prestressed Fiber Reinforced Concrete Beams with Reduced Ratios of Shear Reinforced", Cement and Concrete Composites, Vol. 21, 1999, pp. 213-221.
Furlan, Sydney Jr., and Hanai, J.B. (1997), "Shear Behaviour of Fiber Reinforced Concrete", Cement and Concrete Composites, Vol. 19, 1997, pp. 359-366.
Garas, V.Y., Silva, J., Kurtis, K.E., Kahn, L.F., (2009) Ultra-High Performance Concrete Mix Design, Task 2 Report, Ultra High Performance Concrete for Precast Prestressed Bridge Girders, Georgia DOT Research Project No. 2043, Jan 2009.
Gere, J.M. (2001), Mechanics of Materials, 5th Edition, Brooks/Cole, 2001.
Gettu, R., and Barragan, B.E. (2003), "Direct tension test and interpretation", Rilem Proceedings (Pro 31) Test and Design Methods for Steel Fibre Reinforced Concrete Background and Experiences -, 2003, pp. 15-30.
Goodspeed, C.H., S. Vanikar, and R.A. Cook (1996), "High-performance concrete defined for highway structures". Concrete International, 18(2): 1996, pp. 62-67.
R-2

Graybeal, B. and Hartmann, J. (2003) "Strength and Durability of Ultra-High Performance Concrete," Proceedings of the 3rd International Symposium on High Performance Concrete PCI National Bridge Conference, (Orlando, FL).
Graybeal, B., (2005) Characterization of the Behavior of Ultra-High Performance Concrete, Ph.D. Thesis, Department of Civil and Environmental Engineering, The University of Maryland, College Park, MD, USA, May 2005.
Graybeal, B.A., (2004) "Fabrication of an Optimized UHPC Bridge," Proceedings of the PCI National Bridge Conference, Atlanta, Georgia, Oct. 17-20, 2004.
Graybeal, B.A., (2006) Material Property Characterization of Ultra-High Performance Concrete, FHWA-HRT-06-103, Federal Highway Administration, U.S. Department of Transportation, 2006.
Graybeal, B.A., (2006 b) Structural Behavior of Ultra-High Performance Concrete Prestressed I-Girders, FHWA-HRT-06-115, Federal Highway Administration, U.S. Department of Transportation, 2006.
Gutsch, A., Rostasy, F.S. (1994), "Young concrete under High Tensile Stresses, Creep, Relaxation and Cracking", in: R. Springenschmid (Ed.), Thermal Cracking in Concrete at Early Age, E & FN Spon, London, RILEM, 1994, pp.111 118.
Habel, K., Denari, E., and Brhwiler, E. (2007), "Experimental Investigation of Composite Ultra-High-Performance Fiber-Reinforced Concrete and Conventional Concrete Members", ACI Structures Journal, Vol. 104, No. 1, January-February 2007, pp. 93-101.
Hanna, N. (1997), "Steel Fiber Reinforced Concrete Properties and Resurfacing Applications", Portland Cement Association, Skokie, Ill., Report RD049.01P, 1997.
Hannant, D. J. (1978), "Fibre Cements and Fibre Concretes", John Wiley & Sons, 1978.
Imam, M., Vandewalle, L., Mortelmans, F., and Van Gemert, D. (1997), "Shear Domain of Fibre-Reinforced High-Strength Concrete Beams", Engineering Structures, Vol. 19, No. 9, 1997, pp. 738-747.
Keierleber, B., Bierwagen, D., Wipf, T., Abu-Hawash, A., (2008) "Design of Buchanan County, Iowa, Bridge, Using Ultra High-Performance Concrete and Pi Beam Cross Section," Proceedings of the PCI National Bridge Conference, Orlando, Florida, Oct. 5-7, 2008.
Kormeling, H.A., and Reinhardt, H.W. (1987), "Strain rate Effects on Steel Fibre Concrete in Uniaxial Tension", The International Journal of Cement Composites and Lightweight Concrete, Vol. 9, No. 4, February 1987, pp. 197-204.
Kovler, K. (1999), "A New Look at the Problem of Drying Creep of Concrete under Tension", ASCE Journal of Materials in Civil Engineering, Vol. 11, No. 1, February, 1999, pp. 84-87.
Kovler, K. (1995), "Interdependence of Creep and Shrinkage for Concrete under Tension", ASCE Journal of Materials in Civil Engineering, Vol. 7, No. 2, May, 1995, pp. 96-101.
Kovler, K. (1994), "Testing System for Determining the Mechanical Behavior of Early Age Concrete under Restrained and Free Uniaxial Shrinkage", Journal of Materials and Structures, Vol. 27, 1994, pp. 324-330.
R-3

Kovler, K. (1996), "Why sealed Concrete Swells?", ACI Materials Journal, Vol. 93, No. 4, 1996, pp. 334-340.
Kovler, K., Igarashi, S., and Bentur, A. (1999), "Tensile Creep Behavior of High Strength Concretes at Early Age", Journal of Materials and Structures, Vol. 32, 1999, pp. 383-387.
Krenchel, H., (1974) "Fiber Reinforced Concrete," American Concrete Institute, ACI SP-44, 1974, pp. 45-77.
Krstulovic-Opara, N., and Malak, S. (1997) "Tensile Behavior of Slurry Infiltrated Mat Concrete (SIMCON)", ACI Materials Journal, Vol. 94, No. 1, January-February 1997, pp. 39-46.
Kumar, S. and Santhanam, M. (2004) "Use of a Particle Packing Model to Produce HPC at Optimum Cement Content," The Indian Concrete Journal. pp. 22-28.
Kumar, V.S. and Santhanam, M. (2003) "Particle Packing Theories and Their Application in Concrete Mixture Proportioning: A Review," The Indian Concrete Journal, pp. 1324-1331.
Lafarge Corp. (2000) "Operating Procedure: Cylinder and Prism Preparation," Ductal Batching and Testing Procedures, Reference # T002, Version 2.1, Lafarge Corp. 2000. pp. 1-4.
Lafarge Corp. (2000) "Operating Procedure: Flow Test," Ductal Batching and Testing Procedures, Reference # T006, Version 2.1, Lafarge Corp. 2000. pp. 1-2.
Lafarge Corp. (2000) "Optimisation of the Mixing of Ductal Concretes," Ductal Batching and Testing Procedures, Reference MOBOXX, Lafarge Corp.; 2000. pp. 1-6.
Lankard, D. R. (1975), "Fiber Concrete Applications", (Fibre Reinforced Cement and Concrete), The Construction Press LTD, England, 1975.
Lim, D.H., and Oh, B.H. (1999), " Experimental and Theoretical Investigation on the Shear of Steel Fibre Reinforced Beams", Engineering Structures, Vol. 21. 1999, pp. 937-944.
MacGregor, J.G., and Wight, J.K. (2005), "Reinforced Concrete, Mechanics and Design", 4th Edition, Prentice Hall, 2005.
McHugh, A.J., and Tan, S.R. (1993), " Mechano-Chemical Aspects of the Processing/property/Structure Interactions in Macro-Defect Free Cement", ACBM Journal, Vol. 1, October 1993, pp. 2-11.
Mehta, P.K. and Monteiro, P. J. M. (2005), "Concrete Microstructure, Properties, and Materials" 3rd ed: McGraw-Hill, 2005.
Mehta, P.K. (1999), "Advancements in Concrete Technology", Concrete International, Vol. 21, No. 6, June 1999, pp. 69-76.
Mehta, P.K., and Atcin, P.C. (1990), "Principles Underlying the Production of HighPerformance Concrete," Cement, Concrete, and Aggregates, ASTM, Vol. 12, No. 2, 1990, pp. 70-78.
Mindess, S., and Young, J.F., (1981) Concrete, Prentice-Hall Inc., Englewood Cliffs, New Jersey.
Mindess, S., Young, J. and Darwin, D. (2002) Concrete, Second Edition, Prentice Hall, New Jersey.
R-4

Monosi, S., Pignoloni, G., Collepardi, S., Troli, R., and Collepardi M. (2000), "Modified Reactive Powder Concrete with Artificial Aggregate", ACI International Conference on "Superplasticizers and Other Chemical Admixtures in Concrete", SP 175, October 2000.
Moore, B., Bierwagen, D., (2006) "Ultra High Performance Concrete Highway Bridge," Proceedings of the PCI National Bridge Conference, Grapevine, Texas, Oct. 23-25, 2006.
Naaman, A.E., and Homrich, J.R. "Tensile Stress-Strain Properties of SIFCON", ACI Materials Journal, Vol. 86, No. 3, May-June 1989, pp. 244-251.
Naaman, A.E., and Otter, D., and Najm, H. (1991), "Elastic Modulus of SIFCON in Tension and Compression", ACI Materials Journal, Vol. 88, No. 6, November-December 1991, pp. 603611.
Narayanan, R., and Darwish, I.Y.S. (1987), "Shear in Prestressed Concrete Beams Containing Steel Fibres", The International Journal of Cement Composites and Lightweight Concrete, Vol. 9, No. 2, May 1987, pp. 81-90.
Nawy, E.G. (2006), Prestressed Concrete, A Fundamental Approach, 5th Edition, Prentice Hall, 2006.
Pickett, C. (1942), "The Effect of Change in Moisture-Content on the Creep of Concrete under a Sustained Load", ACI Journal, Vol. 38, 1942, pp. 333-356.
Potrzebowski, J.(1983), "The Splitting Test Applied to Steel Fibre Reinforced Concrete", The International Journal of Cement Composites and Lightweight Concrete, Vol. 5, No. 1, February 1983, pp. 49-53.
Powers, T. C., and Brownyard, T. L. (1948), "Studies on the Physical Properties of Hardened Portland Cement Paste", Bulletin 22 of the Portland Cement Association, Chicago, IL, March 1948.
Poyola, G., Kriven, W. M., and Young, J. F. (1990), " Providing Interfaces in Macro-Defect Free Cement", Cementing the Future, Vol. 12, No. 2, Summer 1990.
Richard, P. and Cheyrezy, M. (1994) "Reactive Powder Concretes with High Ductility and 200 800 MPa Compressive Strength," Proceedings of the V. Mohan Malhotra Symposium on Concrete Technology Past, Present and Future ACI International Conference, SP-144.
Richard, P. and Cheyrezy, M. (1995) "Composition of Reactive Powder Concretes," Cement and Concrete Research, Vol. 25, No 7, pp. 1501-1511.
Rilem TC 162-TDF: "Test and Design Methods fir Steel Fibre Reinforced Concrete, Uniaxial Tension Test for Fibre Reinforced Concrete", Materials and Structures, Vol.34, 2001, pp. 36.
Rossi, P. (2001), "Ultra-High-Performance Fiber-Reinforced Concretes", Concrete International, Vol. 23, 2001, pp. 46-52.
Rossi, P., Coussy, O., Boulay, C., Acker, P., and Malier, Y. (1986), "Comparison between Plain Concrete Toughness and Steel Fibre Reinforced Concrete Toughness", Cement and Concrete Research, Vol. 16, 1986, pp. 303-313.
R-5

Russell, Bruce W. (1992), "Design Guidelines for Transfer, Development and Debonding of Large Diameter Seven Wire Strands in Pretensioned Concrete Girders," Ph.D. Thesis, The University of Texas, Austin, Texas, pp 464
Shah, S. P. (1984), "Fiber Reinforced Concrete," Handbook of Structural Concrete, Editors F. K. Kong, R. H. Evans, E. Cohen, and F. Roll, McGraw-Hill Book Company, New York, 1984.
Shah, S. P., and Weiss, W. J. (1998), "Ultra High Strength Concrete; A Look to the Future" ACI Special Proceedings from the Paul Zia Symposium Atlanta, GA, 1998.
Shah, S.P., and Weiss, and W.J. (1998) "Ultra High Strength Concrete; A Look to the Future" ACI Special Proceedings from the Paul Zia Symposium, Atlanta, GA., 1998
Sritharan, S., Bristow, B.J., and Perry, V. (2003) "Characterizing an Ultra-High Performance Material for Bridge Applications Under Extreme Loads," Proceedings of the 3rd International Symposium on High Performance Concrete PCI National Bridge Conference, (Orlando, FL).
Swamy. R. N., and Barr, B. (2006), "Fibre Reinforced Cements and Concretes: Recent Developments", Elsevier Science Publishing Co., Inc, 1989.
Tao, Z., and Weizu, Q. (2006), "Tensile Creep due to Restraining Stresses in High-Strength Concrete at Early Ages", Cement and Concrete Research, Vol. 36, 2006, pp. 584-591.
Thelandersson, S., Martensson, A., and Dahlblom, O. (1988), "Tension softening and cracking in drying concrete." Journal of Materials and Structures, Vol. 21, No. 6, 1988, pp. 416424.
Todd J.D. (1955) "The Determination of Tensile Stress-Strain Curves for Concrete", Proceedings, Institution of civil Engineers, Vol.4, No.2, Part 1, London, March 1955, pp. 210-211.
Umehara, H., Uehara, T., Iisaka, T., and Sugiyama, A. (1994), "Effect of Creep in Concrete at Early Ages on Thermal Stress", in: R. Springenschmid (Ed.), Thermal Cracking in Concrete at Early Age, E & FN Spon, London, RILEM, 1994, pp.79 86.
Vakili, J. (1983), "An Experimental-Study of Asphalt Concrete Based on a Multiple-Integral Representation of Constitutive Equation of a Non-Linear Viscoelastic Solid", Journal of Rheology, Vol. 27, 1983, pp. 211-222.
Vakili, J. (1984), "Creep Behavior of AsphaltConcrete under Tension", Journal of Rheology, Vol. 28, N0. 5, 1984, pp. 573-580.
Vicenzino, E., Culham, G., Perry, V.H., Zakariasen, D., Chow, T.S.(2005) "First Use of UHPFRC in Thin Precast Concrete Roof Shell for Canadian LRT Station," PCI Journal, V. 50, No. 5, Sept.-Oct. 2005, pp. 50-67.
R-6