Investigation on bond in post‐installed anchored bars and lap splices subjected to direct tension

Moment‐resisting monolithic connections between new and pre‐existing structural members are frequently used whenever existing reinforced‐concrete structures must be extended and/or strengthened. Overlapping post‐installed and cast‐in‐place bars is a common solution to provide reinforcement continuity. Nowadays, post‐installed bars with high‐performance mortars can reach higher bond strength than cast‐in‐place bars configurations. While higher performance could lead to shorter embedment lengths, uncertainties and constraints from design standards limit the potential for cost‐effective solutions. In this work, an experimental investigation on post‐installed lap splices is presented with some novelties in terms of test setup and specimen layout. The bond of a high‐performance injection mortar used in all specimens was assessed by means of tension tests. Lap length, concrete strength, and confinement by transverse reinforcement are the variable factors in the study of the load–slip behavior of the bars and the splitting resistance of the specimens. Results are in good accordance with the design formulas available in standards and empirical data published in the literature.


| INTRODUCTION
Post-installed (PI) reinforcing bars (rebars) provide a practical way to extend and strengthen existing structural systems, as well as to solve the problems arising from misplaced or missing cast-in-place (CI) rebars 1 due to mistakes on the construction site.In such applications, PI rebars are typically anchored using qualified injection mortars in the connections and are either installed as end anchorages or overlapped with CI rebars.In the latter case, PI lap splices are formed with the CI reinforcement and a safe transfer of forces both at the ultimate and at the serviceability limit state becomes possible.The design of PI lap splices is based on standard procedures for CI lap splices. 2,3Therefore, although nowadays some qualified injection systems have demonstrated much higher bond strength, 4,5 the potential improvement of PI lap Discussion on this paper must be submitted within two months of the print publication.The discussion will then be published in print, along with the authors' closure, if any, approximately nine months after the print publication.
splices is not exploited due to uncertainties in performance and execution as well as lack of research.Thus, further experimental testing is needed to validate design models and to improve numerical simulations toward reliability, efficiency, and to eventually open innovative fields of application.Testing on PI lap splices is not as extensive as on PI anchorages and many assumptions rely on database from CI lap splices. 6here is no common consensus on a standard specimen and test setup to evaluate PI lap splices. 4The injection of a mortar to bond rebars adds uncertainties to the complex load transfer mechanism of splices.In this scenario, it is difficult to find a standard recommendation for testing the bond strength in terms of main influence factors.An experimental approach requires sound analysis and qualification of the testing setup. 7A widely accepted layout should be affordable, suitable for repetitions, and provide enough confidence in the results.Standard tests for CI lap splices may not be appropriate for PI rebars or need significant changes.
This study aims at experimentally investigating PI lap splices injected with high-performance mortars to characterize and design concrete splitting.Reference is made to the most relevant parameters affecting bond behavior.The focus is on short-term static loads.Therefore, the influence of fatigue and long-term actions is out of scope.First, pull-out tests on the bond of PI rebars in confined and unconfined conditions are presented.The influence of concrete strength and embedment length is assessed and compared to reference tests of CI rebars.Second, the performance of PI lap splices with short lap length is evaluated.A specimen layout and a test setup are proposed.Comparisons are made with some "reference tests" concerning CI lap splices.The splitting failure and the load-slip curves of the individual rebars in the splices are studied considering key parameters such as transverse reinforcement and concrete strength.
The measured data is compared to design formulas from construction standards, analyzing the deviation and robustness of the tests.The results and conclusions should contribute to a better understanding of PI lap splices and the conduction of future experiments.

| Installation and failure modes
The installation procedure of PI rebars basically consists of drilling into existing concrete, removing the dust, injecting a mortar in the borehole, and placing the rebar before the mortar starts hardening. 1In a lap splice configuration, safe interaction between CI and PI rebars depends on the bond stresses of each rebar and the resistance of the structural section against internal stresses (Figure 1a).PI rebar connections are designed to transmit tensile or compressive forces, in contrast to anchors that are also designed against shear forces as well. 1,8Depending on the nature of the confinement (active ensuring from an applied pressure, or passive exerted by the transverse reinforcement) different failure modes may occur (Figure 1b-d); concrete cone, rebar pull-out, concrete splitting or steel yielding of the rebar. 9

| Design and experiments
PI lap splices are designed as moment-resisting monolithic connections to provide continuity of the reinforcement and avoid relying on the tensile strength of the concrete in the construction joint.Some typical applications are presented in Figure 2.
Qualification of injection mortars for PI rebars is needed to quantify the bond performance and their sensitivity to loading and environmental conditions in order to assure safe connections during the entire working life. 10It requires tests against mechanical loads and durability conditions.No experiments on PI lap splices are prescribed.In Europe, the standardization of PI rebar connections started with the TR 23 11 which in the meantime was replaced by the EAD 330087. 12Since then, qualified systems are designed as if they were CI anchorages or lap splices covered by reinforced concrete standards like Eurocode 2 13 and considering special influence factors for PI rebars specified in the assessment documents.The same applies in North America with the assessment document ICC ES AC 308 14 and the reinforced concrete standard ACI 318. 15The design strength is, therefore, limited by such factors like cracked concrete, temperature, and drilling inaccuracy, whose influence can be minimized by employing qualified personnel for the installation and by carefully considering geometric constraints like concrete cover and rebar spacing.The design equations proposed in reinforced concrete standards come from CI rebars and are mostly semi-empirical.Assuming these equations for PI rebars goes mainly back to the conclusions drawn by Spieth, 2 following his tests and numerical simulations.It is sensitive to note that neither Eurocode 2, 13 nor ACI 318 15 include specific provisions for PI reinforcement.
Though lap splices are today frequently adopted in PI moment-resisting connections, the experimental evidence on their performance in actual construction is scantly indeed. 6The most comprehensive tests on PI lap splices were presented by Spieth, 2 who conducted a systematic investigation on the splitting failure of PI lap splices in beams and plates subjected to bending.Additionally, lap splices were tested in direct tension for a more detailed evaluation.Analysis was focused on the impact on bond strength and slip behavior of different mortars, transverse reinforcements, geometric configurations, and rebar positions in the section.In the experimental campaigns by Randl and Kunz, 5 PI lap splices in connections between cantilevers and simply supported concrete slabs were evaluated using two different types of injection mortars.Varying the splice length and transverse reinforcement, they arrived to the main conclusion that bond stiffness can play an important role in PI rebar connections with high-strength injection mortars.A higher bond stiffness can clearly reduce the transfer length of a certain embedment length.Therefore, pull-out tests on members with small embedment lengths (i.e., 5 rebar diameters) cannot describe the bond stiffness of PI rebars with an embedment length of 20 rebar diameters or higher.
The increasing application of PI lap splices to retrofit and strengthen existing structures requires a more detailed assessment of the load transfer capacity.Nowadays, injection mortars are developed to improve the pullout strength.Nevertheless, PI lap splices have typically small concrete cover and splitting failure hardly allows PI rebar higher bond strength to be fully exploited.On the one hand, modern injection mortars with high stiffness should be tested to avoid bond stress concentrations and premature splitting cracks in service and the ultimate state.On the other hand, it has become increasingly relevant in PI lap splices to reduce installation efforts and material by finding cost-effective solutions.Innovative construction materials, new measuring techniques and the evolution of the design codes require further experimental campaigns and the introduction of standard test methods in lieu of the many bond-test methods developed in the past and still used nowadays.

| General
The bond performance of single PI rebars was quantified in pull-out experiments in confined and unconfined conditions.The failure mode, the bond strength, and the loadslip relationship were evaluated for different embedment lengths.Other factors affecting bond behavior were kept constant throughout the experimental campaign.
Pull-out tests consisted of a plain concrete cylinder (confined setup) or a prism (unconfined setup) with a PI or a CI rebar bonded axially at center under tensile forces.The installation procedure of PI rebars followed the recommendations in the technical assessment documents of the injection mortar.The rebars for both confined and unconfined test conditions were made of steel type B500B according to DIN 488-1, 16 with a nominal diameter of 20 mm and a related ribbed surface greater than 0.056.The mechanical properties of three rebar specimen tested in the laboratory are presented in Table 1.Concrete strength class C20/25 according to Eurocode 2 13 was used in both confined and unconfined tests.In the latter, specimens with concrete strength class C40/50 were also considered.The ready-mixed concretes were cast under an average temperature of 24 C and cured inside the laboratory with a plastic cover to maintain the humidity constant.The mix design and the mechanical properties of the two concretes are listed in Table 2.The concrete compressive strength, splitting tensile strength, and elastic modulus were tested according to DIN EN 12390 [17][18][19] on the same day of the experiments using cylindrical specimens stored under the same conditions as the specimens.The test procedure for the pullout tests met the requirements described in the EOTA Technical Report 48. 20The specimens were placed centered in a 1000 kN universal testing machine and the upper end of the embedded rebar was clamped to the pulling head (see Figure 3).The load was applied to the rebar at a rate of 0.5 mm/min through a servo-hydraulic cylinder.The relative displacements between the rebars and the surrounding undisturbed concrete were recorded continuously employing linear variable displacement transducers (LVDT). 21Such a relative displacement coincides with bar-concrete slip, provided that the bar elongation is neglected (as usually done).

| Confined condition
Details of the test setup and specimens for the confined pull-out experiments are shown in Figure 3a.The concrete strength class of all specimens was C20/25.A seamless steel jacket type S355 with 300 mm diameter and 250 mm height was used as formwork to provide full confinement of the concrete around the rebar.Smooth polytetrafluoroethylene (PTFE) and hardboard sheets of 0.5 mm and 3.0 mm thickness, respectively, were placed between the steel plate and the concrete.These layers reduced friction and made the reaction forces evenly distributed. 22The steel plate provided enough confinement to prevent concrete cone failure.The test was performed until the rebar was completely pulled out of the concrete or until the rebar yielded.
A total of 12 specimens was tested.Table 3 summarizes the average results of the 4 different layouts (3 specimens are replicants).Measured values of slip and load at loss of adhesion (s u,adh ; F u,adh ) and at maximum strength (s max ) are presented.The calculation of the bond strength τ is based on Equation ( 1), where the stress distribution is assumed uniform over the embedment length of the rebar l b , F max is the maximum applied load, and d s is the For an embedment length of 140 mm, PI rebars reached yielding without loss of adhesion.The measured displacement is attributed only to elastic deformations of the rebar.By contrast, CI rebars with the same embedment length failed due to pull-out at a load around half of the yielding load.In other words, the bond strength of PI rebars under tensile static load is at least two times greater than the bond strength of a CI rebar in the same layout.
The averaged maximum load for specimens with PI rebars and embedment length 100 mm are in the range of the steel yielding strength.In fact, one of the rebars was pulled out while the other two failed due to steel yielding and remained in the concrete cylinder.The loss of adhesion occurs when the displacement on the loaded end reaches 1.4 mm.
The PI rebars with an embedment length of 70 mm show an ultimate load 43% greater than the specimens with CI rebars and an embedment length of 140 mm.The loss of adhesion happens when the displacement of the bar on the loaded end is 1.0 mm, 28% less with respect to the case of PI rebars with an embedment length of 100 mm.In contrast, the loss of adhesion in CI rebars begins with a displacement of 0.5 mm, which is 92% less than PI rebars with embedment length of 100 mm.
The load-slip curve of CI rebars is smooth in appearance, while in the case of PI rebars it has relatively marked variations in slope.This observation indicates a less uniform distribution of the bond stress.Figure 5 displays photos with a distance scale of CI and PI pulled-out rebars.While CI rebars have a homogeneous distribution of residuals from the concrete, PI rebars have zones where the loss of adhesion happened between the bar and the mortar (lower part), and other zones where it happened between the mortar and the concrete (upper part).The transition between the two zones approximately occurs in the mid-length section of the embedment.

| Unconfined conditions
The test was designed to allow an unrestricted formation of the failure cone in concrete.There was no confinement close to the concrete surface around the rebar.The specimens were fixed to the floor of the testing machine with steel beams and threaded bars.Elastomers with 10 mm thickness were placed between the concrete and the beams to redistribute the load, reduce the potential confinement, and reduce the rate of load application during earlier stages of the test (see Figure 3b).
Four specimens with different layouts were tested.They consisted of concrete prisms of dimension 400 Â 780 Â 620 mm 3 .The mechanical properties of the embedded rebars are specified in Table 1.The embedment length of the PI rebars varied from 120 mm to 140 mm.The influence of concrete strength on the bond performance was evaluated by employing two strength classes, C20/25 and C40/50 (see Table 2).The layout of the specimens and the results regarding maximum load F max and corresponding bond strength τ (refer to Equation 1) are presented in Table 4.All specimens mainly failed in a mixed mode, with the formation of a concrete cone accompanied by the pull-out of the bar. Figure 6 shows measured curves of tensile load versus slip on the loaded end.A strong variation of the bond capacity was observed when the concrete strength changes.A 44% higher concrete strength delivered ultimate load increase between 22% and 24%.Specimen with 120 mm embedment length showed more slip at maximum load than those with 140 mm embedment length.14% more embedment length yielded between 4% and 7% increase of the ultimate load, which indicates that the equivalent bond strength decreased for a larger embedment length.
Figure 7 shows the rebars after being pulled out of the concrete.In all specimens, the concrete cone formed along the first 2/3 of the embedment length.The cone angle was approximately 54 .Failure in specimens with concrete strength class C40/50 happened at the interface between the rebar and the injection mortar at approximately the last 1/3 part of the embedded length.In specimens with concrete strength class C20/25, failure always occurred between the injection mortar and the concrete.

| General
According to the results in the pull-out experiments, PI rebars with high-performance adhesives bonded in normal strength concrete show higher bond strength than CI rebars.The question arises whether the higher bond strength of PI rebars can be exploited when lapped with

| Specimen layout
The layout of the specimens represents the tensile region of a flexural member in a concrete-to-concrete connection (Figure 8a) designed to fail due to concrete splitting around the plane of the lap splices (Type A failure, according to Eligehausen 27 ).
A few parameters influencing bond performance in PI lap splices have been considered.The "reference" layout complied a lap length of 240 mm, concrete strength class C20/25, and no transverse reinforcement.It was compared using the one-factor-at-a-time approach to specimens with a shorter lap length of 140 mm, a concrete strength class C30/37, and higher confinement around lap splices using transverse reinforcement.For reference, CI lap splices with concrete strength class C20/25 and lap length of 140 and 240 mm were tested.
The geometry of the specimens consisted of a rectangular concrete prism with one pair of CI rebars located at the bottom, overlapped with one pair of either PI or CI rebars at the top (Figure 8b).The prism length was adjusted to the corresponding lap length.All other dimensions were fixed.The nominal diameter of the rebars was 20 mm.The two lap splices had a spacing of 160 mm and were symmetrically positioned in the central plane of the concrete prism to minimize flexural and torsional stresses.The clear distance between overlapped rebars was 20 mm and the concrete cover was 40 mm, complying with the regulations of Eurocode 2 13 and the EAD 330087 12 for PI rebar connections.
Each rebar had a de-bonded length of 80 mm (see Figure 8b) designed to prevent any influence of the concrete surface and avoid cone failure.Polyvinyl chloride (PVC) tubes of 40 mm diameter were placed in the deboned zone.The annular gap between the rebar and the PVC tube was sealed using pipe insulation with 22 mm intern diameter and 9 mm thickness.Possible misalignments between the drilling direction and the nominal axis of the bore for the installment of the PI rebars were minimized. 12To such end, a pilot PVC pipe with 20 mm diameter was installed before casting the concrete where rebars will be post-installed (see Figure 9).After a short curing period, the pilot pipe was removed from the specimen, leaving a 20 mm diameter hole in the concrete, which allowed a more precise drilling direction to enlarge the hole to 25 mm.Additionally, a plastic tube of a nominal diameter of 25 mm covered with foam tape was glued to the PI rebar on the debonded length to create a weak bond zone.
All PI rebars were bonded axially to the specimens with an injected mortar system after 14 days from casting the concrete.The procedure followed the steps provided by the European Technical Assessment (ETA) of the injection system.To keep the rebars vertical during the hardening process of the mortar, supporting elements were placed around the PI rebars and removed before the test.

| Materials
The PI and CI rebars in all specimens were of type B500B according to DIN 488-1. 16They had a nominal diameter of 20 mm, a related ribbed surface greater than 0.056, and a metric threaded end type M24.The mechanical properties of the rebars according to tests conducted in the laboratory are summarized in Table 5.
The details on the two concrete mixtures and the corresponding mechanical properties are listed in Table 6.A cement type CEM I 42,5 R was used in both cases.8][19] The cylinders were cured in the same conditions as the specimens and tested when the experimental campaign began.

| Test setup
All specimens were tested in a 1000 kN universal testing machine.The components of the test setup are shown in Figure 10.The load was applied to the upper rebars through the loading head while the lower rebars were fixed to the reaction base of the machine.Hemispherical nuts were placed between the plate and the screw nuts at the end of each rebar to allow small rotations. 28The steel plates were fixed to the reaction base and the pulling head by prestressing the steel beams using threaded rebars.
The tensile force introduced into both lap splices should be equally distributed to minimize bending moments and lateral forces.To this purpose, the position of the specimens in the testing machine was carefully controlled.The upper and lower surfaces were horizontal and parallel to the loading head and reaction base of the testing frame.Each rebar had a strain gauge installed outside the concrete.Before running the tests, the rebars were prestressed with 10-15 kN tensile force by fixing the nuts manually until the strains in all of them were almost equally distributed.The tests were displacement controlled, with a displacement rate of 0.5 mm/min.They were conducted until the splitting failure of the specimens occurred.
The slip between the rebar and the concrete was recorded through Digital Image Correlation (DIC) technique. 29Coordinates and displacements of point-based

Rebar
Threaded end n f tk (N/mm measurements in strategic positions on the specimens were measured with a frequency of 1 Hz.

| RESULTS AND DISCUSSION
In total, 12 specimens in 6 different layouts were tested.
For each layout, two nominally-identical specimens were tested as proof of test repeatability.An overview of the specimens and the ultimate tensile load of a lap splice layout F max averaged from both lap splices in each specimen are summarized in Table 7.For each layout, the mean value F max and the COV were calculated.The two tests performed for each layout showed consistent results, considering that the variation of the ultimate load between specimens with the same layout was always less than 5%.The statistic difference is relatively small and there is no sign of significant bias in the experiment due to, that is, geometry uncertainties or asymmetric load introduction.In the following sections, the evaluation of load application, load-slip relationship, failure mode, and influence parameters will be discussed.

| Role of the misalignment of the applied load
Possible misalignments of the tensile loads applied to the rebars were evaluated by placing strain gauges on each bar 5 cm apart from the concrete face.The symmetry of the load application depends on the accuracy with which the specimen is built and centered in the machine.Despite extreme care, geometric irregularities could not be avoided, and tolerances of ±1 mm must be accepted as errors in craftsmanship.
Figure 11 shows the percentage difference in strains between the upper (No. 1 and 2) and the lower rebars (No. 3 and 4) of the specimen.For clarity, only one specimen of each layout is presented.It can be noted that most of the load differences are in a range of less than 10% at failure.There is a tendency in specimens with higher deviations that the load differences between rebars are reduced when the overall pulling load of the machine becomes higher.

| Evaluation of bond-slip relation
The roles played on lap-splice behavior by such parameters as lap length, concrete strength and transverse reinforcement are investigated in this section.The average value of the strains and displacements measured at the loaded sections of the two bars protruding from the top of each specimen (bars No. 1 and 2) was calculated for each test, and the same was done for the two bars protruding from the bottom (bars No. 3 and 4).

| Influence of the lap length on PI lap splices
The load-slip behavior of splices with 140 mm and 240 mm lap length are presented in Figure 12.In both layouts, the PI rebars on top have less slip than the CI rebars at the bottom.However, the higher stiffness of PI rebars turned out to be independent from the lap splice length.The load-slip behavior between PI rebars and between CI rebars is not influenced by decreasing the lap splice length from 240 mm to 140 mm.The failure occurs F I G U R E 1 0 Section of the direct tensile test setup and specimens' dimensions in mm.when one of the lower rebars reaches a slip value of 0.38 mm in shorter lap splices and 0.52 mm in larger lap splices.The ultimate load of specimens with a lap length equal to 140 mm is 33% less than specimens with a lap length of 240 mm.Nevertheless, the bond strength is higher in shorter lap splices.Whereas the bond strength of longer lap splices is 6.7 N/mm 2 , the bond strength in shorter lap splices is 8.0 N/mm 2 .

| Comparison between PI and CI lap splices
In Figure 13, the differences between CI and PI lap splices are highlighted.In case of a lap length of 140 mm, the failure occurs when CI rebars reach a slip between 0.37 mm and 0.42 mm.PI rebars show higher stiffness in  comparison to all CI rebars.The slip at maximum load is less than 0.22 mm.No signs of debonding are observed in load-slip curves of PI rebars.In PI lap splices with an overlapping length of 240 mm, the failure starts when the slip in the rebars is 0.54 mm on the CI side and 0.39 mm on the PI side.Rebars on the CI side (bottom) show less stiffness and a higher slip, compared to the rebars on the PI side (top).On contrary, in CI lap splices the top rebars present a higher slip than the lower rebars.
In Figure 14, comparisons are made among the load-slip curves of PI lap splices characterized by different concrete grades and transverse-reinforcement amounts.The load-slip curve of the PI rebars in the case of a higher concrete strength present a similar behavior to PI rebars in the reference case.Nonetheless, CI rebars in higher concrete strength behave differently since they have almost no de-bonding phase before the formation of main splitting cracks.These  bars failed when the slip is 0.60 mm.In the case of higher confinement around the lap splice due to transverse reinforcement, differences concerning the standard layout become obvious.The load-slip behavior of PI rebars becomes stiffer, influencing the displacement of the CI rebars, which shows higher slip values for the same loads.At the ultimate load, the slip in the PI rebars is 0.40 mm, approximately the same as in the standard case.By contrast, slip in the CI rebars is 1.00 mm, while in the standard case the slip is 0.54 mm.

| Evaluation of the failure mode
Specimens fail due to splitting on the side face, as shown in Figure 15.The main cracks are formed in the plane of the lap splices.In the case of CI lap splices, cracks are formed all around the prism, crossing the top and bottom faces between the rebars.As for the PI lap splices, the splitting cracks originate principally around the CI rebars at the bottom.A transverse crack is also observed on the PI side, where the lap splice begins.This behavior is observed in all specimens with concrete class C20/25 and no transverse reinforcement, which indicates a common crack pattern according to the layout.The crack pattern of the specimens with PI lap splices and concrete strength class C30/37 is similar to that of the standard case (Figure 16a).On contrary, the presence of transverse reinforcement around the lap splice modifies the crack distribution (Figure 16b).Longitudinal splitting cracks have smaller width than the reference case.PI rebars provide higher stiffness to the upper part of the specimen, while the transverse reinforcement also provides higher confinement around the spliced rebars.

| Evaluation of the bond strength
Influence factors like lap length, concrete strength, and transverse reinforcement are compared to evaluate the influence on bond behavior and the response of the test setup to highlight potential variations.In Figure 17, the ultimate load of the alternative layouts in contrast to the standard case is shown.The percentage of length variation from a splice between 240 mm and 140 mm is 42%, while the average variation of the ultimate load between both layouts is 39%.The use of higher strength concrete and an inclusion of transverse reinforcement around the lap splice improved the bond strength by 27% and 28%, respectively, on average.
In general, PI and CI lap splices exhibit approximately the same ultimate load, confirming that the failure mode mostly depends on the confinement around the lap splice (i.e., concrete cover), the concrete strength, and the bond strength of the "weaker" rebar in the connection (CI rebar).The higher bond strength of the PI rebar is therefore limited by these influence factors.The test results are compared with the design models to evaluate their accuracy to predict maximum stresses.The comparison is performed through the ratio between empirical and calculated values, meaning that 1.00 would be a perfect correlation.First, the mean ultimate stress in pullout tests is compared to reference values from the models in Eurocode 2 13,30 and the corresponding ETA approval of the injection mortar.Second, the mean ultimate stress in the rebars of each lap splice test is compared to models proposed by standards (Eurocode 2 13 and ACI 318 15 ), guidelines (Model Code 2010 31 and ACI 408 32 ), and Eligehausen. 27

| Pullout tests
In Table 8, the results of the pullout tests performed in unconfined conditions are presented.To calculate the mean resistance of the PI rebars with the formula proposed in Eurocode 2 -Part 4, 30 a combination of concrete cone and pullout was considered as the main failure.The factor for noncracked concrete was 11.0 and the mean compressive strength of the concrete was used instead of its characteristic value.The relation between the experimental and the calculated value indicates a good correlation for both specimens with the same embedment length (140 mm) but different concrete strengths.However, experimental results of specimens with embedment lengths equal to 120 mm showed higher resistance (between 20% and 30%) compared to the reference values of the model in Eurocode 2.
The maximum tensile stresses of the pullout tests in a confined condition and their relation to reference values are shown in Table 9.A mean bond strength of 18.67 N/ mm 2 for the PI rebars against confined pullout failure was obtained from the characteristic value presented in the ETA approval of the injection mortar.A ratio of 0.75 between mean and characteristic (5% quantile) values 1 was considered.A mean bond strength of 6.68 N/mm 2 for the CI rebars was adopted from the formulas in the model of Eurocode 2. 13 The PI rebars had between 43% and 51% higher maximum load.The CI rebars showed a maximum mean stress 38% larger than the calculated value.However, this difference can be explained by a factor introduced in the Eurocode 2 33 that accounts for the effect of cracked concrete and significantly decreases the reference values of the design bond strength.

| Lap splice tests
In Table 10, the comparison between the lap splice test results and design models is presented.The design values given in the standards and guidelines were converted into T A B L E 1 0 Lap-splice test results vs. design models.Formulas and input parameters are presented in Appendix A.

Specimen
Mean f test (N/mm 2 ) mean values by removing the material safety factors and assuming a recommended ratio between mean and characteristic (5% quantile) values of 0.75. 1 The ACI 408, 32 the Model Code 2010 31 and the model of Eligehausen 27 presented better accuracy in predicting the ultimate strength, while the standards Eurocode 2 13 and ACI 318 15 yielded the most conservative ratios.All design models provided safe predictions.Larger differences are observed when evaluating a different concrete strength and accounting for transverse reinforcement around the lapped bars, yet they remain overall on the safe side.

| CONCLUSIONS
An experimental investigation on post-installed rebars used in anchorages and lap splices is presented in this paper, where proposals are also made for the test setup and for the layout of the specimens with a framework characterized by cover splitting.To this purpose, a highperformance injection mortar was used for different values of concrete strength and transverse reinforcement amount.The test results have also been compared with the predictions based on recent design standards.Concerning post-installed rebars used in anchorages, the test results lead to the following conclusions: • In confined conditions, post-installed rebars exhibit roughly a three-time increase in the bond strength compared to cast-in-place rebars.In unconfined conditions, however, the higher bond strength of a single post-installed rebar with an embedment length between 120 and 140 mm cannot be fully exploited because of a mixed pullout and concrete-cone formation, that controls anchorage failure.• Post-installed rebars exhibit higher stiffness, which affects the load-slip relationship.In confined pull-out tests, the displacement measured at the loaded end at loss of adhesion of post-installed rebars was about half of the displacements observed for cast-in-place rebars.
Concerning post-installed lap splices, the test results lead to the following conclusions: • Post-installed lap splices have their bond strength limited by concrete splitting due to the poor tensile strength of the concrete.• Post-installed lap splices exhibit a higher bond stiffness than cast-in-place lap splices, especially in the confined conditions provided by transverse reinforcement.• Post-installed and cast-in-place lap splices characterized by the same bar arrangement approximately fail at the same tensile load.
• Post-installed and cast-in-place lap splices have different crack patterns due to the higher stiffness of the mortar and the brittleness of the concrete surrounding the splice, that is locally activated by the higher bond stresses.• Post-installed lap splices exhibit a higher strength whenever higher-grade concretes and higher amounts of transverse reinforcement are adopted.According to the results of this project, the strength increase may be close to 30%.• The ultimate capacity provided by different design models considered in this paper is consistent with the experimental results, and it is on the safe side.The comparison indicates that these models provide an overall safe design.
Further testing to assess the splitting performance of PI lap splices is recommended, especially in the case of high-performance mortars.For this purpose, the test setup proposed by the authors is versatile.It can be used with different base materials, be easily adapted to accommodate different geometries, and lead to reproducible and reliable results.

2
Typical PI lap splice applications (a) beam and slab connections (b) joint at the foundation of a column or wall.

3
Sections of the pull-out test setup in (a) confined and (b) unconfined conditions and specimens' dimensions in mm.rebar diameter.Average curves of the four layouts are shown in Figure 4, where the load-slip relation is compared.According to the European Assessment Document (EAD) for PI rebar connections, 12 the failure load of tensile tests should show a coefficient of variation (COV) that does not exceed 15%.The maximum average load and its COV fell within the range of expected scatter of pull-out tests with CI rebars and PI rebars.

FF
I G U R E 5 (a) PI and (b) CI rebars pulled out of the specimens.T A B L E 4 Results of pull-out experiments in unconfined conditions.Specimen l b (mm) f cm,cube (N/mm 2 ) F max (kN) τ (N/mm 2 ) I G U R E 6 Load-slip relation of PI rebars under unconfined tensile load.CI rebars.An experimental program is conducted to evaluate the bond behavior of PI lap splices in direct tension.The test setup and specimen layout adopted in this project were based on previous studies on CI lap splices.2,[23][24][25][26]

F I G U R E 7
Specimens after concrete cone failure of the PI rebars.

F I G U R E 9
Formwork of specimens before casting concrete.T A B L E 5 Characteristics of the steel rebars used in the tests.

1
Percentage strain difference during loading of the specimens between (a) rebars 1 and 2, (b) rebars 3 and 4. F I G U R E 1 2 Load-slip relation.Comparison between PI lap splices with different lap length.

F
I G U R E 1 3 Load-slip curves.Comparison between CI lap splices and PI lap splices.

4
Influence of (a) higher concrete strength and transverse reinforcement on the bond-slip curve.Splitting cracks after failure.

6
Splitting cracks after failure.F I G U R E 1 7 Comparison of the measures to improve bond performance in post-installed lap-splices.MODELS 6.1 | General
T A B L E 1 T A B L E 2 Mixture proportions and mechanical properties of concretes for the pull-out tests.cm (N/mm 2 ) f ctm (N/mm 2 ) E (N/mm 2 ) Results of pull-out experiments in confined conditions.
T A B L E 3 F I G U R E 4 Average load-slip relation of PI and CI rebars under confined tensile load.
Mix design and mechanical properties of concretes for the tests.
2)f max (N/mm 2 ) f su (N/mm 2 ) ε su E s (N/mm 2 )F I G U R E 8 Specimens' layout with dimensions in mm.
30ll-out tests in unconfined conditions vs. EC-2 design models.30Formulasand reference values from ETA approval document concerning the mortar are detailed in Appendix A.
Table A1 reports the different formulas used to compare the design models with the test results.The input parameters for Table 8, Table 9, and Table 10 (Chapter 7) are shown in Table A2, Table A3, and Table A4, respectively.T A B L E A 1 Design models to calculate the rebar tensile stress in anchorages and lap splices.Áψ e Áψ s Áψ g