Impact Resisting Mechanisms of Shear-Critical Reinforced Concrete Beams Strengthened with High-Performance FRC

Reinforced concrete (RC) structures typically present brittle failures by shear or punching under impact loading. High-performance fiber-reinforced concrete (HPFRC) has great potential due to its superior strength and energy absorption. The higher price and environmental cost of HPFRC compared to conventional RC can be effectively overcome by partially strengthening impact-sensitive RC members with HPFRC. To study the feasibility of this technique, HPFRC was applied as a tensile layer at the bottom of RC beams. Drop weight impact tests were carried out on beams with two values (35 and 55 mm) of HPFRC thickness, in addition to companion RC beams. Results show that the impact response can be divided into two stages: a first stage governed by local effects and shear plug formation at midspan, and a second stage governed by global beam behavior with formation of shear web cracks. A new resisting mechanism was observed for beams strengthened with HPFRC, as the strengthening layer worked similarly to a stress ribbon retaining the damaged RC and reducing fragmentation-induced debris. Such mechanism was fully achieved by the specimens with 35 mm HPFRC layer but was limited for the specimens with 55 mm HPFRC layer due to impact-induced interface debonding.


Introduction
Concrete structures might be subjected to low to heavy impacts during their service life. Either accidental (rock falls, tornado-debris) or human-induced (terrorist attacks) impact events can produce severe damage or even failure of concrete structures if the impact scenario has not been addressed in the design stage. As the concern of impact hazards has increased in recent years, the capacity of many existing concrete structures which were not designed against impacts is under discussion. In such situations, strengthening or retrofitting structural concrete members can be required [1][2][3].
It has been extensively reported that concrete structures subjected to impact exhibit a high sensitivity to develop brittle failures by shear or punching before yielding to longitudinal reinforcement, even in cases of structures designed to show a ductile response under quasi-static loading [4][5][6][7]. The differences between impact and quasi-static behavior are mainly due to three factors [8][9][10]: (a) inertia forces under dynamic loading lead to a time-dependent distribution of sectional forces different from the one under quasi-static loading; (b) material properties of concrete and steel are affected by the strain rate; (c) impulsive loading and wave propagation produce local damage near the impact zone. Due to the former factors, low energy-absorbing shear failures become dominant for reinforced concrete (RC) beams. In addition, spalling of concrete cover and fragmentation can The application of HPFRCC or UHPFRC to dynamic scenarios can be favored by the positive strain-rate influence on the material properties, as other studies [14,[30][31][32][33][34] have concluded that the tensile properties (peak strength, strain-hardening stage, and fracture energy) are kept in the high strain-rate domain (1-100 s −1 ). The rate sensitivity of UHPFRC in bending has also been verified in impact tests [35,36]. At the structural level, to the authors' knowledge, only Habel and Gauvreau [1] have carried out impact testing on composite UHPFRC-RC slab strips. They studied two configurations with the UHPFRC layer on the compressive and tensile sides of the specimens, respectively. For the specimens with the UHPFRC on the compressive part, the strengthening layer provided good protection against crushing and spalling on the impact zone, but such a configuration did not take advantage of the most remarkable properties of the strengthening material (i.e., strain-hardening and crack-bridging in tension). For the specimens with the UHPFRC on the tensile part, the cantilever test configuration and the limited measuring equipment were insufficient to fully explore the impact response of the UHPFRC-RC system.
In the present paper, the impact performance of shear-critical RC beams strengthened with a thin layer of high-performance fiber-reinforced concrete on the tensile part is studied. Strengthening solutions are necessary for existing structures because even RC members designed to develop a ductile flexural failure mode under quasi-static loading might have shear failures under impact loads [4][5][6][7]. The structural configuration studied in this paper is intended to utilize the superior strength and energy absorption capacity in tension of the strengthening material. By means of the localized application of the strengthening layer at the tensile side, a solution is achieved with smaller costs than those requiring massive material use like R-HPFRCC or R-UHPFRC (which in addition are not applicable to strengthen existing structures). Furthermore, the strengthening material has a more moderate fiber amount included (2% by volume, still able to achieve strain-hardening in tension) than common UHPFRC, thereby contributing to moderate economic and environmental impact. Drop-weight impact tests were carried out on strengthened beam specimens without stirrups. Two configurations were studied for the strengthening layer, as well as a reference series of conventional RC beams tested for comparison.

Test Configuration and Set-Up
Drop-weight testing has shown to be an effective experimental configuration to investigate the response of concrete structures under impact at a laboratory [37]. In the present research, three series of beam specimens were tested with the drop-weight testing machine of the Laboratory of Structures at the Technical University of Madrid (UPM, Spain). The experimental campaign is intended to evaluate the possibilities of a high energy-absorbing strengthening layer for RC beams without stirrups (i.e., with a high tendency to present brittle shear failure, local damage, and fragmentation under impact events).
Tested specimens were 2000 mm long and had a rectangular cross-section of 125 mm (width) × 250 mm (height). The longitudinal reinforcement consisted of 2 steel bars of 16 mm diameter at the top and bottom of the beams ( = 1.6%). A clear cover of 50 mm was considered between the bars and the beam edges in order to allow the installation of a strengthening layer on the tensile side. The test set-up and cross-sectional configuration of each series are plotted in Figure 1. No shear reinforcement was installed, except for two stirrups of 8 mm diameter on the supports to enhance anchorage of longitudinal bars. Reference series RC-U0 consisted of conventional RC beams, while series RC-U35 and RC-U55 included a 35-and 50-mm thick tensile layer of high-performance fiber-reinforced concrete, respectively, to simulate rehabilitation or strengthening of reference series. The reinforcement layout of reference series RC-U0 is a typical one of shear-critical RC members [12], which can be designed according to the most current shear models [38]. In series RC-U55, the interface between the conventional concrete and the strengthening material was placed at the level of the  Impact tests were carried out with a three-point bending scheme, with a span length of 1600 mm between supports, as plotted in Figure 1. A steel mass of 200 kg was freely dropped to the top side of the midspan of specimens from a height of 1.75 m (impact velocity = 5.9 m/s). The striking end of the projectile and the supports were cylindrical with 29 mm radius. At both supports, two steel yokes prevented uplift of the specimens. The impact load and reaction forces were measured with HBM piezoelectric force washers suited for highly dynamic solicitations with a sensitivity of 4.3 pC/N (CFW-700 and CFW-190, respectively). The vertical acceleration at the specimens' midspan was measured with a PCB piezoelectric accelerometer of 5 mV/g sensitivity, placed at mid-height of the cross-section (125 mm from bottom). The midspan acceleration is a representative measurement of the beam response and allows integration to achieve midspan displacements as explained in Section 2.3. The measurements were taken at a sampling rate of 40 kHz with an HBM expressoDAQ module. A comprehensive description of the testing machine is reported elsewhere [12,13]. An overview of the impact test configuration in the laboratory is given in Figure 2. It is noted that beam dimensions and test configuration are identical to those used in previous research of reinforced SFRC beams [12], except for the concrete cover, which was 40 mm, so that a comparative database including different cross-sectional schemes can be available for researchers. Impact tests were carried out with a three-point bending scheme, with a span length of 1600 mm between supports, as plotted in Figure 1. A steel mass of 200 kg was freely dropped to the top side of the midspan of specimens from a height of 1.75 m (impact velocity = 5.9 m/s). The striking end of the projectile and the supports were cylindrical with 29 mm radius. At both supports, two steel yokes prevented uplift of the specimens. The impact load and reaction forces were measured with HBM piezoelectric force washers suited for highly dynamic solicitations with a sensitivity of 4.3 pC/N (CFW-700 and CFW-190, respectively). The vertical acceleration at the specimens' midspan was measured with a PCB piezoelectric accelerometer of 5 mV/g sensitivity, placed at mid-height of the cross-section (125 mm from bottom). The midspan acceleration is a representative measurement of the beam response and allows integration to achieve midspan displacements as explained in Section 2.3. The measurements were taken at a sampling rate of 40 kHz with an HBM expressoDAQ module. A comprehensive description of the testing machine is reported elsewhere [12,13]. An overview of the impact test configuration in the laboratory is given in Figure 2. It is noted that beam dimensions and test configuration are identical to those used in previous research of reinforced SFRC beams [12], except for the concrete cover, which was 40 mm, so that a comparative database including different cross-sectional schemes can be available for researchers. Appl. Sci. 2020, 10, x FOR PEER REVIEW 5 of 18

Specimens' preparation and material characterization
The beam specimens were manufactured at the Laboratory of Structures at UPM. In order to ease the placement of the strengthening layer, the beams were manufactured upside down. The conventional concrete was supplied by a local contractor. For series RC-U35 and RC-U55, the thickness corresponding to the strengthening layer was left empty during concreting of the conventional concrete (i.e., no demolition works of the concrete cover were carried out). The upper surface of the RC layer was prepared with a slight treatment consisting of wire-brushing and waterpressure cleaning. Though a more exhaustive surface treatment (e.g., sand blasting) is typically recommended [26], it has been preferred to keep a less demanding interface closer to real conditions of structures strengthened on the bottom side. An additional method for application of bottom strengthening is by bonding a prefabricated laminate with epoxy adhesive [39,40]. Before pouring the strengthening material, the surface was washed again and kept moist. Fabrication and pouring of the strengthening material was carried out 15 months after manufacturing the RC.
The conventional concrete was designed for a characteristic strength of 30 MPa. The maximum aggregate size was 12 mm, the content of Portland cement I 42.5 R was 350 kg/m 3 and the water/cement ratio was 0.5. The average cylinder (150 mm × 300 mm) compressive strength at 28 days was 38.1 MPa. At testing age, the average cylinder (150 × 300 mm) compressive and indirect (Brazilian) tensile strength were 40.8 MPa and 2.3 MPa, respectively. The reinforcing steel bars had a characteristic yield and ultimate strength of 500 MPa and 575 MPa, respectively.
The strengthening material consisted of a dry mortar premix of compact-reinforced concrete (CRC kindly supplied by Hi-Con A/S, Hjallerup, Denmark), which was mixed with water and fibers following the supplier's instructions. A 2% volumetric content of short steel fibers (lf = 12.5 mm, df = 0.3 mm, fy = 2950 MPa) was used. The flow diameter in the slump flow test was larger than 550 mm, which allows classifying the mixture as self-compacting according to the French Association for Civil Engineering, AFGC [19]. The beams were tested 2 months after strengthening. The average cubic (150

Specimens' Preparation and Material Characterization
The beam specimens were manufactured at the Laboratory of Structures at UPM. In order to ease the placement of the strengthening layer, the beams were manufactured upside down. The conventional concrete was supplied by a local contractor. For series RC-U35 and RC-U55, the thickness corresponding to the strengthening layer was left empty during concreting of the conventional concrete (i.e., no demolition works of the concrete cover were carried out). The upper surface of the RC layer was prepared with a slight treatment consisting of wire-brushing and water-pressure cleaning. Though a more exhaustive surface treatment (e.g., sand blasting) is typically recommended [26], it has been preferred to keep a less demanding interface closer to real conditions of structures strengthened on the bottom side. An additional method for application of bottom strengthening is by bonding a prefabricated laminate with epoxy adhesive [39,40]. Before pouring the strengthening material, the surface was washed again and kept moist. Fabrication and pouring of the strengthening material was carried out 15 months after manufacturing the RC.
The conventional concrete was designed for a characteristic strength of 30 MPa. The maximum aggregate size was 12 mm, the content of Portland cement I 42.5 R was 350 kg/m 3 and the water/cement ratio was 0.5. The average cylinder (150 mm × 300 mm) compressive strength at 28 days was 38.1 MPa. At testing age, the average cylinder (150 × 300 mm) compressive and indirect (Brazilian) tensile strength were 40.8 MPa and 2.3 MPa, respectively. The reinforcing steel bars had a characteristic yield and ultimate strength of 500 MPa and 575 MPa, respectively.
The strengthening material consisted of a dry mortar premix of compact-reinforced concrete (CRC kindly supplied by Hi-Con A/S, Hjallerup, Denmark), which was mixed with water and fibers following the supplier's instructions. A 2% volumetric content of short steel fibers (l f = 12.5 mm, d f = 0.3 mm, f y = 2950 MPa) was used. The flow diameter in the slump flow test was larger than 550 mm, which allows classifying the mixture as self-compacting according to the French Association for Civil Engineering, AFGC [19]. The beams were tested 2 months after strengthening. The average cubic (150 mm) compressive strength of the strengthening material was 126.4 MPa and 135.4 MPa at 28 days and testing age, respectively. Flexural-tensile tests were carried out on un-notched and notched prismatic (100 × 100 × 500 mm, with span length between supports of 420 mm in all cases) samples at the age of beam tests. Un-notched samples were tested in a 4-point bending scheme. The results are represented in Figure 3a in terms of load-midspan deflection and equivalent bending stress-midspan deflection diagrams. A detail of the pre-peak stage is also plotted. Very thin cracks at a separation of 12-16 mm were observed in the central part of the specimens before localization of a macro-crack under peak strength, thereby showing the strain-hardening stage. The equivalent bending strength was 14.1 MPa on average. Notched samples were tested in a 3-point bending configuration with a 5 mm depth notch at midspan. Load-crack width results are plotted in Figure 3b. The average tensile properties of the strengthening material were estimated from inverse analysis of flexural tests according to López and colleagues [41] and the results are listed in Table 1. The free shrinkage strain was also measured in 2 cylinders of 150 × 300 mm, showing an average strain of −180 × 10 −6 from manufacturing to testing age. Hereafter, the strengthening material will be referred to as high-performance fiber-reinforced concrete (HPFRC), as it has high compressive strength and tensile strain-hardening.  Figure 3a in terms of load-midspan deflection and equivalent bending stress-midspan deflection diagrams. A detail of the pre-peak stage is also plotted. Very thin cracks at a separation of 12-16 mm were observed in the central part of the specimens before localization of a macro-crack under peak strength, thereby showing the strain-hardening stage. The equivalent bending strength was 14.1 MPa on average. Notched samples were tested in a 3-point bending configuration with a 5 mm depth notch at midspan. Load-crack width results are plotted in Figure 3b. The average tensile properties of the strengthening material were estimated from inverse analysis of flexural tests according to López and colleagues [41] and the results are listed in Table 1. The free shrinkage strain was also measured in 2 cylinders of 150 × 300 mm, showing an average strain of −180 × 10 -6 from manufacturing to testing age. Hereafter, the strengthening material will be referred to as highperformance fiber-reinforced concrete (HPFRC), as it has high compressive strength and tensile strain-hardening.

Test results
The impact solicitation of the specimens can be understood from the measurements taken by the load cells and the accelerometer. Exemplarily, the records of test RC-U0-IMP-1 are represented in Figure 4. Reactions measured at both supports are symmetrical in all tests, whereby the total reaction is represented. Both impact force and total reaction had a two-stage shape over time, first with a highfrequency peak followed by a second stage with a smaller frequency. The peak of the impact force was higher than that of the total reaction. A time gap of 2-3 ms was observed between the beginning of the impact event and the activation of the support reactions due to the time required for the wave propagation from the midspan to the supports. The difference between impact force and total reaction makes it clear that inertia forces developed, which can be verified by the acceleration measured at midspan (Figure 4). Though strains were not measured in the present experimental

Test Results
The impact solicitation of the specimens can be understood from the measurements taken by the load cells and the accelerometer. Exemplarily, the records of test RC-U0-IMP-1 are represented in Figure 4. Reactions measured at both supports are symmetrical in all tests, whereby the total reaction is represented. Both impact force and total reaction had a two-stage shape over time, first with a high-frequency peak followed by a second stage with a smaller frequency. The peak of the impact force was higher than that of the total reaction. A time gap of 2-3 ms was observed between the beginning of the impact event and the activation of the support reactions due to the time required for the wave propagation from the midspan to the supports. The difference between impact force and total reaction makes it clear that inertia forces developed, which can be verified by the acceleration measured at midspan (Figure 4). Though strains were not measured in the present experimental campaign, previous tests under similar conditions have shown that strain rates can be in the range of 1-30 s −1 [12].
rather than the impact load unless the effect of inertia forces is removed from the impact load, which in turn is equivalent to the total reaction [12,42,43]. Total reaction-midspan deflection diagrams are represented in Figure 6 for all tested series. The midspan deflection was calculated by double integration of the measured acceleration, in agreement with previous research [12,[43][44][45]. As observed in Figure 6, there is a shift of the origin of total reaction-deflection diagrams due to the fact that a deflection at the midspan already developed when the impact effect reached the supports (refer to Figure 4 as well).
From the crack patterns at failure ( Figure 5), it is clear that all beam specimens presented extensive shear cracking. The conventional RC beams (series RC-U0) failed by shear with various diagonal cracks, multiple fragmentation, and cover spalling, whose induced debris is one the most undesirable hazards for humans and goods under structures subjected to impacts. The crack patterns of the beams strengthened with a HPFRC layer showed the development of two types of shear cracks ( Figure 5): on the one hand, two diagonal cracks with an inclination of 45° or higher forming a shear plug at the midspan; on the other hand, two diagonal cracks in both shear spans with a shape similar to shear-flexure cracks typically developed in quasi-static tests of RC [12]. Yi et al. [8] classified the former crack categories as type I (shear plug) and type II (diagonal cracks at shear spans), as schematized in Figure 7b. The most interesting difference between the failure pattern of the two strengthened series can be observed at the HPFRC layer: in series RC-U55, debonding occurred at the interface between HPFRC and RC, and the HPFRC between the two type II cracks was split out; meanwhile, the HPFRC in series RC-U35 was able to stay in contact with the RC and to develop a resisting mechanism similar to a stress ribbon able to retain the damaged RC part. This last mechanism can be very interesting for real structures, as it has the potential to protect humans and goods below the structure.  The shape of the load-deflection diagrams of Figure 6 can be useful to understand the different behavior of tested series. Such diagrams can be divided into two stages: first, a high-peak triangular stage and a secondary post-peak stage. While the amplitude of the first peak provides the load carrying capacity, the second stage gives information on the post-peak deformation and energy absorption capacity after first stage, which is governed by the HPFRC strengthening layer. Even though the average values of the first peak were rather similar for the three series (except for some scatter within series RC-U0; refer to Section 3, where the differences are analyzed), significant differences can be observed in the second stage. For RC-U0 series, the second stage was rather small,  Pictures of deformed shapes and failure patterns of all tested specimens are given in Figure 5, while the main test results are summarized in Table 2. The strength (R max ) has been taken as the peak of Appl. Sci. 2020, 10, 3154 8 of 18 the total reaction at both supports. It has to be noted that such a value differs from the maximum impact load due to the effect of inertia forces. The total reaction has to be used as a strength indicator rather than the impact load unless the effect of inertia forces is removed from the impact load, which in turn is equivalent to the total reaction [12,42,43]. Total reaction-midspan deflection diagrams are represented in Figure 6 for all tested series. The midspan deflection was calculated by double integration of the measured acceleration, in agreement with previous research [12,[43][44][45]. As observed in Figure 6, there is a shift of the origin of total reaction-deflection diagrams due to the fact that a deflection at the midspan already developed when the impact effect reached the supports (refer to Figure 4 as well).

Phenomenology of resisting mechanisms
The brittleness and high tendency to develop shear cracking and fragmentation of RC b under impact loading were confirmed by the results of RC-U0 series. Test results indicate th impact response of strengthened specimens mobilized more complex resisting mechanisms those reported for quasi-static loading scenarios. Previous researchers [46] explained tha contribution of the HPFRC layer to the shear capacity under quasi-static loading can be due, o one hand, to the increase of the shear strength of the RC layer and, on the other hand, to addi shear resisting mechanism. The former can be explained by the capacity of the HPFRC layer to co the width of the critical shear crack, thereby increasing aggregate interlocking, as well as enla the compression zone depth, which in turn increases its capacity to carry shear stresses additional contribution was assigned by Noshiravani and Brühwiler [29] to the capacity o strengthening layer for bending in double curvature along the interlayer debonding zone, as dep in Figure 7a. It has to be noted that critical shear cracks under quasi-static loading can be assim to previously defined type II cracks. From the crack patterns at failure (Figure 5), it is clear that all beam specimens presented extensive shear cracking. The conventional RC beams (series RC-U0) failed by shear with various diagonal cracks, multiple fragmentation, and cover spalling, whose induced debris is one the most undesirable hazards for humans and goods under structures subjected to impacts. The crack patterns of the beams strengthened with a HPFRC layer showed the development of two types of shear cracks ( Figure 5): on the one hand, two diagonal cracks with an inclination of 45 • or higher forming a shear plug at the midspan; on the other hand, two diagonal cracks in both shear spans with a shape similar to shear-flexure cracks typically developed in quasi-static tests of RC [12]. Yi et al. [8] classified the former crack categories as type I (shear plug) and type II (diagonal cracks at shear spans), as schematized in Figure 7b. The most interesting difference between the failure pattern of the two strengthened series can be observed at the HPFRC layer: in series RC-U55, debonding occurred at the interface between HPFRC and RC, and the HPFRC between the two type II cracks was split out; meanwhile, the HPFRC in series RC-U35 was able to stay in contact with the RC and to develop a resisting mechanism similar to a stress ribbon able to retain the damaged RC part. This last mechanism can be very interesting for real structures, as it has the potential to protect humans and goods below the structure.
shear resisting mechanism. The former can be explained by the capacity of the HPFRC layer to control the width of the critical shear crack, thereby increasing aggregate interlocking, as well as enlarging the compression zone depth, which in turn increases its capacity to carry shear stresses. The additional contribution was assigned by Noshiravani and Brühwiler [29] to the capacity of the strengthening layer for bending in double curvature along the interlayer debonding zone, as depicted in Figure 7a. It has to be noted that critical shear cracks under quasi-static loading can be assimilated to previously defined type II cracks. The present experimental results show that two types of shear cracks (I and II as defined previously, see Figure 7b) developed. Type I cracks opened immediately upon impact as a local effect, while type II cracks were due to flexure-shear interaction at the webs produced by inertia forces, as similarly reported by previous researchers [6,8,47]. From the video records of completed tests, the sequence of crack development and activation of internal forces are as follows: The impact load is firstly resisted by inertia forces in the midspan, before the activation of reaction forces. At this stage, the midspan zone is highly stressed and type I cracks develop. In the second stage, when reaction forces are activated and the impact force is cancelled, inertia forces are fully equilibrated by the reaction forces. Shear forces and bending moments in the second stage are responsible for type II cracks. Due to the similar crack patterns of all tested specimens, it seems that the formation of type I and II cracks was not affected by the presence of a HPFRC layer. Rather, the contribution of the stress ribbon mechanism of the HPFRC layer was activated upon the formation of such cracks and responsible for the retention of the damaged RC part of the beams and control of fragmentation. The crack pattern of RC-U35 series indicates that the full tensile capacity of the HPFRC was utilized by the stress ribbon mechanism, as the HPFRC layer shows visible macrocracks at an average distance of 180 mm to each other ( Figure 5). In RC-U35 series, the stress ribbon mechanism was able to bridge the damaged zone between the two type II cracks. The full utilization of the tensile capacity of the HPFRC confirms conclusions made in previous research (see [14,[30][31][32][33][34]) on the positive strain-rate influence on the tensile properties of HPFRC in the impact domain. Such strain rate has been in the range of 1-30 s -1 for specimens tested under the same conditions as those of the present paper [12]. These results also agree with the conclusions of Ulzurrun and Zanuy [13] for SFRC regarding the rate The shape of the load-deflection diagrams of Figure 6 can be useful to understand the different behavior of tested series. Such diagrams can be divided into two stages: first, a high-peak triangular stage and a secondary post-peak stage. While the amplitude of the first peak provides the load carrying capacity, the second stage gives information on the post-peak deformation and energy absorption capacity after first stage, which is governed by the HPFRC strengthening layer. Even though the average values of the first peak were rather similar for the three series (except for some scatter within series RC-U0; refer to Section 3, where the differences are analyzed), significant differences can be observed in the second stage. For RC-U0 series, the second stage was rather small, which explains the almost negligible post-peak capacity, in agreement with the observed failure mode. In contrast, RC-U35 and RC-U55 series showed a longer second stage, but a remarkable difference can be observed at the end of this second stage: in the two tests of RC-U35 series, a significant part of the deflection was recovered as the load released, which reminds the unloading stage of a specimen under quasi-static loading. The descending branch can be understood as a sign of the integrity of the HPFRC layer which explains the resisting mechanism achieved by the HPFRC layer in RC-U35 series. In contrast, the former descending branch cannot be observed in the tests of RC-U55 series, where the HPFRC was debonded by the impact.

Phenomenology of Resisting Mechanisms
The brittleness and high tendency to develop shear cracking and fragmentation of RC beams under impact loading were confirmed by the results of RC-U0 series. Test results indicate that the impact response of strengthened specimens mobilized more complex resisting mechanisms than those reported for quasi-static loading scenarios. Previous researchers [46] explained that the contribution of the HPFRC layer to the shear capacity under quasi-static loading can be due, on the one hand, to the increase of the shear strength of the RC layer and, on the other hand, to additional shear resisting mechanism. The former can be explained by the capacity of the HPFRC layer to control the width of the critical shear crack, thereby increasing aggregate interlocking, as well as enlarging the compression zone depth, which in turn increases its capacity to carry shear stresses. The additional contribution was assigned by Noshiravani and Brühwiler [29] to the capacity of the strengthening layer for bending in double curvature along the interlayer debonding zone, as depicted in Figure 7a. It has to be noted that critical shear cracks under quasi-static loading can be assimilated to previously defined type II cracks.
The present experimental results show that two types of shear cracks (I and II as defined previously, see Figure 7b) developed. Type I cracks opened immediately upon impact as a local effect, while type II cracks were due to flexure-shear interaction at the webs produced by inertia forces, as similarly reported by previous researchers [6,8,47]. From the video records of completed tests, the sequence of crack development and activation of internal forces are as follows: The impact load is firstly resisted by inertia forces in the midspan, before the activation of reaction forces. At this stage, the midspan zone is highly stressed and type I cracks develop. In the second stage, when reaction forces are activated and the impact force is cancelled, inertia forces are fully equilibrated by the reaction forces. Shear forces and bending moments in the second stage are responsible for type II cracks. Due to the similar crack patterns of all tested specimens, it seems that the formation of type I and II cracks was not affected by the presence of a HPFRC layer. Rather, the contribution of the stress ribbon mechanism of the HPFRC layer was activated upon the formation of such cracks and responsible for the retention of the damaged RC part of the beams and control of fragmentation. The crack pattern of RC-U35 series indicates that the full tensile capacity of the HPFRC was utilized by the stress ribbon mechanism, as the HPFRC layer shows visible macrocracks at an average distance of 180 mm to each other ( Figure 5). In RC-U35 series, the stress ribbon mechanism was able to bridge the damaged zone between the two type II cracks. The full utilization of the tensile capacity of the HPFRC confirms conclusions made in previous research (see [14,[30][31][32][33][34]) on the positive strain-rate influence on the tensile properties of HPFRC in the impact domain. Such strain rate has been in the range of 1-30 s −1 for specimens tested under the same conditions as those of the present paper [12]. These results also agree with the conclusions of Ulzurrun and Zanuy [13] for SFRC regarding the rate sensitivity of the pull-out mechanisms mobilized by straight steel fibers, showing better dynamic enhancement and lower fiber ruptures than hook-end fibers.
In RC-U55 series, the full development of the stress ribbon by the HPFRC was limited by the interface debonding, which according to the videos recorded during testing, took place very soon upon impact. Debonding can be explained by the wave sequence sketched in Figure 7c: upon impact, a compressive wave propagates firstly from the top to the bottom of the beam at midspan. Thereafter, a tensile wave is reflected and, in case a weak layer is crossed by the reflected wave, debonding can occur. According to test results, the interface between old concrete and HPFRC was able to resist the reflected wave in RC-U35 series but not in RC-U55 series. It is apparent that, for RC-U55 series, the presence of longitudinal reinforcement has weakened the interface which was unable to resist the reflected wave. This result can be explained by the reduced contact area between old concrete and HPFRC due to the presence of the steel bars; according to the dimensions of the beam (width of 125 mm) and the diameter of the steel bars (2 × 16 mm), the contact area at the interface of RC-U55 series was 26% smaller than that of RC-U35 series.

Load-Deflection Diagrams and Energy Absorption
The two stages of tests described in Section 3.1, related to type I and II cracks, respectively, can be associated to the two stages of the total reaction-midspan deflection diagrams of Figure 6 (hereafter, referred to as load-deflection diagrams). The main characteristics of load-deflection diagrams are listed in Table 2. In addition, the maximum and residual deflections, and the energies absorbed in the two stages of tests are plotted in Figure 8a,b. The first stage of load-deflection diagrams was similar for all tested series, which is explained by the local response upon impact and the formation of type I cracks in all tests. This agrees with the similar average peak load (R max ) and energy absorbed in the first stage (E 1 , calculated as the area below the load-deflection diagram in the first stage) shown in all series. The second stage of load-deflection diagrams is associated to the global response of the beams, the formation of type II cracks, and the activation of the stress ribbon mechanism in strengthened beams. From Figure 8b, the energy E 2 (calculated as the area below the load-deflection diagram in the second stage) absorbed by strengthened specimens was significantly larger than unstrengthened beams. On average, the increase of absorbed energy E 2 of RC-U35 and RC-U55 was 2.1 and 2.7 times the value of RC-U0, respectively. The capacity of the HPFRC layer to retain the damage of the specimens can also be demonstrated by the maximum midspan deflection (Figure 8a), which on average was 36% (RC-U35) and 39% (RC-U55) smaller than that of RC-U0 series, respectively.

Weight loss and debris
An additional indicator of impact damage is the loss of weight of the specimens due to split-out fragments. The analysis of weight loss has relevance to understand the capacity of the strengthening layer to protect goods or humans under the impacted member against induced debris. The weight loss of tested specimens is represented in Figure 8c as a function of the thickness of the HPFRC layer. It is apparent that that RC-U35 series was able to significantly reduce the generation of debris to 54% with respect to RC-U0 series. In contrast, the interfacial debonding shown by RC-U55 series was very harmful, as the generation of fragments was 28% higher than RC-U0 series due to spalling of the strengthening layer. Accordingly, it is clearly not recommended to design the strengthening layer with the interface at the same level as the longitudinal reinforcement.

Comparison with other configurations
In order to evaluate the capacity of the HPFRC-RC configuration to resist impact loads, a wider study was carried out by including results of previous research. Eight additional tests with identical geometry, reinforcement arrangements (concrete cover 10 mm smaller), boundary conditions, and impact load were considered from two other studies [12,22]. Among all the newly considered specimens, two were RC beams identical to series RC-U0 of this paper (only with a slightly different concrete strength, fc = 48.5 MPa), four specimens consisted of reinforced SFRC beams (R-SFRC), and two specimens were reinforced HPFRC beams (R-HPFRC). In the R-SFRC beams, SFRC was used instead of conventional concrete. The SFRC consisted of a matrix with the same concrete as the companion RC beams and a certain amount of short straight steel fibers (lf = 10 mm, df = 0.16 mm, fy = 3000 MPa): two specimens contained 0.5% fibers by volume and two specimens 1.0%. In the R-HPFRC beams, the conventional concrete was replaced with HPFRC with 1.6% fiber amount (lf =10 mm, df = 0.16 mm, fy = 3000 MPa).
The RC beams presented the same impact behavior as in the present paper. The R-SFRC beams developed flexural failure with opening of a shear plug (type I cracks) in the midspan. In addition to the shear plug, type II diagonal cracks opened in the webs but the bridging capacity of the SFRC was able to avoid their propagation. The residual width of diagonal cracks and shear plugs was larger in the specimens with 0.5% fiber content than with 1.0%. The R-HPFRC developed flexural failure mode with a major bending crack at midspan and no shear cracking at all.
The results of all considered tests are represented in Figure 9 in terms of strength (maximum value of total reaction) and maximum midspan displacement produced by the impact. As it can be noted, the highest strength was obtained by the R-HPFRC beams followed by the R-SFRC beams with The development of interface debonding in strengthened specimens can be also analyzed with the shape of the last descending part of load-deflection diagrams ( Figure 6). The integrity of the HPFRC layer in RC-U35 series corresponds with a descending part where the midspan deflection is partially recovered (i.e., the last descending branch can be seen as an unloading stage rather than a full collapse). In contrast, the last descending part of the load-deflection curves of RC-U55 series shows no deflection recovery, but sudden decrease of the load carried, which explains the spalling of the HPFRC layer. The existence of interface debonding can also be studied from the residual midspan deflection (v res in Table 2, also represented in Figure 8a); it is clear that the specimens of series RC-U35 are only where the residual deflection after testing was significantly smaller than the maximum deflection.

Weight Loss and Debris
An additional indicator of impact damage is the loss of weight of the specimens due to split-out fragments. The analysis of weight loss has relevance to understand the capacity of the strengthening layer to protect goods or humans under the impacted member against induced debris. The weight loss of tested specimens is represented in Figure 8c as a function of the thickness of the HPFRC layer. It is apparent that that RC-U35 series was able to significantly reduce the generation of debris to 54% with respect to RC-U0 series. In contrast, the interfacial debonding shown by RC-U55 series was very harmful, as the generation of fragments was 28% higher than RC-U0 series due to spalling of the strengthening layer. Accordingly, it is clearly not recommended to design the strengthening layer with the interface at the same level as the longitudinal reinforcement.

Comparison with Other Configurations
In order to evaluate the capacity of the HPFRC-RC configuration to resist impact loads, a wider study was carried out by including results of previous research. Eight additional tests with identical geometry, reinforcement arrangements (concrete cover 10 mm smaller), boundary conditions, and impact load were considered from two other studies [12,22]. Among all the newly considered specimens, two were RC beams identical to series RC-U0 of this paper (only with a slightly different concrete strength, f c = 48.5 MPa), four specimens consisted of reinforced SFRC beams (R-SFRC), and two specimens were reinforced HPFRC beams (R-HPFRC). In the R-SFRC beams, SFRC was used instead of conventional concrete. The SFRC consisted of a matrix with the same concrete as the companion RC beams and a certain amount of short straight steel fibers (l f = 10 mm, d f = 0.16 mm, f y = 3000 MPa): two specimens contained 0.5% fibers by volume and two specimens 1.0%. In the R-HPFRC beams, the conventional concrete was replaced with HPFRC with 1.6% fiber amount (l f = 10 mm, d f = 0.16 mm, f y = 3000 MPa).
The RC beams presented the same impact behavior as in the present paper. The R-SFRC beams developed flexural failure with opening of a shear plug (type I cracks) in the midspan. In addition to the shear plug, type II diagonal cracks opened in the webs but the bridging capacity of the SFRC was able to avoid their propagation. The residual width of diagonal cracks and shear plugs was larger in the specimens with 0.5% fiber content than with 1.0%. The R-HPFRC developed flexural failure mode with a major bending crack at midspan and no shear cracking at all.
The results of all considered tests are represented in Figure 9 in terms of strength (maximum value of total reaction) and maximum midspan displacement produced by the impact. As it can be noted, the highest strength was obtained by the R-HPFRC beams followed by the R-SFRC beams with 1.0% fiber content and the R-SFRC with 0.5% fiber. The deformation capacity under impact can be evaluated by means of maximum deflection obtained at the midspan. From Figure 9, the average midspan deflection of HPFRC-strengthened RC specimens was 37% smaller than that of RC beams failing by brittle shear and severe fragmentation, while that of R-HPFRC specimens failing by flexure was 75% smaller than that of RC. The comparison of maximum deflection indicates that the HPFRC-RC beams showed an intermediate deformation capacity between RC and R-HPFRC. The smaller strength and higher deflection of RC-U55 with respect to RC-U35 was due to the full development of the additional resisting mechanism as stress ribbon by RC-U35 series.
Appl. Sci. 2020, 10, x FOR PEER REVIEW 14 of 18 1.0% fiber content and the R-SFRC with 0.5% fiber. The deformation capacity under impact can be evaluated by means of maximum deflection obtained at the midspan. From Figure 9, the average midspan deflection of HPFRC-strengthened RC specimens was 37% smaller than that of RC beams failing by brittle shear and severe fragmentation, while that of R-HPFRC specimens failing by flexure was 75% smaller than that of RC. The comparison of maximum deflection indicates that the HPFRC-RC beams showed an intermediate deformation capacity between RC and R-HPFRC. The smaller strength and higher deflection of RC-U55 with respect to RC-U35 was due to the full development of the additional resisting mechanism as stress ribbon by RC-U35 series. Figure 9. Comparison of results of the tests of the present paper and those from [12,22].

Conclusions
The results presented in this paper are intended to understand the potential of shear-critical HPFRC-RC members subjected to impacts. According to the results, the following conclusions can be drawn:

Conclusions
The results presented in this paper are intended to understand the potential of shear-critical HPFRC-RC members subjected to impacts. According to the results, the following conclusions can be drawn: RC beams presented brittle shear failure with significant fragmentation. Strengthening with HPFRC reduced fragmentation but types I and II cracks also formed. The beams strengthened with HPFRC layer presented an additional resisting mechanism with the HPFRC working similarly to a stress ribbon. This additional contribution was activated in the second stage of tests characterized by global behavior. The stress ribbon mechanism provided an energy absorption in the second stage (E 2 ) 2.1 and 2.7 times that of the unstrengthened beams, in the case of specimens with 35 mm and 55 mm strengthening layers, respectively. II. Interface debonding can occur when the tensile reflected wave induced by the impact crosses the interface between concretes, as observed in the last stage of impact tests of RC-U55 series. In such series, the interface was placed at the level of the longitudinal reinforcement, which reduced the contact surface between concretes and weakened the impact behavior. Accordingly, it is recommended to avoid such design configuration. III. In comparison with results from previous research of R-HPFRC and R-SFRC beams failing by flexure under impact, HPFRC-RC presents an intermediate behavior between them and conventional RC in terms of strength and deformation capacity. IV. Strengthening with HPFRC tensile layer presents interesting potential for existing RC structures as a new resisting mechanism provided by the HPFRC working similarly as a stress ribbon can reduce impact debris, thereby protecting humans and goods below the structure. Future research is still needed in order to quantify the contribution of the new resisting mechanism. V.
The research presented in this paper has shown that a structural solution for existing shear-critical RC members susceptible to suffer impacts can be based on the application of a thin HPFRC layer on the tensile side. In order to achieve the stress ribbon mechanism found for this structural solution, structural designers are recommended to consider the following guides: a. A fiber amount of at least 2% for the HPFRC to achieve strain hardening. b.
The interface between the old concrete and the HPFRC should not be placed at the level of the longitudinal reinforcement. c.
A surface preparation of the old concrete with hydro-jetting or sand-blasting before application of the HPFRC is recommended. In case the tensile side is at the bottom, a prefabricated HPFRC laminate can be alternatively bonded with epoxy adhesive.