Experimental Campaign of a Low-Cost and Replaceable System for Passive Energy Dissipation in Precast Concrete Structures

: This research develops a new low-cost energy dissipation system, capable of being implemented in residential structures in developing countries with high seismic activity, in which the current solutions are not economically viable. These residential structures are entirely made of precast concrete elements (foundations, walls, and slabs). A solution is developed that consists of a new connection between a precast foundation and a structural wall, which is capable of dissipating almost all the seismic energy, and therfore protecting the rest of the building from structural damage. To validate the solution, a testing campaign is carried out, including a ﬁrst set of “pushover” tests on isolated structural walls, a second set of “pushover” tests on structural frames, and a ﬁnal set of seismic tests on a real-scale three-storey building. For the ﬁrst and second set of tests, ductility is analyzed in accordance with ACI 374.2R-13, while for the third one, the dynamic response to a reference earthquake is evaluated. The results reveal that the solution developed shows great ductility and no relevant damage is observed in the rest of the building, except in the low-cost energy dissipation system. Once an earthquake has ﬁnished, a precast building implemented with this low-cost energy dissipation system is capable of showing a structural performance level of “immediate occupancy” according to ACI 374.2R-13.


Introduction
Precast concrete construction represents a very important percentage of all civil works in the world, given its enormous advantages from a constructive point of view. The reduction of the manufacturing time, the improvement in the quality of the work (due to the improvement of the working conditions), and the reduction of uncertainties related to the geometric and structural deviation of the solution with respect to the project are some of the advantages of this constructive procedure.
Precast concrete is especially efficient in residential structures, which are generally made up of a few groups of different structural elements (columns, slabs, footings, walls, etc.), formed by many identical units. In particular, precast concrete is especially interesting in developing countries, where it can be difficult to find enough skilled labour to perform in situ constructions.
A significant number of these developing countries are in areas of high seismic activity. This is an inconvenience for precast solutions, since it is penalized by international regulations by granting lower reduction coefficients (R) for energy dissipation. This is due to the lower ductility of the connections Secondly, pushover tests were carried out on structural frames, composed of two precast structural walls placed over two precast footings and connected with a precast slab. The aim of this second phase is to evaluate the ductility of the system, including the dissipation capacity of both the connection footing wall and the connection wall slab, in accordance with the aforementioned regulation.
Thirdly, seismic tests using a shake table were carried out on a real-scale three-storey precast concrete building, consisting of two precast structural walls placed over two precast footings, two intermediate precast slabs, and a lightweight roof. In this case, the aim of this third phase is to characterize the dynamic response of the entire structure to a reference earthquake and the energy dissipation capacity of the building.
A customized unidirectional shake table was designed and manufactured specifically for the third test phase. This testing facility was capable of appling a horizontal acceleration up to 1·g to the structural elements with a height up to 6 m, a weight up to 40 tons, and a frequency up to 8 Hz.
All the precast concrete elements, as well as the low-cost energy dissipation system were invented, developed, and designed by the Spanish company ICONKRETE 2012, S.L., and therefore this company is the owner of this structural solution and the testing results. The precast elements were manufactured by the company ZENET in its factory in Escalonilla (Toledo, Spain). The test was carried out in the Laboratory of Large Structures of the University of Burgos (Burgos, Spain).
The structure of this paper is as follows: In Section 2, the experimental program is presented; in Section 3, the results of the tests are described and discussed; and finally, in Section 4, the conclusions are shown.

Experimental Program
In this section, the three testing sets of specimens are described, i.e., the isolated structural walls, the structural frames, and the real-scale three-storey precast concrete building, as well as the low-cost energy dissipation system and the shake table. Additionally, the testing procedure is described.

Isolated Structural Wall
As explained before, the isolated structural walls are composed of a precast structural wall and a precast footing, connected through the low-cost energy dissipation system.
The structural wall is a conventional reinforced concrete precast element 3.0 m high, 2.0 m wide, and 16 cm thick. The wall is placed on a reinforced concrete precast footing, 1.1 m wide, 2.0 m deep, and 0.6 m high. In both cases, the concrete quality is C30/37, according to Eurocode 2 [35]. The compressive strength was obtained following the method described in standard EN 12390-3 [36]. The footing includes a longitudinal pocket 36 cm wide and 36 cm high, where the structural wall is placed. Underneath the wall, a 2 cm neoprene band is placed. The structural wall is rigidly connected to the footing through the low-cost energy dissipation system, which is described later (Figure 1).

Structural Frame
The structural frame is composed of two isolated structural elements, as described above, connected with a structural concrete slab. The inner distance between the walls is 3.84 m. The slab is composed of a self-supporting precast prestressed concrete slab with a thickness of 8 cm and an upper layer of 14 cm of cast in situ reinforced concrete. The total thickness of the concrete slab is 22 cm. In The structural wall is reinforced on both faces with steel rebars 8 mm in diameter spaced 150 mm in both longitudinal and vertical directions.

Structural Frame
The structural frame is composed of two isolated structural elements, as described above, connected with a structural concrete slab. The inner distance between the walls is 3.84 m. The slab is composed of a self-supporting precast prestressed concrete slab with a thickness of 8 cm and an upper layer of 14 cm of cast in situ reinforced concrete. The total thickness of the concrete slab is 22 cm. In all cases, the concrete quality is C30/37, according to Eurocode 2 [35].
A "flexible" connection between the walls and the slab was used, which is described next. The self-supporting precast prestressed concrete slab rests on a steel corner profile anchored to the walls using a set of mechanical anchorages. Additionally, a row of conventional steel rebars sew the joint between the wall and the cast-in situ slab. The rebars are L-shaped with a length of 800 and 200 mm, respectively. The long leg of the rebar is placed horizontally inside the cast-in situ reinforced concrete slab, while the short leg is placed vertically inside the wall. The diameter of the rebars are 20 mm, with a spacing of 50 cm. From the structural point of view, this connection is very effective under vertical loads, such as self-weight, dead loads, and vertical live loads. Under horizontal loads, similar to the ones caused by the earthquake, this connection is able to withstand negative bending moments, but not positive bending moments, resulting in a semi-rigid joint. (Figure 2).

Structural Frame
The structural frame is composed of two isolated structural elements, as described above, connected with a structural concrete slab. The inner distance between the walls is 3.84 m. The slab is composed of a self-supporting precast prestressed concrete slab with a thickness of 8 cm and an upper layer of 14 cm of cast in situ reinforced concrete. The total thickness of the concrete slab is 22 cm. In all cases, the concrete quality is C30/37, according to Eurocode 2 [35].
A "flexible" connection between the walls and the slab was used, which is described next. The self-supporting precast prestressed concrete slab rests on a steel corner profile anchored to the walls using a set of mechanical anchorages. Additionally, a row of conventional steel rebars sew the joint between the wall and the cast-in situ slab. The rebars are L-shaped with a length of 800 and 200 mm, respectively. The long leg of the rebar is placed horizontally inside the cast-in situ reinforced concrete slab, while the short leg is placed vertically inside the wall. The diameter of the rebars are 20 mm, with a spacing of 50 cm. From the structural point of view, this connection is very effective under vertical loads, such as self-weight, dead loads, and vertical live loads. Under horizontal loads, similar to the ones caused by the earthquake, this connection is able to withstand negative bending moments, but not positive bending moments, resulting in a semi-rigid joint. (Figure 2).

Real-Scale Three-Storey Precast Concrete Building
Finally, the real-scale three-storey precast concrete building is composed of two precast concrete walls, two precast footings, two concrete slabs, and a flexible steel roof. The inner distance between the two walls is 2.25 m. In all cases, the concrete quality is C30/37, according to Eurocode 2 [35].
The structural walls are conventional reinforced concrete precast elements 5.62 m high, 2.0 m wide, and 16 cm thick. They are placed on the same reinforced concrete precast footings as described above.
Each of the two concrete slabs is composed of a self-supporting precast prestressed concrete slab with a thickness of 8 cm and an upper layer of 14 cm of cast-in situ reinforced concrete. The connection between the slab and the wall is the same as the one described in the previous subsection. The slabs are located at 2.45 m and 4.20 m high. Finally, the roof is located at the top of the walls. It is a lightweight roof formed by an aluminium sheet which is bolted to two I-beams. (Figure 3).
Each of the two concrete slabs is composed of a self-supporting precast prestressed concrete slab with a thickness of 8 cm and an upper layer of 14 cm of cast-in situ reinforced concrete. The connection between the slab and the wall is the same as the one described in the previous subsection. The slabs are located at 2.45 m and 4.20 m high. Finally, the roof is located at the top of the walls. It is a lightweight roof formed by an aluminium sheet which is bolted to two I-beams. (Figure 3).

Low-Cost Energy Dissipation System
The low-cost energy dissipation system is a device placed on the connection between the footing and the wall. As previously explained, the footing includes a longitudinal pocket 36 cm wide and 36 cm high, where the structural wall is placed. It consists of a set of threaded bars placed orthogonally to the wall, in such a way that they cross both the footing and the wall. In this case, the system consists of a total of 10 threaded bars, 20 mm diameter, and placed in two rows, spaced 300 mm in both longitudinal and vertical directions. The property class of the steel for the threaded bars is 3.6, according to ISO 898-1 [37].

Low-Cost Energy Dissipation System
The low-cost energy dissipation system is a device placed on the connection between the footing and the wall. As previously explained, the footing includes a longitudinal pocket 36 cm wide and 36 cm high, where the structural wall is placed. It consists of a set of threaded bars placed orthogonally to the wall, in such a way that they cross both the footing and the wall. In this case, the system consists of a total of 10 threaded bars, 20 mm diameter, and placed in two rows, spaced 300 mm in both longitudinal and vertical directions. The property class of the steel for the threaded bars is 3.6, according to ISO 898-1 [37].
Since the wall has a thickness of 16 cm, there are two 10 cm gaps between the wall and the footing, one at each side of the wall ( Figure 4). Additionally, nuts located at both sides of the wall and the footing are required to fix the wall in its proper position.
When an earthquake occurs, it is expected that most of the energy is consumed in the plastic deformation of the bars, preventing the rest of the structure from damage. The number, position, distribution, and diameter of the bars, as well as the steel quality must be specifically designed to each particular structure, depending on the dimensions of the structure and the location of the building.
One additional advantage of this solution is that bars are easily replaceable once an earthquake has occurred. deformation of the bars, preventing the rest of the structure from damage. The number, position, distribution, and diameter of the bars, as well as the steel quality must be specifically designed to each particular structure, depending on the dimensions of the structure and the location of the building.
One additional advantage of this solution is that bars are easily replaceable once an earthquake has occurred.

Shake Table
A unidirectional shake table was specifically designed and manufactured for this research, in order to fulfill the testing requirements of the real-scale three-storey precast concrete building.
The shake table is composed of a cast steel slab 4.14 m long, 2.87 m wide, and 0.65 m high. It rests on six circular elastomeric bearing pads, reinforced with steel plates. The dimensions of the bearings are 350 mm diameter and 137 mm high.
The seismic loads are applied using a tension-compression MTS 201.70F dynamic actuator (MTS, Eden Prairie, MN, USA), with a capacity of ±1000 kN. The actuator has a load cell MTS 661.31F-01 (MTS, Eden Prairie, MN, USA), with a range of ±1000 kN and an error of below 1% of the range. During the testing, the actuator provided a longitudinal displacement. In order to prevent undesirable transversal movements, four stoppers were placed in both lateral sides of the steel slab ( Figures 5 and 6).

Shake Table
A unidirectional shake table was specifically designed and manufactured for this research, in order to fulfill the testing requirements of the real-scale three-storey precast concrete building.
The shake table is composed of a cast steel slab 4.14 m long, 2.87 m wide, and 0.65 m high. It rests on six circular elastomeric bearing pads, reinforced with steel plates. The dimensions of the bearings are 350 mm diameter and 137 mm high.
The seismic loads are applied using a tension-compression MTS 201.70F dynamic actuator (MTS, Eden Prairie, MN, USA), with a capacity of ±1000 kN. The actuator has a load cell MTS 661.31F-01 (MTS, Eden Prairie, MN, USA), with a range of ±1000 kN and an error of below 1% of the range. During the testing, the actuator provided a longitudinal displacement. In order to prevent undesirable transversal movements, four stoppers were placed in both lateral sides of the steel slab ( Figures 5 and 6).
Appl. Sci. 2020, 10, x FOR PEER REVIEW 6 of 23 Since the wall has a thickness of 16 cm, there are two 10 cm gaps between the wall and the footing, one at each side of the wall ( Figure 4). Additionally, nuts located at both sides of the wall and the footing are required to fix the wall in its proper position.
When an earthquake occurs, it is expected that most of the energy is consumed in the plastic deformation of the bars, preventing the rest of the structure from damage. The number, position, distribution, and diameter of the bars, as well as the steel quality must be specifically designed to each particular structure, depending on the dimensions of the structure and the location of the building.
One additional advantage of this solution is that bars are easily replaceable once an earthquake has occurred.

Shake Table
A unidirectional shake table was specifically designed and manufactured for this research, in order to fulfill the testing requirements of the real-scale three-storey precast concrete building.
The shake table is composed of a cast steel slab 4.14 m long, 2.87 m wide, and 0.65 m high. It rests on six circular elastomeric bearing pads, reinforced with steel plates. The dimensions of the bearings are 350 mm diameter and 137 mm high.
The seismic loads are applied using a tension-compression MTS 201.70F dynamic actuator (MTS, Eden Prairie, MN, USA), with a capacity of ±1000 kN. The actuator has a load cell MTS 661.31F-01 (MTS, Eden Prairie, MN, USA), with a range of ±1000 kN and an error of below 1% of the range. During the testing, the actuator provided a longitudinal displacement. In order to prevent undesirable transversal movements, four stoppers were placed in both lateral sides of the steel slab ( Figures 5 and 6).  The pushover tests carried out on the isolated structural walls and the structural frames were developed according to the American standard ACI374.2R-13 [34]. According to this standard, four levels of structural performance under seismic events are defined, which are "operational",  The pushover tests carried out on the isolated structural walls and the structural frames were developed according to the American standard ACI374.2R-13 [34]. According to this standard, four levels of structural performance under seismic events are defined, which are "operational", "immediate occupancy", "life safety", and "collapse prevention", moving from least to greatest lateral drift ratio (Figure 7).

Pushover Tests on Isolated Structural Walls and Structural Frames
The pushover tests carried out on the isolated structural walls and the structural frames were developed according to the American standard ACI374.2R-13 [34]. According to this standard, four levels of structural performance under seismic events are defined, which are "operational", "immediate occupancy", "life safety", and "collapse prevention", moving from least to greatest lateral drift ratio ( Figure 7). According to the goal of this research, the "immediate occupancy" level is considered for design purposes, which implies that the building can be used once the seismic event has passed. At this level, the stiffness requirements are high, the behavior of the structure must be within the elastic-linear range and, consequently, the seismic loads developed are important. The standard used does not consider other criteria apart from those concerning structural damage. That is the reason why nonstructural issues, such as furniture overturning or partition damage, have not been considered in this research. However, it is worth noting that the "immediate occupancy" level implies a very low risk of life-threatening injury as a result of structural damage. According to the goal of this research, the "immediate occupancy" level is considered for design purposes, which implies that the building can be used once the seismic event has passed. At this level, the stiffness requirements are high, the behavior of the structure must be within the elastic-linear range and, consequently, the seismic loads developed are important. The standard used does not consider other criteria apart from those concerning structural damage. That is the reason why non-structural issues, such as furniture overturning or partition damage, have not been considered in this research. However, it is worth noting that the "immediate occupancy" level implies a very low risk of life-threatening injury as a result of structural damage.
The lateral drift ratio is defined as the quotient between the horizontal displacement of the structure at the loading point and the distance from this point to the centroid of the connection between the wall and the footing. In this case, the maximum allowable lateral drift ratio (see Figure 7) is 0.5%. Additionally, some other criteria must be fulfilled [34]:

1.
No plastic behavior is observed in the structure, neither concrete nor steel rebars; 2.
Crack width should be below 1.6 mm; 3.
No concrete crushing is observed.
Regarding the testing procedure, it consisted of applying a quasi-static horizontal load near the top of the structural element (wall or frame). Displacement controlled reverse cyclic tests were performed. The number of loading cycles for each amplitude of the imposed displacement was kept at two. The value of the amplitude depends on the critical drift ϕ y , which is defined as the drift associated with yielding. The first couple of cycles corresponded to a drift equal to 0.5·ϕ y , the second couple of cycles, to a drift equal to ϕ y , the third one, to a drift equal to 2·ϕ y , and for the remaing couple of cycles the amplitude increased in 1·ϕ y , i.e., 3·ϕ y , 4·ϕ y , 5·ϕ y , and so on to the conclusion of the test. The test concluded when the maximum load of one cycle was more than 20% lower than the total maximum load of all cycles, according to ACI 374.2R-13 [34].
In the case of the tests on the isolated structural walls, the load was applied at a height of 2.5 m and the critical drift ϕ y corresponded to a horizontal displacement of 10.6 mm at the load application Appl. Sci. 2020, 10, 1213 8 of 22 point. In the case of the tests on the structural frames, the load was applied at a height of 2.94 m and the critical drift ϕ y corresponded to a horizontal displacement of 12.6 mm at the load application point.
The tests on both the isolated structural walls and the structural frames were carried out using a tension-compression MTS 201.70F dynamic actuator (MTS, Eden Prairie, MN, USA), with a capacity of ±1000 kN. The actuator had a load cell MTS 661.31F-01 (MTS, Eden Prairie, MN, USA), with a range of ±1000 kN and an error of below 1% of the range. The tests were displacement controlled. This testing procedure provided greater safety against unexpected collapse and it better meet the requirement of the ACI 374.2R-13 [34].
In order to analyze the structural behavior of the testing specimens, a number of sensors were used, including inclinometers (model PST300, Pewatron AG, Zurich, Switzerland), linear potentiometer displacement transducers (ranged from 50 to 500 mm, Novotechnik, Ostfildern, Germany), and uniaxial strain gauges (150 mm length, Tokyo Sokki Kenkyujo Co., Ltd., Tokyo, Japan). Figures 8 and 9 show the position of the sensors in each of the two sets of testing specimens, i.e., the isolated structural walls and the structural frames.
In the case of the tests on the isolated structural walls, the load was applied at a height of 2.5 m and the critical drift φy corresponded to a horizontal displacement of 10.6 mm at the load application point. In the case of the tests on the structural frames, the load was applied at a height of 2.94 m and the critical drift φy corresponded to a horizontal displacement of 12.6 mm at the load application point.
The tests on both the isolated structural walls and the structural frames were carried out using a tension-compression MTS 201.70F dynamic actuator (MTS, Eden Prairie, MN, USA), with a capacity of ±1000 kN. The actuator had a load cell MTS 661.31F-01 (MTS, Eden Prairie, MN, USA), with a range of ±1000 kN and an error of below 1% of the range. The tests were displacement controlled. This testing procedure provided greater safety against unexpected collapse and it better meet the requirement of the ACI 374.2R-13 [34].
In order to analyze the structural behavior of the testing specimens, a number of sensors were used, including inclinometers (model PST300, Pewatron AG, Zurich, Switzerland), linear potentiometer displacement transducers (ranged from 50 to 500 mm, Novotechnik, Ostfildern, Germany), and uniaxial strain gauges (150 mm length, Tokyo Sokki Kenkyujo Co., Ltd., Tokyo, Japan). Figures 8 and 9 show the position of the sensors in each of the two sets of testing specimens, i.e., the isolated structural walls and the structural frames.

Seismic Tests on Real-Scale Three-Storey Precast Concrete Building
The third phase consists of seismic tests on a real-scale three-storey precast concrete building. In this case, a representative earthquake was reproduced in the laboratory, namely "El Centro" (an earthquake that occurred in the city of El Centro, California, USA in 1979). More specifically, the accelerogram belonging to an orientation of 220° was considered, because this was the most unfavourable in terms on maximum horizontal accelerations. The accelerogram was obtained from the USGS (United States Geological Survey, USA) (Figure 10a).
The accelerogram is first transformed in a diagram of horizontal displacement verus time, which is the input signal introduced to the dynamic actuator control equipment (Figure 10b). Additionally, the response spectrum of the earthquake is shown (Figure 10c).

Seismic Tests on Real-Scale Three-Storey Precast Concrete Building
The third phase consists of seismic tests on a real-scale three-storey precast concrete building. In this case, a representative earthquake was reproduced in the laboratory, namely "El Centro" (an earthquake that occurred in the city of El Centro, California, USA in 1979). More specifically, the accelerogram belonging to an orientation of 220 • was considered, because this was the most unfavourable in terms on maximum horizontal accelerations. The accelerogram was obtained from the USGS (United States Geological Survey, USA) ( Figure 10a).
The third phase consists of seismic tests on a real-scale three-storey precast concrete building. In this case, a representative earthquake was reproduced in the laboratory, namely "El Centro" (an earthquake that occurred in the city of El Centro, California, USA in 1979). More specifically, the accelerogram belonging to an orientation of 220° was considered, because this was the most unfavourable in terms on maximum horizontal accelerations. The accelerogram was obtained from the USGS (United States Geological Survey, USA) (Figure 10a).
The accelerogram is first transformed in a diagram of horizontal displacement verus time, which is the input signal introduced to the dynamic actuator control equipment (Figure 10b). Additionally, the response spectrum of the earthquake is shown (Figure 10c). The accelerogram is first transformed in a diagram of horizontal displacement verus time, which is the input signal introduced to the dynamic actuator control equipment ( Figure 10b). Additionally, the response spectrum of the earthquake is shown (Figure 10c). Figure 10c reveals that this earthquake causes the highest acceleration for structures with a natural period of 0.06 s, i.e., for structures with a natural frequency of 16.7 Hz. However, this earthquake is not only very dangerous for rigid structures, but it also provides acceleration values greater than the ground acceleration up to a period of 0.25 s, i.e., a frequency of 4 Hz.
The earthquake was not applied directly on the building, but in a progressive way, similar to foreshocks before the main earthquake. A total of six foreshocks were applied to the building before the main earthquake. To perform it, the ordinate of the seismic signal (i.e., the diagram of horizontal displacement vs. time) was multiplied by a factor. For the first foreshock, the factor was 0.05, i.e., the shape of this earthquake is homothetic to the real "El Centro" earthquake but the displacements are only 5% of the main earthquake. For the rest of the foreshocks, the factors were 0.1, 0.3, 0.5, 0.7, and 0.9 respectively. Finally, the main earthquake was applied.
The sensors used to monitor this test were load cell, accelerometers on the shake table, the intermediate slabs at the top of the walls, and displacement transducers in several positions ( Figure 11). only 5% of the main earthquake. For the rest of the foreshocks, the factors were 0.1, 0.3, 0.5, 0.7, and 0.9 respectively. Finally, the main earthquake was applied.
The sensors used to monitor this test were load cell, accelerometers on the shake table, the intermediate slabs at the top of the walls, and displacement transducers in several positions ( Figure  11).  Figure 12 shows a general view of the testing. As can be observer, in addition to the self-weight of the building, a dead load of 1.50 kN/m 2 was included on the intermediate slab. This load was materialized using water tanks.
Additionally, a one-cycle impulsive test was performed, before and after the seismic tests, in order to measure the natural frequency and the damping ratio of the building and to compare them with the excitation frequency of the earthquake. Moreover, the comparison of the natural frequency and damping ratio values, before and after the seismic event (including the main earthquakes and the foreshocks), provide useful information about the damage caused by the seismic tests.
In this case, a displacement-time one-cycle square wave signal was applied, with an excitation frequency of 5 Hz and an amplitude of 1 mm.

Experimental Results and Discussion
Next, the experimental results of the testings are exposed. In this case, the results of the most representative test of each phase are shown. Additionally, a one-cycle impulsive test was performed, before and after the seismic tests, in order to measure the natural frequency and the damping ratio of the building and to compare them with the excitation frequency of the earthquake. Moreover, the comparison of the natural frequency and damping ratio values, before and after the seismic event (including the main earthquakes and the foreshocks), provide useful information about the damage caused by the seismic tests.
In this case, a displacement-time one-cycle square wave signal was applied, with an excitation frequency of 5 Hz and an amplitude of 1 mm.

Experimental Results and Discussion
Next, the experimental results of the testings are exposed. In this case, the results of the most representative test of each phase are shown.

Pushover Tests on Isolated Structural Wall
As previously explained, the aim of this testing phase is to evaluate the ductility of the low-cost energy dissipation system by considering diagrams of horizontal load versus drift, drift versus strain in concrete, and bending moment versus rotation.

Experimental Results and Discussion
Next, the experimental results of the testings are exposed. In this case, the results of the most representative test of each phase are shown.

Pushover Tests on Isolated Structural Wall
As previously explained, the aim of this testing phase is to evaluate the ductility of the low-cost energy dissipation system by considering diagrams of horizontal load versus drift, drift versus strain in concrete, and bending moment versus rotation.   Some interesting conclusions can be obtained from Figure 13. First, a symmetric behavior under push and pull is obsesrved, as expected. Second, it can be observed that the behavior of the specimen is linear elastic up to the critical drift ϕ y (which is 0.5% according to ACI374.2R-13 [34]). In consequence, this solution agrees with one of the requirements of the American standard mentioned above.
Once the drift is greater than 0.5%, a progressive plastic behavior is observed, i.e., the specimen begins to dissipate energy at the expense of a higher deformation. The area enclosed by the hysteresis loop is proportional to the energy dissipated during the testing and represents the structural element capacity to mitigate the earthquake effect inelastically.
The maximum loads obtained during the testing were 13.0 kN in the push phase and 12.0 kN in the pull phase. In both cases, these loads correspond to a drift of 2%, which is four times greater than the critical drift. The loads obtained at the critical drift were 6.5 kN in the push phase and 6.2 kN in the pull phase. Figure 14 shows the relationship between the drift of the wall and the average vertical strain of the concrete at the base of the wall, at both the dorsal face (where the actuator is placed) and the frontal face (the opposite side). In this case, a positive value of strain denotes tension, and a negative value denotes compression. capacity to mitigate the earthquake effect inelastically.

Diagrams of Drift versus Strain in Concrete
The maximum loads obtained during the testing were 13.0 kN in the push phase and 12.0 kN in the pull phase. In both cases, these loads correspond to a drift of 2%, which is four times greater than the critical drift. The loads obtained at the critical drift were 6.5 kN in the push phase and 6.2 kN in the pull phase. Figure 14 shows the relationship between the drift of the wall and the average vertical strain of the concrete at the base of the wall, at both the dorsal face (where the actuator is placed) and the frontal face (the opposite side). In this case, a positive value of strain denotes tension, and a negative value denotes compression.  The diagrams in Figure 14 show an asymmetric behavior, i.e., compression strains are greater than tension strain. This could be because under tension, small microcracks in concrete occur, relaxing tension stress in concrete (and as a counterpart increasing the tension stress of the reinforcement), resulting in smaller values of tension strain.

Diagrams of Drift versus Strain in Concrete
This result confirms that the plastic behavior shown by the isolated structural wall is completely caused by the low-cost energy dissipation device. Moreover, once the plastic behavior of the structural wall is observed, a progressive decrease of the maximum vertical strain of concrete occurs.  The diagrams in Figure 14 show an asymmetric behavior, i.e., compression strains are greater than tension strain. This could be because under tension, small microcracks in concrete occur, relaxing tension stress in concrete (and as a counterpart increasing the tension stress of the reinforcement), resulting in smaller values of tension strain.
This result confirms that the plastic behavior shown by the isolated structural wall is completely caused by the low-cost energy dissipation device. Moreover, once the plastic behavior of the structural wall is observed, a progressive decrease of the maximum vertical strain of concrete occurs. This is because the elastic energy stored in the wall progressively flows to the energy dissipation device, preventing the wall from structural damage. Figure 15 shows the diagram of bending moment versus rotation of the connection between the wall and the footing. The bending moment is defined as the product of the horizontal force and the vertical distance between the force and the centroid of the low-cost energy dissipation system. The rotation is measured using an inclinometer placed at the level of the centroid of the energy dissipation device (see Figure 8). Positive values mean push, and negative values mean pull. Figure 15 confirms the findings shown in previous figures. On one side, it is observed that up to the critical drift ϕ y , the connection shows a linear elastic behavior. The slopes of the curves are high and quite similar under loading and unloading. The area under the hysteresis loop is small, which means that there is no energy dissipation. Additionally, no lose of stiffness is observed.

Diagram of Bending Moment versus Rotation
Once the structure reaches a drift of 2%, a plastic behavior starts to occur. Then, a progressive decrease of the stiffness of the connection is observed. The area under the hysteresis loop gradually increases, which reveals that there is a progressive energy dissipation. In general, a symmetric behavior of the connection is observed.
Appl. Sci. 2020, 10, x FOR PEER REVIEW 13 of 23 This is because the elastic energy stored in the wall progressively flows to the energy dissipation device, preventing the wall from structural damage. Figure 15 shows the diagram of bending moment versus rotation of the connection between the wall and the footing. The bending moment is defined as the product of the horizontal force and the vertical distance between the force and the centroid of the low-cost energy dissipation system. The rotation is measured using an inclinometer placed at the level of the centroid of the energy dissipation device (see Figure 8). Positive values mean push, and negative values mean pull.  Once the structure reaches a drift of 2%, a plastic behavior starts to occur. Then, a progressive decrease of the stiffness of the connection is observed. The area under the hysteresis loop gradually increases, which reveals that there is a progressive energy dissipation. In general, a symmetric behavior of the connection is observed.

Pushover Tests on Structural Frames
As previously explained, the aim of this testing phase is to evaluate the ductility of the frame, including the low-cost energy dissipation system, as well as the "flexible" connection between the slab and the wall. To obtain it, the diagrams of load versus drift, drift versus strain in concrete, and bending moment versus rotation are shown.
The tests carried out on the structural frames are not reversal (i.e., push and pull), but they are push and "unpush" (i.e., push the frame up to the maximum displacement of each cycle and return it back to the displacement until zero).

Pushover Tests on Structural Frames
As previously explained, the aim of this testing phase is to evaluate the ductility of the frame, including the low-cost energy dissipation system, as well as the "flexible" connection between the slab and the wall. To obtain it, the diagrams of load versus drift, drift versus strain in concrete, and bending moment versus rotation are shown.
The tests carried out on the structural frames are not reversal (i.e., push and pull), but they are push and "unpush" (i.e., push the frame up to the maximum displacement of each cycle and return it back to the displacement until zero). Some interesting findings can be observed from Figure 16. First, it is highlighted that a linearelastic behavior is observed until the critical drift φy. This result agrees with the American standard ACI374.2R-13 [34]. Once the critical drift is reached, a progressive plastification of the structure occurs. The area under the hysteresis loop progressively increases, which denotes that an energy dissipation process occurs.

Diagram of Horizontal Force versus Drift
The maximum load is reached for a drift of 3%, i.e., six times the critical draft. Beyond this value, the load does not increase or decrease, but it remains almost constant. However, the area under the hysteresis loops significantly increases. The structural solution shows a huge capacity of energy dissipation without losing structural capacity. Some interesting findings can be observed from Figure 16. First, it is highlighted that a linear-elastic behavior is observed until the critical drift ϕ y. This result agrees with the American standard ACI374.2R-13 [34]. Once the critical drift is reached, a progressive plastification of the structure occurs. The area under the hysteresis loop progressively increases, which denotes that an energy dissipation process occurs.
The maximum load is reached for a drift of 3%, i.e., six times the critical draft. Beyond this value, the load does not increase or decrease, but it remains almost constant. However, the area under the hysteresis loops significantly increases. The structural solution shows a huge capacity of energy dissipation without losing structural capacity. Figure 17 shows the relationship between the drift of the walls and the average vertical strain of the walls' concrete, near but below the connection with the slab, at both the internal and the external faces. In this case, a positive value of strain denotes tension, and a negative value denotes compression.

Diagrams of Drift versus Strain in Concrete
the load does not increase or decrease, but it remains almost constant. However, the area under the hysteresis loops significantly increases. The structural solution shows a huge capacity of energy dissipation without losing structural capacity. Figure 17 shows the relationship between the drift of the walls and the average vertical strain of the walls' concrete, near but below the connection with the slab, at both the internal and the external faces. In this case, a positive value of strain denotes tension, and a negative value denotes compression. Figure 17 reveals the behavior of the connection between the slab and the wall. First, a linearelastic behavior of the connection, up to a drift of 2%, is observed. In Wall 1, the one in contact with the actuator, compression strain is observed in the exterior face, as well as tension strain in the interior face. On the contrary, in Wall 2, tension strain is observed in the exterior face, as well as compression strain in the interior face. In each loading cycle, the loading and the unloading branches are almost identical and the area under the hysteresis loop is small, which denotes an absence of energy dissipation. During this first phase of the testing, the visual inspections revealed very small horizontal cracks in the walls (especially in the external face of Wall 2 where the tension strain was larger) with a crack width below 0.2 mm, i.e., clearly smaller than 1.6 mm which is the maximum allowable crack width defined by the ACI374.2R-13 [34]. No concrete crushing occurred. The maximum measured strain values belong a drift of 2%. Beyond this value, there was a progressive decrease in the maximum measured strain, which denotes that the stiffness of the connection between the slab and the wall decreased and a plastic hinge appeared in this connection. Moreover, this plastic hinge showed an asymmetric behavior, i.e., its structural behavior was different when it was subjected to a positive bending (tension in the inner face of the wall and the lower face of the slab) or a negative bending (tension in the outer face of the wall and the upper face of the slab).

Diagrams of Drift versus Strain in Concrete
Because of the type of loading cycles of the test, the connection between Wall 1 and the slab was always under positive bending while the connection between Wall 2 and the slab was always under negative bending. Wall 1 showed values of strain (both tension and compression) lower than the values observed in Wall 2. This means that the connection shows a stiffness under positive bending smaller than the one under negative bending.
Once the drift was beyond 2%, a clear plastic behavior started to be observed. In each loading cycle, the loading and the unloading branches were different and the area under the hysteresis loop increased with the cycles, which denoted an increased ability of the connection to dissipate energy. In this case, the visual inspections carried out during the testings revealed small horizontal cracks in the walls (especially in the external face of Wall 2 where the tension strain was larger). However, the crack widths were always below 1.6 mm which is the maximum allowable crack width defined by the ACI374.2R-13 [34]. No concrete crushing occurred.  Figure 17 reveals the behavior of the connection between the slab and the wall. First, a linear-elastic behavior of the connection, up to a drift of 2%, is observed. In Wall 1, the one in contact with the actuator, compression strain is observed in the exterior face, as well as tension strain in the interior face. On the contrary, in Wall 2, tension strain is observed in the exterior face, as well as compression strain in the interior face. In each loading cycle, the loading and the unloading branches are almost identical and the area under the hysteresis loop is small, which denotes an absence of energy dissipation. During this first phase of the testing, the visual inspections revealed very small horizontal cracks in the walls (especially in the external face of Wall 2 where the tension strain was larger) with a crack width below 0.2 mm, i.e., clearly smaller than 1.6 mm which is the maximum allowable crack width defined by the ACI374.2R-13 [34]. No concrete crushing occurred.
The maximum measured strain values belong a drift of 2%. Beyond this value, there was a progressive decrease in the maximum measured strain, which denotes that the stiffness of the connection between the slab and the wall decreased and a plastic hinge appeared in this connection. Moreover, this plastic hinge showed an asymmetric behavior, i.e., its structural behavior was different when it was subjected to a positive bending (tension in the inner face of the wall and the lower face of the slab) or a negative bending (tension in the outer face of the wall and the upper face of the slab).
Because of the type of loading cycles of the test, the connection between Wall 1 and the slab was always under positive bending while the connection between Wall 2 and the slab was always under negative bending. Wall 1 showed values of strain (both tension and compression) lower than the values observed in Wall 2. This means that the connection shows a stiffness under positive bending smaller than the one under negative bending.
Once the drift was beyond 2%, a clear plastic behavior started to be observed. In each loading cycle, the loading and the unloading branches were different and the area under the hysteresis loop increased with the cycles, which denoted an increased ability of the connection to dissipate energy. In this case, the visual inspections carried out during the testings revealed small horizontal cracks in the walls (especially in the external face of Wall 2 where the tension strain was larger). However, the crack widths were always below 1.6 mm which is the maximum allowable crack width defined by the ACI374.2R-13 [34]. No concrete crushing occurred. Figure 18 shows the diagrams of bending moment versus rotation of the connections between Walls 1 and 2 and the slab. The bending moment is defined as the product of the horizontal force and the vertical distance between the force and the centroid of the low-cost energy dissipation system. This is, in fact, a "global bending moment" of the frame, and not the real moment of the connection between the wall and the slab. The rotation is defined as the variation of the inner angle between the wall and the slab. The behavior observed in Figure 18 agrees with the one shown in Figure 16. The first hysteresis cycles (up to a drift of 2%) reveal a linear-elastic behavior of the connections. The loading and the unloading branches are very similar, and the areas enclosed by the hysteresis loop are very small.

Diagram of Bending Moment versus Rotation
Beyond a drift of 2%, the structure begins to show a plastic behavior. In each hysteresis loop, the slope of the curve bending moment versus rotation progressively decreases, and the area enclosed by the hysteresis loop progressively increases. Consequently, the energy dissipation capacity of the connection between the wall and the slab increases. Additionally, the permanent rotation corresponding to null bending moment increases in each cycle, which denotes that the connection suffers damage in each cycle.
An unexpected behavior is observed during the last cycle in the diagram concerning Wall 1. In particular, there is an interruption of the data capture from rotations around 0.03 rad. This can be explained because the measurement range of the transducers used to calculate rotations was exceeded. As a result, the transducers are detached from the concrete surfaces and there is no data collection.
When both connections are compared, it is observed that the one placed in Wall 1 (and consequently subjected to positive bending) shows less stiffness than the one in Wall 2 (subjected to negative bending). This finding agrees with the results of Figure 17. Moreover, the permanent rotation corresponding to null bending moment in the connection of Wall 1 is larger than the one in Wall 2. The behavior observed in Figure 18 agrees with the one shown in Figure 16. The first hysteresis cycles (up to a drift of 2%) reveal a linear-elastic behavior of the connections. The loading and the unloading branches are very similar, and the areas enclosed by the hysteresis loop are very small.
Beyond a drift of 2%, the structure begins to show a plastic behavior. In each hysteresis loop, the slope of the curve bending moment versus rotation progressively decreases, and the area enclosed by the hysteresis loop progressively increases. Consequently, the energy dissipation capacity of the connection between the wall and the slab increases. Additionally, the permanent rotation corresponding to null bending moment increases in each cycle, which denotes that the connection suffers damage in each cycle.
An unexpected behavior is observed during the last cycle in the diagram concerning Wall 1. In particular, there is an interruption of the data capture from rotations around 0.03 rad. This can be explained because the measurement range of the transducers used to calculate rotations was exceeded. As a result, the transducers are detached from the concrete surfaces and there is no data collection.
When both connections are compared, it is observed that the one placed in Wall 1 (and consequently subjected to positive bending) shows less stiffness than the one in Wall 2 (subjected to negative bending). This finding agrees with the results of Figure 17. Moreover, the permanent rotation corresponding to null bending moment in the connection of Wall 1 is larger than the one in Wall 2.

Seismic Tests on Real-Scale Three-Storey Precast Concrete Building
Once the two testing phases have been completed (the first one on isolated structural walls and the second one on structural frames), the seismic tests on a real-scale three-storey precast concrete building were performed. The aim of this third testing phase is to validate the structural solution implemented on a real building subjected to an earthquake. In this case, the structural behavior of the building under the seismic events mainly depends on the connections, both the low-cost energy dissipation systems placed on the connections between the walls and the footings and the flexible connections between the walls and the slabs.
To obtain it, the following parameters are monitored during a real seismic event: Longitudinal displacement of the shake table, longitudinal displacements of the building at the storey levels, and longitudinal accelerations of the shake table and the building. The results shown in the following figures belongs only to the main earthquake and not to the foreshocks. Figure 19 shows the diagrams of the longitudinal displacement versus time during the "El Centro" earthquake at the following locations: shake table, Concrete Slab 1, Concrete Slab 2 and lightweight roof.

Longitudinal Displacement of the Shake Table and the Building
Appl. Sci. 2020, 10, x FOR PEER REVIEW 17 of 23 Figure 19 shows the diagrams of the longitudinal displacement versus time during the "El Centro" earthquake at the following locations: shake table, Concrete Slab 1, Concrete Slab 2 and lightweight roof.    Moreover, the visual inspections carried out after the seismic tests revealed that no structural damage is observed in the building (i.e., no cracks in the walls or slabs appeared and, of course, no concrete crushing occurred). This means that the seismic energy was completely dissipated by the connections, i.e., by the low-cost energy dissipation systems placed on the connections between the walls and the footings and the flexible connections between the walls and the slabs. The main aim of the research, which is the design and validation of a low-cost energy dissipation system, as well as the flexible connection between the walls and the slabs, has been reached.
At the end of the seismic event, the residual longitudinal displacements of the both concrete slabs and the flexible roofs are almost zero, which means that the building recovers its original position, that is, the walls recover their upright position.
Finally, it is concluded that the building reached the performance level of "immediate occupancy", according to ACI374.2R-13 [34]. Table and the Building Figure 20 shows the diagrams of the longitudinal acceleration versus time during the "El Centro" earthquake at the following locations: shake  Figure 20 shows the diagrams of the longitudinal acceleration versus time during the "El Centro" earthquake at the following locations: shake table, Concrete Slab 1, Concrete Slab 2, and lightweight roof.    20 reveals that this structure is especially resistant to the earthquake "El Centro", since the maximum measured longitudinal accelerations of the concrete slabs is smaller than the one on the shake table. The maximum measured longitudinal acceleration at the lightweight roof is a bit larger than the one on the shake table. Specifically, the maximum measured longitudinal acceleration is 2.8 m/s 2 on the shake table, 1.7 m/s 2 on the first concrete slab, 2.1 m/s 2 on the second concrete slab, and 4.0 m/s 2 on the top of the building.

Longitudinal Accelerations of the Shake
The dominant excitation frequency of the earthquake "El Centro" is around 1.6 Hz (Figure 21), and the natural frequency of the building is around 3.4 Hz before the seismic tests ( Figure 22) and 2.6 Hz after them ( Figure 23). This large difference between the excitation frequency and the natural frequency implies that the longitudinal accelerations that the earthquake causes in the building are small. Consequently, the horizontal inertial forces are also small, as well as the internal forces caused by the earthquake. The seismic event results in a reduction of the natural frequency of the building by 0.8 Hz, that is, 23%. Since no concrete cracks were observed, it is concluded that the damage is completely focused on the connections, both the low-cost energy dissipation system and the flexible connections between slab and walls. The seismic event results in a reduction of the natural frequency of the building by 0.8 Hz, that is, 23%. Since no concrete cracks were observed, it is concluded that the damage is completely focused on the connections, both the low-cost energy dissipation system and the flexible connections between slab and walls. Additionally, the measured damping factor of the building is 4.7% before the seismic tests and 6.2% after them. This increase in the damping factor is also a good indicator of the damage caused by the seismic tests. The seismic event results in a reduction of the natural frequency of the building by 0.8 Hz, that is, 23%. Since no concrete cracks were observed, it is concluded that the damage is completely focused on the connections, both the low-cost energy dissipation system and the flexible connections between slab and walls. Additionally, the measured damping factor of the building is 4.7% before the seismic tests and 6.2% after them. This increase in the damping factor is also a good indicator of the damage caused by the seismic tests. The seismic event results in a reduction of the natural frequency of the building by 0.8 Hz, that is, 23%. Since no concrete cracks were observed, it is concluded that the damage is completely focused on the connections, both the low-cost energy dissipation system and the flexible connections between slab and walls.
Additionally, the measured damping factor of the building is 4.7% before the seismic tests and 6.2% after them. This increase in the damping factor is also a good indicator of the damage caused by the seismic tests.
When the seismic response of the building is compared to Eurocode 8 [38], it is observed that the measured elastic response spectrum S(T), defined as the ratio between the maximum acceleration of the building and the ground acceleration (i.e., the maximum acceleration of the shake table) is 1.43, which is smaller than the theoretical S(T) provided by this European standard. This means that the standard is conservative, as expected.
Additionally, it is highlighted that the natural frequency of the structure is below 4 Hz before the seismic event (Figure 22b), which is the lowest threshold of the dangerous region of the "El Centro" earthquake (see Figure 10c) and no relevant accelerations are developed during the seismic events. Moreover, the loss of stiffness caused by the seismic events reduces the natural frequency (Figure 23b) of the structure and, consequently, reduces the horizontal accelerations caused by the earthquake, which prevents the structure from aftershock earthquakes.

Conclusions
This paper shows the design and the laboratory validation tests of a new low-cost energy dissipation system, for application in precast concrete structures composed of precast footings, precast structural walls, and precast concrete slabs. This energy dissipation system basically consists of a specific connection between the precast footing and the precast structural wall, formed by a set of threaded steel bars that connect both elements. During an earthquake, the steel bars undergo plastic deformation, absorbing most of the energy generated by the earthquake and preventing damage to the rest of the building. The additional advantage of this solution is that steel bars can be easily replaced after the seismic event. Additionally, a flexible connection between walls and slab has been used.
A testing campaign was carried out, including three phases. First, pushover tests were carried out on isolated structural walls formed by one precast structural wall and a precast footing. Second, pushover tests were carried out on structural frames, composed of two precast structural walls placed over two precast footings and connected to a precast slab. Thirdly, seismic tests using a shake table were carried out on a real-scale three-storey precast concrete building, consisting of two precast structural walls placed over two precast footings, two intermediate precast slabs, and a flexible steel roof.
The aim of this structural solution is to fulfill the requirements of the American standard ACI374.2R-13 [34] and more specifically, fulfill the structural performance level of "immediate occupancy", which means that the building can be used without collapse risk once the seismic event has occured.
The pushover tests on isolated structural walls revealed that this solution exhibits a linear-elastic behavior until and beyond the critical drift (which is 0.5%) and no cracks were observed in the structures. The maximum load was reached at a drift of 3%, which was six times greater than the critical drift. Beyond this drift, the structure began to show a plastic behavior. However, no structural damage was observed in the concrete elements, which means that the majority of the energy dissipated by the structural element was through the low-cost energy dissipation device. Additionally, a great ductility of the solution was observed.
The pushover tests on structural frames revealed that the flexible connection between the walls and the slab exhibited an excellent structural behavior. Again, this solution exhibited a linear-elastic behavior until and beyond the critical drift (which is 0.5%) and no cracks were observed in the structures. The maximum load was reached at a drift of 2%, which was four times greater than the critical drift.
Beyond this drift, the structure began to show a plastic behavior. At this moment, the flexible connections worked as an asymmetrical plastic hinge, able to transmit relevant negative bending moments but almost negligible positive bending moments. This reduction in the overall horizontal stiffness of the frame resulted in an increased plastic behavior of the structure and, in consequence, an increased capacity to dissipate seismic energy. In this case, small horizontal cracks in the walls (especially in the external face of the Wall 2 where the tension strain is larger) were observed.
The seismic tests revealed an excellent behavior of the real-scale three-storey precast concrete building. The structure was subjected to a main earthquake and six foreshocks. The earthquake used was "El Centro" (the eathquake that occurred in the city of El Centro, California, USA in 1979). Additionally, two one-cycle impulsive tests were performed, one before the seismic events and the other after the seismic events, in order to measure the dynamic parameters of the building (natural frequency and damping ratio) before and after the seismic events.
The visual inspections carried out after the seismic tests revealed that no structural damage was observed in the building (i.e., no cracks in the walls or slabs appeared and, of course, no concrete crushing occurred). This means that the seismic energy was completely dissipated by the connections, i.e., by the low-cost energy dissipation systems placed on the connections between the walls and the footings and the flexible connections between the walls and the slabs.
At the end of the seismic event, the residual longitudinal displacements of the both concrete slabs and the lightweight roofs were almost zero, i.e., the building recovered its original position.
The impulsive tests revealed that the seismic events caused a decrease in the natural frequency and an increase in the damping ratio, which illustrates the damage given by the seismic tests.
Finally, it is concluded that the building reached the performance level of "immediate occupancy", according to ACI374.2R-13 [34].