Field testing of gravel-rubber mixtures as geotechnical seismic isolation

We present the results of the forced-vibration experiments performed at the large-scale prototype structure of EuroProteas founded on gravel-rubber mixture (GRM) layers acting as a means of Geotechnical Seismic Isolation (GSI). Three GRM with different rubber content per mixture weight (0%, 10%, and 30%) but the same mean grain size ratio were used as foundation soil. Each GRM-structure system was subjected to harmonic forces in a wide range of excitation frequencies and force amplitude. It was found that a 0.5 m thick GRM foundation soil layer with 30% rubber content can effectively isolate the structure. The strong effect of the rubber fraction was expressed in the detected period elongation and the dominating rocking component which leads to a more “rigid-body” response of the structure. Moreover, the developed base shear and base moment are significantly reduced regardless of the excitation frequency, while the increased damping of the system and the important energy dissipation demonstrate the effectiveness of the GRM foundation soil layer. Overall, the experimental results demonstrated that the use of GRM as a GSI system can be considered as a low-cost alternative seismic isolation technique.


Introduction
Over the past decades, several seismic isolation techniques have been used in civil engineering practice, aiming to reduce the response of the structures and their seismic damage. The most widely adopted "structural seismic isolation" methods include high damping elastomeric bearings, sliding elements, and passive energy dissipated systems such as tuned-mass dampers or tuned-liquid dampers (Constantinou et al. 1998;Naeim and Kelly 1999). These methods are based either on decoupling the structure from the ground motion 1 3 or energy dissipation by incorporating high damping materials. There are many examples of the application of these techniques for the construction of earthquake-resistant structures. However, it is unsurprising that most applications concern important buildings or infrastructure with sensitive equipment. Moreover, the increased construction and maintenance cost of these isolation systems prohibit their use in conventional structures, especially in developing countries.
Thus, the urge to develop alternative low-cost seismic isolation systems has led to the proposal of various novel seismic isolation methods (Kelly 2002;Banović et al. 2019). Many researchers proposed the exploitation of soil's deformability to achieve a "natural" passive isolation mechanism (Trifunac and Todorovska 1998). The ground motion is attenuated through the soil's nonlinear response and yielding below the structure in combination with the foundation rocking, uplift, and sliding (Anastasopoulos et al. 2010;Gazetas 2015). Despite the appealing effect of the rocking isolation mechanism, the potentially sizeable residual differential settlement of the structure and the need for its realignment can be considered a drawback.
Recently, researchers have shown an increased interest in the novel concept of Geotechnical Seismic Isolation (GSI) as defined in Tsang (2009). The main idea of this strategy is to surround the structure's foundation with a low-modulus material that dissipates the seismic energy before it is transmitted to the superstructure. Among other materials proposed for the foundation isolation, such as soil-rubber mixtures (SRM) and gravel-rubber mixtures (GRM) (Anastasiadis et al. 2012a;, volcanic pumice -rubber mixtures (Tsinaris et al. 2018), geosynthetic materials (Yegian and Kadakal 2004), polyurethane foam (Gatto et al. 2022) or PVC (Tsiavos et al. 2020), have attracted significant research interest. Numerous studies (Edil and Bosscher 1994;Edeskär 2006) have reported that granulated rubber, recently included in ASTM (2008) and CEN/TS14243 (2010) standards, is characterized by low unit weight, low bulk density, high hydraulic conductibility, and high elastic deformability. Furthermore, extensive laboratory studies on SRM and GRM have shown that the rubber content is one of the main parameters that control the dynamic properties of the mixture, as the minor strain dynamic stiffness ( G o ) decreases and the initial damping ( D o ) rises with increasing rubber content (Anastasiadis et al. 2012b;Senetakis et al. 2012b;Pistolas et al. 2018).
Additionally, using SRM or GRM manufactured from disposed scrap tires in civil engineering projects has obvious environmental advantages. Until recently, few developed countries (USA, Japan, etc.) have imposed regulations to manage waste tire recycling. Most developing and agricultural countries have not yet explored alternative uses of scrap tires that are disposed of in landfills raising severe environmental concerns (Tasalloti et al. 2021). Simultaneously, rubber mixtures can be considered an affordable alternative for the seismic isolation of conventional structures, especially in developing countries. Tsang (2008), in the first systematic numerical study on the effectiveness of the use of a SRM in the seismic isolation of structures, reported that a SRM layer placed underneath a structure could effectively reduce its horizontal and vertical ground motion response. At the same time, its low cost could benefit the developing countries. Since then, several studies have examined the advantages of SRM or GRM numerically as a GSI system in the form of a layer underlying the foundation of a structure (Mavronicola et al. 2010;Pitilakis et al. 2015;Brunet et al. 2016;Tsang and Pitilakis 2019;Pistolas et al. 2020). So far, however, very few experimental studies on the investigation of the application of GRM as a GSI system have been conducted or are under development, while most of them are mainly limited to laboratory and small-scale testing (Kaneko et al. 2013;Xiong et al. 2014;Anbazhagan et al. 2017;Tsiavos et al. 2020;Tsang et al. 2021;Hernández et al. 2020).
Recently, the first large-scale field experimental study of the dynamic response of the EuroProteas prototype structure founded on GRM layers was performed . The experimental findings showed that a thin GRM layer with a height of 0.50 m, including 30% rubber content per mixture weight, could effectively isolate the structure from the foundation soil. Moreover, based on the analysis of ambient noise, free-and forced vibration test measurement, it was successfully demonstrated that the stiffness of the GSI-structure system was substantially decreased, while its damping was increased.
The present study provides further important insights on the response of the GSI-structure system in the forced vibration tests. Four forced-vibration test series were performed on the EuroProteas structure resting on three GRM layers with different rubber fractions per mixture weight (0%, 10%, and 30%). The effect of the rubber content of the GRM on the response of the system is examined in terms of period lengthening, base shear and moment and kinetic energy. Overall, it is shown that a structure founded on a GSI layer composed of a GRM with 30% rubber content per weight can be effectively isolated.

EuroProteas full-scale model structure
EuroProteas prototype structure ( Fig. 1) was constructed at the EuroSeisTest experimental facility, 30 km northeast of the city of Thessaloniki, in Greece (http:// euros eisdb. civil. auth. gr/ sfsi). It was designed to serve as a simple test structure that promotes soil-foundationstructure interaction (SFSI) phenomena. The combination of its large superstructure mass and the soft foundation soil enables SFSI effects and the soil's nonlinear behavior due to the excitation of the soil-structure system. EuroProteas is founded on a reinforced concrete slab (C20/25) of dimensions 3.00 × 3.00 × 0.40 m. Two identical reinforced concrete slabs representing the superstructure are supported by four steel columns of section QHS 150 × 150 × 10 mm clamped on the foundation slab. The columns are connected with steel X-braces of section L 100 × 100 × 10 mm in all directions to ensure the structure's symmetry. The mass of each slab is estimated at around 9.16 mg, assuming a uniform unit weight of 25 kN/m 3 for the reinforced concrete. The total mass of the structure is calculated approximately at 28.5 Mg. The total height of EuroProteas is 5 m. More details on the prototype's design and construction can be found in Pitilakis et al. (2018) and Vratsikidis et al. (2021).
The upper roof slab and the X-braces are removable, allowing the structure's mass and stiffness modification, respectively. As a result, the fixed-base frequency of EuroProteas can be adjusted in a range that was calculated numerically between 1.78 and 13.06 Hz ). In the experimental program presented in this paper, the configuration of EuroProteas involved X-bracing on all four sides and two roof slabs. Its fixed-base fundamental natural frequency was defined at 9.13 Hz.

Foundation soil improved with GRM
EuroProteas structure is located at the center of the EuroSeisTest experimental facility, which was established since 1993 and is maintained under the responsibility of the Research Unit of Soil Dynamics and Geotechnical Earthquake Engineering (SDGEE, http:// sdgee. civil. auth. gr/) of the Aristotle University of Thessaloniki. Several geophysical and geotechnical studies, including Cross-Hole and Down-Hole tests, surface wave inversion tests, sampling boreholes, N SPT and CPT tests, and laboratory tests on undisturbed soil samples have been conducted to define the foundation soil stratigraphy and its dynamic properties (Pitilakis et al. 1999;Raptakis et al. 2000;Manakou et al. 2010;Pitilakis et al. 2018). The results of the field and laboratory tests represented a 7 m thick upper layer of silty clayey sand, which overlies a layer of clayey to silty sand with gravels between 7 and 22 m and a layer of marly silt to silty sand until the depth of 30 m. The shear wave velocity was estimated between 100 and 150 m/s in the uppermost 5 m, while it increases to more than 250 m/s at 25 m depth (Fig. 2).
As this study examines the effect of the rubber content of the GRM layer on the dynamic response and the overall performance of the GSI system, we considered three GRM layers with different fractions of rubber and gravel as foundation soil materials. The characteristics of the gravel and rubber materials used in the preparation of the mixtures are presented in Pitilakis et al. (2021). The first mixture consisted only of gravel without any rubber (GRM100/0) to serve as a benchmark GSI case, while the rubber contents of the other two mixtures were defined to be 10% (GRM90/10) and 30% (GRM70/30). The mean grain size ratio (D 50,R /D 50,G ) of the GRM90/10 and GRM70/30 was the same and equal to 0.16. According to Chew et al. (2022) the matrix material of the GRM90/10 is gravel and its response is anticipated to be similar to a gravel-like material. On the other hand, GRM70/30, having a volumetric rubber content approximately equal to 50%, is considered a material with intermediate behavior, as both gravel and rubber form the skeleton of the mixture. Thus, it is expected to increase the damping of the system as it can sustain larger deformation during loading.
To prevent any possible segregation (Kim and Santamarina 2008), we added a small amount of water (approximately 2-3% per mixture) while mixing the gravel and rubber materials. It should be noted that at the end of the experiments no evidence of segregation was observed. The thickness of the layer was set to 0.5 m to avoid deep excavation issues and reduce the construction cost of the proposed seismic isolation solution. We excavated three 3.20 × 3.20 m square plan pits down to 0.5 m depth and filled them with the three mixtures. A thin geotextile layer was used to cover the base and the pit walls to prevent the GRM from mixing with the underlying and surrounding soil (Fig. 3). The structure was placed at the top of each GRM foundation layer (Fig. 4).The physical properties of the GRM used in the tests are presented in Table 1. The minimum and maximum relative densities of the mixtures were calculated according to ASTM (2000b) (Method C) and ASTM (2000a) specifications, respectively. A detailed description of the installation of the GRM layers on site is given in Pitilakis et al. (2021). Figure 5 shows the positions of the sensors used in the tests. Eight triaxial accelerometers (Etna2, Kinemetrics Inc. and CMG-5TCDE, Guralp Systems Ltd) were used to record the motion of EuroProteas. Four of them were installed on the roof; two along the axis parallel to the direction of shaking (in-plane) and the other two at the opposite corners of the slab to capture possible out-of-plane motion. Four more accelerometers were bolted on the foundation slab forming a cross shape to capture the foundation's translational, rocking, and possible out-of-plane motion. Furthermore, four laser sensors

Experimental setup
The experimental campaign included ambient noise measurements, and free-and forced-vibration tests. A detailed description of the ambient noise measurements and the free-vibration tests is thoroughly presented in Pitilakis et al. (2021). In contrast, this paper will focus on the forced-vibration tests.
The three GSI-structure systems were subjected to forced vibration testing by the eccentric mass vibrator system MK-500U (ANCO Engineers Inc.), provided by the Institute of Engineering Seismology and Earthquake Engineering (ITSAK-EPPO). The unidirectional dual counter-rotating shaker that can produce a maximum sinusoidal horizontal force amplitude of 50 kN was installed at the geometrical center of the upper roof slab. The eccentricity of the pairs of the steel plates that act as weights was adjusted between 0.15 and 11.31 kg-m. The harmonic output force was calculated as where E is the total eccentricity of the shaker in kg-m and = 2 f is the rotational speed of the shaker in rad/s. Four identical series of forced-vibration experiments were performed at each GSIstructure system based on the four different eccentricities of the shaker. The frequency and the amplitude of the applied harmonic forces that are presented in this study are summarized in Table 2. At each excitation frequency, the structure was shaken for a time window of 25 s to reach a steady state.
(1) F s = E 2 sin t  Figure 6 presents the maximum acceleration amplitude recorded by the instruments installed at the roof of EuroProteas versus the excitation frequency in the forced-vibration tests 3 and 4 when the strongest harmonic forces were applied. The tests were performed while the structure was founded on the three GRM layers. The peak amplitude is defined as the mean absolute value of the maxima and the minima for a small number of cycles when the structure's vibration reaches a steady state. The motion of the structure founded on the gravelly layer reached a peak when the excitation frequency was set to 4Hz in both tests. This finding matches the results of the system identification presented in Pitilakis et al. (2021). The structure's response founded on the layer with 10% rubber content per mixture weight is almost similar to the reference case (100% gravel), indicating that adding a low rubber fraction has a marginal influence on the dynamic response of the structure. Furthermore, the peak values recorded at the roof slab of the GRM90/10-EuroProteas system are slightly lower, indicating a slight increase in the system's damping due to the presence of the rubber in the foundation soil layer.

Structural response
On the other hand, when the structure is lying on the GRM70/30 layer, a clear peak is noticed at the excitation frequency of 2.5Hz, confirming the shift in the resonant frequency of the GSI-structure system as identified by Pitilakis et al. (2021). This change in the predominant frequency is attributed to the decrease in the system's stiffness due to the effect of the rubber. Moreover, the increase in the rubber content of the GRM to 30% results in a significant reduction in the recorded acceleration in almost the whole excitation frequency range due to the effect of the rubber on the material damping of the GRM. For frequencies less than 3 Hz, the structure's response founded on the GRM70/30 is more significant. However, this result is explained by the fact that, in this case, the structure is excited close to its first modal frequency. These findings corroborate the behavioral zones identified by the analyses performed by Chew et al. (2022). The effectiveness of the rubber content of the GRM layer in the reduction of the structural response was also assessed in terms of an acceleration reduction factor defined as where ü rs,R is the peak acceleration recorded at each excitation frequency at the roof of EuroProteas when founded on one of the GRM layers with rubber content greater than 0%, and ü rs,G is the corresponding peak acceleration value recorded when the structure is founded on the gravelly layer. Figure 7 presents the acceleration reduction factor evaluated for the two GRM layers in the forced vibration tests 3 and 4 versus the peak acceleration recorded at the top of the GRM100/0-EuroProteas system at the same excitation frequencies. In both tests, the effectiveness of the GRM layer with 30% rubber content is relatively high, ranging from 0.3 to 0.8. In addition, the values of the reduction factor seem to be independent of the amplitude of the motion. The negative results are attributed to the greater response of the GRM70/30-Europroteas system when excited close to its resonant frequency. In contrast, the GRM layer of 10% rubber fraction seems to have almost no effect on the overall response of the GSI-structure system as the values of the reduction factor are found to be very low or even negative.

Foundation rocking amplification
The GSI-EuroProteas systems can be modeled as a flexible-base single degree of freedom structure. The response at the top of the SDOF model, ü rs is assumed to be the sum of three components where ü f = foundation translational acceleration estimated by subtracting the translational acceleration related to the foundation rotational acceleration, h f̈f (h f is the height of the foundation slab), from the horizontal acceleration recorded at the top of the foundation (2) 1 −ü rs,R ∕ü rs,G (3) u rs =ü f +ü s + ḧf

(a)
Exp 4 (b) Fig. 7 Acceleration reduction factor estimated for the GSI-structure systems with 10% and 30% rubber content versus the roof peak acceleration of the structure founded on the gravelly soil at the same input frequency slab; ü s = structural translational acceleration due to bending; ḧf = translational acceleration related to the foundation rotation defined as the product of the height of the roof, h, and the foundation rotational acceleration ̈f . The latter is calculated as the difference between the vertical components of the two instruments installed on the foundation edges divided by their intermediate distance. The maximum amplitude of the motion components calculated when the structure is excited in a wide frequency range is presented in Fig. 8. The presence of 10% rubber content in the GRM layer only slightly affects the structure's response. A modest reduction is noticed in the horizontal translational acceleration due to the bending of the structure, which indicates that the increase in the rubber content leads to a more rigid-body response of the structure. This argument is also supported

(e)
Exp 4 (f) Fig. 8 Peak amplitude of the three components of the roof response of EuroProteas founded on the three GRM layers at different excitation frequencies in tests 3 and 4 by the almost similar values of the horizontal acceleration due to the foundation rotation when the structure is founded on the GRM100/0 and GRM90/10, taking into consideration the fact that the total horizontal acceleration of the GRM90/10-EuroProteas system was slightly reduced as shown in Fig. 6. When the rubber content is 30%, both the horizontal translational acceleration due to the bending of the structure and the horizontal acceleration due to the rocking of the foundation are significantly reduced for shaker frequencies over 3Hz. When the structure is excited in lower frequencies, its response when founded on the GRM70/30 is greater as the excitation frequency is close to its resonant frequency.
Moreover, the response of the EuroProteas-GRM70/30 system is remarkably increasing at 7Hz and at 6Hz in test 3 and 4, respectively. This is possibly attributed to the excitation of the structure close to its 3rd modal frequency, estimated at approximately 7Hz . Regarding the foundation translational acceleration, there is no clear trend on the effect of the rubber content on it as it can be considered as only a tiny portion of the overall response of the structure in all three cases.

Base shear and moment
The motion of a single degree of freedom system subjected to an external harmonic force F s produced by an eccentric mass shaker can be described by the following three equations representing the foundation translation, foundation rotation, and structural translation (Chopra 2012;Tileylioglu et al. 2011) where m s = superstructure mass; m f = foundation mass; I f = mass moment of inertia of the foundation slab; c i = damping terms; k s = structural stiffness; k x = lateral foundation stiffness; and k yy = rocking foundation stiffness. The base shear and moment rotation are calculated based on the first two rows of Eq. 4. The maximum amplitude of the base shear and moment calculated when the structure is excited in a wide frequency range of harmonic forces in tests 3 and 4 are presented in Fig. 9.
The amplitude values of the base shear and the base moment developed when the structure is founded on the GRM100/0 are similar to the corresponding values for the GRM90/10 in test 3. However, when the largest amplitude harmonic forces are applied to the structure in test 4, a decrease in the amplitude is noticed when the excitation frequency is close to the first modal frequency of the system. This discrepancy could be attributed to the system's more "rigid body" response due to the presence of the 10% rubber content, which also led to a reduction in the horizontal translational acceleration due to the bending of the structure, as observed previously in Fig. 8. A remarkable drop in the base shear and the base moment by over 50% and in some excitation frequencies reaching 90% is noticed when the rubber fraction of the GRM is increased to 30% for frequencies over 3Hz. The magnitude of these quantities is greater compared to the GRM100/0 and GRM90/10 only for frequencies around the resonant frequency of the GRM70/30-structure system.
In Fig. 10, the developed base shear and base moment for each frequency versus the recorded foundation motion components are presented. When the rubber content is 30% per weight in the GRM, it is noticed that the developed base shear and moment are (4) m s (ü f + ḧf +ü s ) + c sus + k s u s = F s significantly reduced compared to the GRM100/0-structure system for the same foundation translation and rotation.

GSI-structure system energy
An energy approach is adopted to quantify the relative amount of the kinetic energy dissipated by the three modes of response as described in Sect. 3.2. The time-varying kinetic energy that is dissipated through the three response modes is calculated at each excitation frequency as follows where I str = m str H str and I f = m f (B 2 + 4h 2 f )∕12 are the mass moment of inertia of the superstructure and the foundation mass about the centroidal axis of rocking at the base of the structure (B is the foundation width). The maximum amplitude of the kinetic energy related to each of the three response components when the structure is excited in a wide frequency range in tests 3 and 4 is presented in Fig. 11. The kinetic energy due to the motion of the structure is remarkably decreased when the rubber content is increased to 30%. Although smaller, a significant decrease is also noticed for the GRM90/10. The energy due to the rocking of the foundation slab is almost equal to that of the horizontal response of the structure when founded on the GRM100/0 and GRM90/10. However, when the rubber content is increased to 30%, the kinetic energy values due to the rocking component, although reduced compared to the GRM100/0, are remarkably more substantial than the corresponding values due to the other modes. Regarding the kinetic energy due to the translational motion of the foundation, it can be considered negligible for all the GSI-structure systems as it is also noticed from the very low values of the horizontal motion of the foundation (Fig. 8).

Conclusions
This study presented and analyzed the forced-vibration tests performed at the large-scale prototype structure of EuroProteas founded on gravel-rubber mixture (GRM) layers acting as a means of geotechnical seismic isolation (GSI). The main goal of this experimental program was to investigate and assess the effect of the rubber content of the GRM on the overall performance of the GSI-structure system. An eccentric mass shaker was installed at the roof of the structure to apply harmonic forces in a wide frequency and force amplitude range to the structure placed on three GRM layers with rubber content of 0%, 10%, and 30% per weight.
The experimental outcome showed that: • The presence of 10% rubber content in the GRM layer only slightly affects the structure's response, which is almost similar to the response of EuroProteas founded on the gravelly foundation soil layer. • The resonant frequency of the structure is shifted to a lower value demonstrating the decreased system's stiffness. In contrast, the recorded acceleration in almost the whole excitation frequency range is reduced due to the rubber's increased material damping. • An increase in the rubber content to 30% results in a significant decrease of both the horizontal translational acceleration due to the bending of the structure and the horizontal acceleration due to the rocking of the foundation. However, the not as significant reduction of the rocking component indicates the "rigid-body" response of the structure. • The amplitude of the developed base shear and base moment is remarkably reduced by over 50% and seems to be independent of the excitation frequency and the motion of the foundation when the structure is founded on the GRM70/30. • The kinetic energy of the GSI-structure system is significantly decreased when the rubber content is increased to 30%.

(c)
Exp 4 (d) Fig. 11 Amplitude of the kinetic energy components of the three GSI-structure systems at each excitation frequency in tests 3 and 4 • Before any application of GRM layers as GSI, all environmental issues related with potential leaching of chemicals from granulated rubber particles should be properly investigated.