This study aimed at investigating the seismic performance of hollow fibre-reinforced polymer (FRP) piles compared to traditional piles in soft clay deposits using shaking table tests. A laminar shear box with dimensions of 1.0 × 1.0 × 1.0 m was used to contain the soil medium and allow it to respond similarly to the free field. Two types of composite group piles (2 × 2) made of carbon fibre-reinforced polymer and glass fibre-reinforced polymer, along with aluminium piles, were manufactured and embedded as end-bearing piles within the soil. Several monitoring instruments were used to observe the soil–pile response under variety of ground motions adopted from the 2010 Val-des-Bois Earthquake in Canada and the 1995 Kobe Earthquake. Seismic response of the foundation was strongly dependent on the stiffness provided by the soil, which was a function of the degree of softening and intensity of shaking. The foundation motion of model piles was higher than those of the free field mainly because of strong kinematic soil–pile interaction. Among the model piles, FRP piles revealed lower foundation motion due to their lower flexural stiffness compared to aluminium piles. This could make FRP piles a remarkably viable option compared to conventional piles under seismic loading.
Introduction
Pile foundations are widely used to support high load structures in compressible and weak ground strata throughout the world. However, traditional piling materials may have limited lifespan in harsh environment because of the corrosion of steel and the degradation of concrete and timber. Fibre-reinforced polymers (FRPs) have recently been employed as pile materials in multiple forms including concrete filled FRP piles, hollow FRP tubes, and FRP sheet piles (Guades et al., 2012). For example, FRP piles were used as a substitute to precast concrete piles for Route 40 highway bridge over Nottaway River in Virginia (Fam et al., 2003). In latest projects such as Statue Liberty Dock, more than 200 timber piles were replaced by FRP piles due to deterioration of the original piles. Long Beach bridge is another project where reinforced concrete piles were replaced with FRP pipe piles in New York. FRP piles can withstand aggressive corrosive environments such as the splash zone in case of marine piles, where they would be ultimately submerged and exposed (Stapleman, 1997). Similarly, highway overhead sign poles and pedestrian bridge piers need to retain their integrity in cold regions where salt is used for de-icing the roads. While past studies may have discouraged the application of hollow FRPs due to drivability difficulty and possible buckling, these issues, however, can be resolved by considering concrete filled FRPs or use of them in softer ground conditions. Lu et al. (2022) conducted a preliminary study on the behaviour of FRP piles under axial and lateral loading in sandy soil. The study revealed that surface roughness, confined pressure, and relative density are key factors in determining the shear resistance of soil under axial loading. Furthermore, the bearing capacity and flexural stiffness of the piles under lateral loading are influenced by the type of FRP, pile diameter, and the aging of the material.
Several studies have been conducted to investigate the load transfer and flexural response of FRP piles to validate the application of FRP composites in construction industry (e.g. Fam, 2000; Giraldo and Rayhani, 2014; Hosseini and Rayhani, 2017; Iskander et al., 2001; Sakr et al., 2007). Shakir et al. (2016) investigated the lateral impact response of concrete filled steel tube with and without CFRP strengthening, and the results showed that Results of these analyses revealed that FRP piles may well be driven to reasonable capacities for load-bearing piles. It was also reported that driving forces may have minor influence on the flexural strength of CFFT piles. Concrete filled CFRP tubes with fibre orientation along the length of the pile were reported to demonstrate higher load carrying capacity whereas the unconfined concrete piles reached lower capacity (Murugan et al., 2017). In terms of flexural stiffens, Shao (2003) reported considerable ductility in concrete columns with glass FRP tubes as a replacement of the entire internal steel. Mirmiran and Shahawy (1996) also reported that the entire steel reinforcement in a concrete column or pile could be replaced by the FRP tube without affecting its load carrying capacity under static loading. Ibrahim et al. (2022) investigated the impact behaviour of composite reinforced concrete beams with pultruded I-GFRP. Their results demonstrated that concrete compressive strength is a significant factor in enhancing the impact resistance of composite specimens, particularly in comparison to those without GFRP I-beams. In another study, Dai et al. (2023) examined the effect of a glass fibre-reinforced concrete (GFRC) isolation layer on the dynamic response of tunnels through shaking table tests. The findings indicated that the GFRC layer effectively reduces the dynamic response in critical parts of the tunnel lining. In addition, Elmasry et al. (2014) conducted a numerical analysis of concrete-filled FRP piles subjected to seismic loading. Their study, which focused on a 5 × 3 group pile system with lumped mass at the pile cap, revealed that lower stiffness in the piles leads to larger deflections under seismic forces.
The lucrative capabilities of FRP composite piles encourages further investigation on performance of these piles under seismic loading in prone regions. However, there is lack of investigation to understand seismic response of these FRP piles in soft clay deposits. Past earthquakes have revealed complications due to the amplification of earthquake motion in soft clay deposits, and soil–pile–structure interaction (SPSI). Catastrophic structural damage and failure to pile foundations were observed in the Kobe (1995), Loma Prieta (1989), and Mexico City (1985) earthquakes, and clearly highlighted the role of local site conditions and soil–structure interaction (SSI) in modifying and altering the characteristics of strong motions (Mendoza and Romo, 1989; Nghiem, 2009). The challenge of SPSI in the seismic analysis and design of structures has become increasingly critical, as it may be inevitable to build structures at locations with less favourable geotechnical conditions in seismically active regions. Table 1 summarises past shaking table tests in 1g and centrifuge environment. Several of these model tests have been used to evaluate the impact of SSI in seismic behaviour of foundations (e.g. Feng et al., 2019; Torabi and Rayhani, 2014; Varghese et al., 2021).
Summary of past shaking table tests using various type of containers
| Shaking table type | References | Remarks |
|---|---|---|
| 1g Shaking Table | Tsukamoto et al. (2012); Massimino and Maugeri (2013); Hokmabadi (2014); Yang et al (2020); Wu et al. (2020); Chiou et al. (2021) | Foundation type: shallow foundation; single or group piles in different arrangements-conventional piles (e.g. concrete, steel, aluminium, polypropylene) |
| — | Soil medium: dry sand, loose sand, medium dense sand, and saturated sand, synthetic clay, glyben clay | |
| Centrifuge Shaking Table | Kagawa et al. (2004); Banerjee and Lee (2013); Liu et al. (2018); Zhu et al. (2021) | Foundation type: single or group pile |
| — | Soil medium: liquefiable sand, Slurry clay |
This paper presents the results of a series of 1g shaking table tests to study the seismic response of scaled end bearing hollow FRP tubes made of carbon and glass fibres in comparison to traditional piles. The impact of soil and piles’ response on foundation input motion were analysed for both FRP tubes compared to conventional aluminium group piles in synthetic soft clay by considering SPSI, which is significantly important on the performance-based design of foundations in seismic regions.
Shaking table experimental programme
Model laminar container
The shaking table tests were performed on a 1g shake table located at Civil and Environmental Engineering laboratory of Carleton University. In this shaker, a high-performance servo-hydraulic system is used to apply 1D horizontal base input displacement to soil–pile models and model containers in response to an applied input signal. The table can accommodate a maximum load of 10 tons and a maximum peak to peak displacement of 300 mm at the nominal operating frequency range of 0–50 Hz. A laminar shear container with inner dimensions of 1.0 m × 1.0 m with a height of 1.0 m was used to contain the soil–pile models and conduct the shaking tests (Figure 1).
A large metallic structure, featuring multiple horizontal ridges, is positioned on a yellow base within a laboratory setting. The structure serves a specific engineering purpose, as indicated by various hoses and cables that are connected to it. In the background, there are double doors painted yellow and blue, which lead into another area of the facility, displaying a clinical and industrial atmosphere. Additionally, a fire extinguisher is mounted on the wall nearby, emphasising the safety measures present in the workspace. The environment appears organised with various tools and equipment, suggesting ongoing research or testing activities.Shaking table in Civil Engineering Lab, Carleton University
A large metallic structure, featuring multiple horizontal ridges, is positioned on a yellow base within a laboratory setting. The structure serves a specific engineering purpose, as indicated by various hoses and cables that are connected to it. In the background, there are double doors painted yellow and blue, which lead into another area of the facility, displaying a clinical and industrial atmosphere. Additionally, a fire extinguisher is mounted on the wall nearby, emphasising the safety measures present in the workspace. The environment appears organised with various tools and equipment, suggesting ongoing research or testing activities.Shaking table in Civil Engineering Lab, Carleton University
Preparation of model piles and pile groups
The objective of the scale modelling technique for shaking test programme is to reach dynamic similarity, where the model and prototype experience homologous forces. Geometrical analysis is the key for scale model representation in this test programme. In 1 g scale modelling, where ρ is density, E is modulus of elasticity, a is acceleration, and g is gravitational acceleration, the dimensionless product a/g (i.e. Froude’s number) must be kept equal to unity indicating that the ratio of model to prototype specific stiffness is equal to the geometric scaling factor (Iai, 1989; Meymand, 1998; Moncarz and Krawinkler, 1981; Rocha, 1957). By defining scaling conditions for density and acceleration, the displacement, length, and time, the scale factors can all be expressed in terms of the geometric scaling factor (λ), and a complete set of dimensional scaling relations (ratio of prototype to model) can be driven for all model variables. Thus, a geometric scaling factor of 10 (prototype to model) was adopted to determine the model pile sizes and simulate a practical prototype of pile groups in field. The variables contributing to the primary modes of system response are summarised in Table 2 (e.g. Hokmabadi, 2014; Lee et al., 2012; Sulaeman, 2010; Turan and El Naggar, 2008). As mentioned, a scale factor of 1:10 was used to determine the model pile dimensions in order to derive a practical prototype of steel pipe piles in industry (Table 3). Scaling limitations imposed a maximum prototype pile length of 10 m, which provided a L/d ratio of 18.2, acceptable for an end-bearing pile. A series of model CFRP and GFRP group piles were manufactured with commercially available glass and carbon fibre-reinforced fabrics used for reinforcement and retrofit of structural members. The model FRP piles were manufactured by fitting the fibre sheets around a steel tube. The fibre sheets were saturated in epoxy prior to fitting around the steel tube. After 48 h of curing time, the piles were extracted from the steel tube and cut to the required length. Figure 2 illustrates the final configuration of model piles along with the fabricated group piles. Each group pile consisted of four (2 × 2) hollow CFRP or GFRP piles with 55 mm dia. and an L/d ratio of 18.2. Piles within each group were distanced at ≈200 mm from each other. The group piles were assembled using a pile cap made of a thick timber plate with a dimension of 260 mm × 260 mm and thickness of 76 mm. This cap section was fabricated with four holes, which piles are spaced at a centre-to-centre distance of about 3.6 times their diameter (s/d = 3.6) with respect to the boundary condition. Pile caps were loaded by lead weights of ≈9 kg which was tightened by passing a threaded steel rod to keep the structural load fixed. A similar pile group was also manufactured using four identical aluminium pipes with similar dimensions to simulate conventional piles since aluminium falls in the range of acceptable modulus, and it can be configured as a thin wall section to meet the EI criterion (i.e. flexural stiffness). Several past researchers have also considered aluminium as model pile in past shaking table tests (e.g. Azizkandi et al., 2020; Bao et al., 2012; Bathurst et al., 2007; Chen and Ueng, 2010; Finn and Gohl, 1992; Makris et al., 1997; Yao, 1980).
Scaling relations for shaking table tests in terms of geometric scaling factor (λ)
| Mass density | 1 | Acceleration | 1 | Length | λ |
|---|---|---|---|---|---|
| Stiffness | λ2 | Shear wave velocity | λ½ | Stress | λ— |
| Modulus | λ— | Time | λ½ | Strain | 1 |
| Displacement | λ— | Frequency | λ-½ | EI | λ5 |
Protype and model pile design
| Geometric properties | Prototype | Model pile (e.g. carbon) | Difference: % |
|---|---|---|---|
| Outside diameter (mm) | 550 | 55 | 0.0% |
| Length (m) | 10 | 1.0 | 0.0% |
| Wall thickness (mm) | 13 | 5.2 | — |
| L/D ratio | 18.2 | 18.2 | 0.0% |
| Modulus of elasticity (MPa) | 200 000 (steel) | 7000 (carbon) | — |
| EI (flexural stiffness, kN.m2) | 39 500 | 0.446 (λ5 scaled: 44 600) | 11% |
The images depict three different views of composite materials. The first image shows a cylindrical rod covered in a smooth white material, secured with a black band. The second image features a long, textured black bar, detailed with a grid pattern, indicating its construction method or material composition. The third image displays multiple dark rods standing upright, with wires and additional materials extending from them, indicating ongoing assembly or testing in a workshop environment. Tools and other materials are visible in the background, indicating an active workspace.(a) Fabrication of glass FRP pile, (b) fabricated carbon FRP pile, and (c) 2 × 2 group piles
The images depict three different views of composite materials. The first image shows a cylindrical rod covered in a smooth white material, secured with a black band. The second image features a long, textured black bar, detailed with a grid pattern, indicating its construction method or material composition. The third image displays multiple dark rods standing upright, with wires and additional materials extending from them, indicating ongoing assembly or testing in a workshop environment. Tools and other materials are visible in the background, indicating an active workspace.(a) Fabrication of glass FRP pile, (b) fabricated carbon FRP pile, and (c) 2 × 2 group piles
Model soil characteristics and preparation
The use of a reconstituted soil or field soil as the model soil is a common method in centrifuge testing, where a natural soil is mined and then mixed with water to form a slurry that can easily be placed in the test container. Consequently, the soil is then consolidated to reach the aimed strength profile. The consolidation process offers the advantage of fixing anisotropy and a stress history into the soil. Nevertheless, this technique is not feasible and attainable in 1g shaking table tests due to large scale of the container and also, the extensive time that is required for consolidation in a 1g environment. Most importantly, the reconstituted soil cannot satisfy the scale modelling requirement specifically the undrained shear strength and dynamic shear modulus which are obeyed in this study. Therefore, it is essential to develop a synthetic model soil as the soil medium to comply with the scaling law in shaking tests and, additionally, this can ease in control of saturation level and soil sample consistency/uniformity throughout all tests.
A synthetic clay mixture with proper scaled stiffness and strength properties was considered to provide characteristics of soft clay soil medium for the shaking table tests. Upon examining several mixing ratios, ultimately, the desired proportion of soil mix components kaolinite and bentonite powders with specified amount of 70% and 30%, respectively, along with addition of 100% water were used to simulate soft clay behaviour in model tests. The index properties of the model clay mixture were characterised through a series of laboratory tests. The Atterberg limits, determined in accordance with (ASTM, 2024a) yielded a liquid limit (LL) of ≈68% and a plastic limit (PL) of about 37%, specifying a high plasticity soft clay. The specific gravity (Gs) of the soil was found to be 2.74 (ASTM, 2024c), while the internal friction angle was estimated at around 21°. Furthermore, to determine particle size distribution of the clay mixture, hydrometer analysis conducted in accordance with ASTM (2024b) showed over 80% of the particles were finer than 2 µm, indicating the dominance of clay-sized particles in the soil model. Upon mixing the soil in small batches of 19 Litre buckets using hand drill mixer and also ensuring a homogenous mix sample, the synthetic clays were covered in 55 gal barrels and allowed to cure inside the sealed container. It was observed that this curing process allowed a more even distribution of water throughout the soil (in terms of consistency). The undrained shear strength of the mix at a cure age of five days was determined to be 6 kPa using vane shear test, and the shear wave velocity at the same cure age was estimated at about 35 m/s. This benchmark cure age was used as it was expected that the time from soil placement to time of testing and the time between tests would be approximately five days. The prototype values implied by these model properties with a geometric scaling factor of 10 are a static undrained shear strength of 55 kPa and a shear wave velocity of 110 m/s at a density of 1.43 Mg/m3.
Pile–soil models preparation and instrumentation
Upon placing the model piles within the container and mounting instruments, the soil mixture was then placed into the laminar shear container in several layers using drum lifter to provide easier handing in placing the synthetic clay; and upon completion of each test, the soil model was removed. The uniformity of the soil model was checked by performing vane shear tests to establish shear strength profiles before and after shaking tests. It should be noted that two shaking table tests, conducted following the same procedures, were completed. Test 1 compared the performance of CFRP piles versus aluminium piles, while Test 2 compared GFRP piles to aluminium piles. This approach provides a reliable method for comparing pile performance, as each test used at least one set of similar pile groups (i.e. aluminium) for reference. Several instruments and sensors including accelerometers, LVDTs and strain gauges were positioned within the soil and mounted on model piles and pile caps to record the test data. Two accelerometers were positioned at depths of 200 mm and 800 mm within the soil to record the free field motions in the soil. Similarly, two accelerometers were installed on a pile per each group at the same depths as the free field accelerometers to measure the pile–soil interaction effects. Also, one accelerometer was mounted on the pile cap of each group pile and head masses to detect possible translation and rocking motions. String potentiometers were also fastened to the laminar box at depths of 200 and 800 mm to record the displacement along the laminar box, and additionally 2 LVDTs were mounted on pile caps to record any potential flexibility induced by the piles during shaking. Furthermore, another LVDT was positioned vertically on top of the soil surface to record the soil settlement during each shake test. Lastly, a total of eight bending strain gauges were mounted on both sides of individual piles in each group to measure the strain and consequently the change in bending moment of the piles during the shaking tests. Upon preparation of the laminar shear container, the shaking system was set for employing earthquake excitation. Figure 3 illustrates the configuration of model piles instrumentations, dimensions and components of prepared pile–soil model. A2 and A3 are accelerometers attached to a pile per group near the base; A5 and A6 were mounted on the same pile at a depth close to the surface; A1 and A4 are located inside the soil at the same level as A2 and A5 to record free field motions in the soil (far from the piles); A7 and A8 were mounted on the pile caps.
The image presents a technical diagram and photographs of a test setup for assessing pile performance. The top diagram labels various components including a Vertical Lateral Variable Displacement Transducer, L V D T, Fibre Reinforced Polymer, F R P, pile, accelerometer, lead weight, pile cap, aluminium pile, and strain gauge, along with reference markers A 1 to A 8. The lower photograph displays the experimental apparatus with labelled elements such as a latex membrane, lead weight, pile cap, soil sample, vertical L V D T, horizontal L V D T, and aluminium laminar components. It depicts a large rectangular setup on a shaking table, measuring 1000 millimetres by 1245 millimetres, with notations on dimensions provided in the corner.Pile–soil model configuration and dimensions in laminar box
The image presents a technical diagram and photographs of a test setup for assessing pile performance. The top diagram labels various components including a Vertical Lateral Variable Displacement Transducer, L V D T, Fibre Reinforced Polymer, F R P, pile, accelerometer, lead weight, pile cap, aluminium pile, and strain gauge, along with reference markers A 1 to A 8. The lower photograph displays the experimental apparatus with labelled elements such as a latex membrane, lead weight, pile cap, soil sample, vertical L V D T, horizontal L V D T, and aluminium laminar components. It depicts a large rectangular setup on a shaking table, measuring 1000 millimetres by 1245 millimetres, with notations on dimensions provided in the corner.Pile–soil model configuration and dimensions in laminar box
Input earthquake motions
The testing models were subjected to real earthquake excitations with different peak accelerations and frequency contents. The input ground motions represent a set of earthquakes data recorded during the 2010 Val-des-Bois earthquake in eastern Canada, and the 1995 Kobe earthquake in the metropolitan area of western Japan. Each shaking model test was subjected to a total of six excitations at intensity levels of 50%, 100%, and 200% known as the Ottawa event and 5%, 10%, and 20% of the 1995 Kobe event. Three low level accelerations with a peak horizontal acceleration (PHA) of 0.02–0.08g were targeted to ensure that response remained in the elastic range. Three mid-range to high-range signals with a PHA of about 0.04–0.16g were also applied to impart an intermediate level excitation and induce non-linear site and pile response. To clarify, this manuscript presents two sets of shaking table tests. Test 1 compares the CFRP pile group to the aluminium pile group, while Test 2 compares the GFRP pile group to the aluminium pile group.
Acceleration time histories
Figure 4 shows sample acceleration times histories for the free field motion (A4) and corresponding pile motions for the CFRP (A5) and the conventional aluminium pile (A6) under the Kobe 20 (0.16 g) shaking event. The peak accelerations in terms of ‘g’ for each accelerometer for free field motions as well as both the pile groups are presented in Table 4. From the time histories of acceleration, the soil conditions had a significant effect on the site response, as expected. The peak acceleration within the soil profile showed a major amplification from the base excitation to A1 and A4 along the soil profile for weaker seismic events (e.g. O50 and O100). As the input acceleration at the base of the soil model increased, the free filed motion at both the soil levels (A1 and A4) showed a decreasing trend (de-amplification). As it will be discussed later, this reduction in free field motion could be related to higher strain induced in the soil and the soil non-linear response during stronger input motions (e.g. K20). However, the recorded peak accelerations within the soil (free field) for all applied excitations were significantly less than those measured on piles at the same depth. This difference in peak ground motions underscores the importance of considering soil–pile interaction (SPI) impacts on foundation input motions in seismic prone areas.
The image consists of four graphs arranged in a two by two layout, each displaying acceleration measured in g over time measured in seconds. The top left graph is titled free field surface K 20, showing oscillations in acceleration for a duration of 40 seconds. The top right graph is also titled free field surface K 20, similar in parameters to the first graph. The bottom left graph is titled C F R P surface K 20, exhibiting different oscillation patterns over the same timeframe, and the bottom right graph is labelled C F R P base K 20, indicating a distinct set of acceleration measurements over 40 seconds. All graphs use the same axes, with the Y-axis labelled acceleration g and the X-axis labelled time seconds.Acceleration time histories for free field and CFRP at K20 shaking
The image consists of four graphs arranged in a two by two layout, each displaying acceleration measured in g over time measured in seconds. The top left graph is titled free field surface K 20, showing oscillations in acceleration for a duration of 40 seconds. The top right graph is also titled free field surface K 20, similar in parameters to the first graph. The bottom left graph is titled C F R P surface K 20, exhibiting different oscillation patterns over the same timeframe, and the bottom right graph is labelled C F R P base K 20, indicating a distinct set of acceleration measurements over 40 seconds. All graphs use the same axes, with the Y-axis labelled acceleration g and the X-axis labelled time seconds.Acceleration time histories for free field and CFRP at K20 shaking
Summarised peak accelerations at free field, pile cap, and along the pile shafts for all events
| Model piles | Input events | Input motions: g | A1: g | A2: g | A3: g | A4: g | A5: g | A6: g | A7: g | A8: g |
|---|---|---|---|---|---|---|---|---|---|---|
| CFRP and AL (T1) | O50 | 0.020 | 0.032 | 0.061 | 0.076 | 0.024 | 0.059 | 0.081 | 0.045 | 0.098 |
| O100 | 0.041 | 0.071 | 0.100 | 0.121 | 0.067 | 0.094 | 0.129 | 0.097 | 0.140 | |
| O200 | 0.082 | 0.105 | 0.219 | 0.217 | 0.101 | 0.185 | 0.221 | 0.175 | 0.259 | |
| K5 | 0.041 | 0.075 | 0.121 | 0.157 | 0.076 | 0.110 | 0.169 | 0.101 | 0.179 | |
| K10 | 0.082 | 0.131 | 0.244 | 0.258 | 0.163 | 0.273 | 0.293 | 0.263 | 0.297 | |
| K20 | 0.164 | 0.216 | 0.455 | 0.490 | 0.252 | 0.448 | 0.539 | 0.419 | 0.590 | |
| GFRP and AL (T2) | O50 | 0.020 | 0.040 | 0.076 | 0.071 | 0.030 | 0.058 | 0.078 | 0.038 | 0.097 |
| O100 | 0.041 | 0.081 | 0.124 | 0.116 | 0.055 | 0.117 | 0.127 | 0.108 | 0.147 | |
| O200 | 0.082 | 0.104 | 0.197 | 0.224 | 0.108 | 0.181 | 0.235 | 0.176 | 0.264 | |
| K5 | 0.041 | 0.080 | 0.119 | 0.152 | 0.085 | 0.105 | 0.162 | 0.104 | 0.178 | |
| K10 | 0.082 | 0.138 | 0.201 | 0.271 | 0.160 | 0.272 | 0.291 | 0.273 | 0.303 | |
| K20 | 0.164 | 0.239 | 0.514 | 0.510 | 0.259 | 0.498 | 0.557 | 0.534 | 0.575 |
The distribution of acceleration varies along the shaft of all model group piles. All FRP piles experienced smaller acceleration close to their base and followed by a rise in peak acceleration near the surface in all shaking events. This behaviour was similarly observed in response motion of Aluminium piles along the shaft. Although, there was significantly higher response at the aluminium pile cap compared to the CFRP/GFRP group pile caps. This also means that superstructures rested on aluminium group piles would have experienced significantly greater acceleration compared to structures resting over FRP piles. This can be related to the higher stiffness of the traditional piles (i.e. aluminium) and hence higher deviation between the foundation and free field motions resulted from kinematic SPI. The higher peak accelerations in the aluminium pile cap may also result in generation of the base shear and induced moments that consequently can appear as displacement and perhaps rotation of the foundation relative to the free field.
Spectral analysis and discussion
Free field soil response
Soft clay deposits are recognised to specifically amplify earthquake motions in weaker seismic events, though very strong shaking (e.g. K20) may attenuate surface motions due to soil non-linearity and stiffness degradation. It can also be hypothesised that such motions, however, may induce shear failure in soft soil deposits.
The free field response of the soil was established based on results from the two accelerometers positioned inside the laminar shear container far from pile foundations. From the collected acceleration data (Table 4), spectral response was established to describe the frequency content of the earthquake motions measured on the piles and within the soil (Figure 5). Free field acceleration response profile revealed significantly higher acceleration within the soil profile compared to the input motion in relatively weaker events, which is signifying the amplification of motions along the soil profile. This behaviour can be observed in O100 and K5 events where peak input motion of 0.041g is clearly demonstrating amplification response of soil and elastic behaviour of soil at base and surface. As the intensity of input acceleration in the soil models increased (e.g. K20), the level of amplification in the soil profile decreased. This lower amplification for soft soils exposed to strong shakings could be related to the non-linear response of soft clay under strong excitations as well as stiffness degradation. It could also be associated with higher material damping which further reduces the response as the soil may have experienced shear failure limiting the stress that can be transmitted to the surface soil layers.
The image presents three graphs depicting spectral acceleration in terms of gravitational acceleration against period in seconds. The top graph corresponds to input motion O 100, the middle graph to input motion K 5, and the bottom graph to input motion K 20. Each graph features three lines, an orange solid line labelled F F close to base, a blue dashed line labelled F F beneath the surface, and a grey dashed line labelled input motion. The X-axis indicates the period in seconds, ranging from 0 to at least 10 seconds, while the Y-axis represents spectral acceleration from 0 to a maximum value indicative of the motion amplitude. The layout shows clear differentiation of responses across various input motions while sharing a consistent format, with labelled axes and legends for clarity.Response spectra for free field motions during the O100, K5, and K20 shaking events
The image presents three graphs depicting spectral acceleration in terms of gravitational acceleration against period in seconds. The top graph corresponds to input motion O 100, the middle graph to input motion K 5, and the bottom graph to input motion K 20. Each graph features three lines, an orange solid line labelled F F close to base, a blue dashed line labelled F F beneath the surface, and a grey dashed line labelled input motion. The X-axis indicates the period in seconds, ranging from 0 to at least 10 seconds, while the Y-axis represents spectral acceleration from 0 to a maximum value indicative of the motion amplitude. The layout shows clear differentiation of responses across various input motions while sharing a consistent format, with labelled axes and legends for clarity.Response spectra for free field motions during the O100, K5, and K20 shaking events
Pile–soil interaction analysis
Previous studies have shown that the kinematic SSI could deviate the foundation input motions from the free field motions and, hence, affect seismic response of superstructures. This interaction can be defined as frequency dependant transfer function involving both the free field motion and the base-slab motion while the base and structure are assumed massless. As a result, this may cause wave inclination or foundation rocking due to base-slab averaging and embedment depth (Stewart et al., 1999). This impact can be explained by assessing the kinematic SPI on foundation input motions. The spectral acceleration closer to ground surface for all the model piles were compared with the spectral response of the free field motions at the same depth. This pile/soil response spectral ratio (RRS) can evidently illustrate the divergence of the pile behaviour throughout the shaft as the result of its adhesion and material characteristics which affect the pile foundation motions. The results of spectral ratio are plotted for all the model piles measured at the O100, K5, and K20 events in Figure 6.
The image features a series of four line graphs showcasing the relationship between Relative Residual Strain, R R S, and period measured in seconds. The top row contains graphs for O 100 on the left and O 100 close to the base on the right, while the bottom row includes graphs for K 5 on the left and K 20 on the right. Each graph uses different line styles to represent three materials, Carbon Fibre Reinforced Polymer, C F R P, Glass Fibre Reinforced Polymer, G F R P, and Aluminium, A L. The X-axis denotes the period ranging from 0.01 to 100 seconds, while the Y-axis represents R R S, showing values up to approximately 2.5. The legends indicate which line corresponds to each material and condition. Each graph displays patterns of peaks for different material types, indicating variability in response based on material and position.Pile–soil response spectral ratio (RRS) for all models during O100, K5, and K20 shaking events
The image features a series of four line graphs showcasing the relationship between Relative Residual Strain, R R S, and period measured in seconds. The top row contains graphs for O 100 on the left and O 100 close to the base on the right, while the bottom row includes graphs for K 5 on the left and K 20 on the right. Each graph uses different line styles to represent three materials, Carbon Fibre Reinforced Polymer, C F R P, Glass Fibre Reinforced Polymer, G F R P, and Aluminium, A L. The X-axis denotes the period ranging from 0.01 to 100 seconds, while the Y-axis represents R R S, showing values up to approximately 2.5. The legends indicate which line corresponds to each material and condition. Each graph displays patterns of peaks for different material types, indicating variability in response based on material and position.Pile–soil response spectral ratio (RRS) for all models during O100, K5, and K20 shaking events
The spectral acceleration recorded on all the model piles were shown to be significantly higher (almost twice) than those recorded at the same level within the soil (free field motion). The ratio of foundation to free field motions were substantially higher at the frequency range of 1–20 Hz. The period of vibration was also shown to be longer for the foundation motions compared to the free field records. This increase in spectral acceleration for the pile models and change in frequency content of the seismic waves are clearly associated with the difference in flexural stiffness of the pile materials compared to that of the soil and, hence, significant kinematic interaction between the model piles and soil. This result can be attributed to the inability of the piles to follow the wave motion due to their flexural rigidity. Therefore, during earthquake loading, seismic waves are reflected as the foundation is being stressed by seismic loading and the foundation will experience oscillatory motion that differs substantially from the free field response.
An equivalent comparison can be made for the different model piles to evaluate the impact of pile materials on the pile–soil interaction and, in turn, the seismic response of different foundation systems using the spectral acceleration response of piles. Figure 7 shows similar linear response for FRP piles and aluminium pile at O100 and K5 excitations near the surface, while larger amplification was recorded at the base of aluminium pile as a result of the kinematic forces at the motion direction close to the base. Nevertheless, the frequency of shaking was substantially filtered at the surface and alternatively continued at linear state. This outcome is correspondingly differed at the stronger motion of K20, where higher RRS values were observed with drastically greater amplifications at the same frequency level of 10–100 and 1–10 Hz at both locations (BS and CB).
The image features a grid of six graphs, each plotting spectral acceleration in units of gravitational acceleration against period in seconds. The horizontal X-axis represents the period, extending from 0 to 10 seconds, while the vertical Y-axis indicates spectral acceleration. Each graph compares three materials, Carbon Fibre Reinforced Polymer, C F R P, Glass Fibre Reinforced Polymer, G F R P, and Aluminium, A L, distinguished by different line styles. The rows include graphs labelled O 100 and K 5 at the top, and K 20 at the bottom, with each label indicating a specific condition being tested. The left side graphs represent surface conditions, while the right side graphs depict base conditions, showing how spectral acceleration varies across these conditions at the specified periods.Spectral acceleration for model piles at O100, K5, and K20 excitations
The image features a grid of six graphs, each plotting spectral acceleration in units of gravitational acceleration against period in seconds. The horizontal X-axis represents the period, extending from 0 to 10 seconds, while the vertical Y-axis indicates spectral acceleration. Each graph compares three materials, Carbon Fibre Reinforced Polymer, C F R P, Glass Fibre Reinforced Polymer, G F R P, and Aluminium, A L, distinguished by different line styles. The rows include graphs labelled O 100 and K 5 at the top, and K 20 at the bottom, with each label indicating a specific condition being tested. The left side graphs represent surface conditions, while the right side graphs depict base conditions, showing how spectral acceleration varies across these conditions at the specified periods.Spectral acceleration for model piles at O100, K5, and K20 excitations
The applied ground motion at the base of aluminium piles seems to be almost twice the free field response at the same level. This means the response of aluminium piles were dominated at the base by kinematic forces due to their higher flexural stiffness as they have suffered larger deformation and magnitude of acceleration at the toe. Comparing response of both FRP piles at Kobe20 (i.e. CFRP and GFRP) shows that the CFRP piles experienced slightly higher levels of acceleration and led to higher amplifications compared to the GFRP piles near the surface. Therefore, it seems that the response of CFRP piles could be an indication of more dominant interaction than the GFRP. As stated earlier, the elastic modulus (E) of the CFRP materials (as an indicator of stiffness and rigidity) was higher than the glass fibres at around 7.0 GPa compared to the GFRP materials at 5.8 GPa. This difference in rigidity could have led to higher motions transferred through the CFRP piles and hence higher-level amplification of foundation input motions affected the overall response as also observed in traditional piles. Nevertheless, the frequency content of the amplified motions appears to be consistent for all model pile groups. Thus, the SPI has not deviated the frequency content of seismic motions recorded for the pile systems and only amplified the input motion.
Impact of SPI on pile cap motion
The characteristic of earthquake motions at the structure is affected by the response of the underlying soil through soil amplification and SSI. This behaviour can further be examined by the influence of pile characteristics on seismic response of structures (or vice versa). The spectral response of the pile cap was compared for all model piles using the collected data from attached accelerometer on the superstructure of each pile group. Figure 8 demonstrates the response spectra measured on accelerometers placed over the pile caps under three seismic excitations (O100, K5, and K20). Both FRP pile caps demonstrated rather similar amplification, and it was much lower than the level of amplification experienced on the aluminium pile cap and its structure. This reveals that the composite material characteristics including their flexural rigidity could have critical impact on the response of the pile along the shaft, pile cap, and the structure. During the K5 event, the high-frequency portion of input motions was significantly filtered at surface of the piles which is due to the dynamic interaction of the surrounding soil with the piles. Interestingly, this filtering is not observed for the acceleration recorded at the aluminium pile cap, but the cap experienced almost 40% increase in acceleration from the response of input motion.
The image features three graphs displaying spectral acceleration measured in gravitational acceleration against period measured in seconds. The top graph labelled O 100 presents results for Carbon Fibre Reinforced Polymer, C F R P, Glass Fibre Reinforced Polymer, G F R P, and Aluminium, A L, at the cap, along with a dashed line representing data for F F beneath the surface, covering a period range from 0.01 to 100 seconds. The middle graph for K 5 displays similar comparisons, while the bottom graph for K 20 follows the same structure. Each graph employs different line styles to distinguish between the material categories, and the axes are clearly labelled. The data in each graph flows from left to right, with varying peaks and trends for the different materials.Pile cap response spectra at K5 and K20 shaking events for all model piles
The image features three graphs displaying spectral acceleration measured in gravitational acceleration against period measured in seconds. The top graph labelled O 100 presents results for Carbon Fibre Reinforced Polymer, C F R P, Glass Fibre Reinforced Polymer, G F R P, and Aluminium, A L, at the cap, along with a dashed line representing data for F F beneath the surface, covering a period range from 0.01 to 100 seconds. The middle graph for K 5 displays similar comparisons, while the bottom graph for K 20 follows the same structure. Each graph employs different line styles to distinguish between the material categories, and the axes are clearly labelled. The data in each graph flows from left to right, with varying peaks and trends for the different materials.Pile cap response spectra at K5 and K20 shaking events for all model piles
Greater flexibility of FRP piles and the difference in the interconnection of FRP piles to the cap may split the kinematic forces along the pile shaft through the cap and its structure, whereas traditional aluminium piles have reflected the induced kinematic forces to the cap and structure which this can result on damage or failure of structural members including partitions of any structure. Consequently, FRP composites may be a suitable option compared to traditional piles especially in seismic prone areas and offshore projects due to its larger flexibility and of lower intensification of foundation input motions.
Effect of shaking on pile–soil interface
During seismic loading, pile–soil gaps may develop between the soil and the pile near the ground surface which could result in a reduction of soil–pile interface strength. This condition may further affect the overall SPSI and radiation damping due to the higher stiffness contrast between the soil and piles. In addition, piles may be exposed to foundation input motion at much higher frequencies than the surrounding soil, and soil–pile contact may force the soil to also oscillate at these high frequencies, resulting in the transmission of high frequency energy away from the pile into the surrounding soil (Meymand, 1998). On the other hand, since radiation damping occurs at high frequencies and low levels of soil damping, and it cannot propagate through the established gaps launched between the pile and soil, it can further affect the response of the pile cap due to the critical impact of radiation damping. Consequently, it dominates the structure’s inertial forces and reduces the overall effects of the spectral de-amplification.
It is essential to comprehend that upon completion of K20, model piles have experienced gapping (Figure 9). This shaking induced gapping could lead to strength degradation and contribute to a partial loss of soil–pile adhesion. This may have played a role in the level of acceleration recorded on piles (close to surface) and pile caps through reduction of radiation damping and contributing to amplification of foundation input motions. Although all pile caps were loaded at constant light head mass, however at strong shaking of K20, a small partial inertial interaction could also have caused subsequent near surface soil–pile gapping and softened zones. The soil–pile gap was found to be slightly wider for the aluminium piles compared to the FRP model piles at K20 event. Therefore, it is probable that the response of aluminium pile cap was related to the gapping at the surface where potentially significant amplification was felt at the head/cap due to loss of soil–pile interface stiffness.
The image consists of two close up photographs of clay structures depicting two columns, each partially embedded in a textured surface of clay. In the left image, a red arrow points horizontally with a label indicating a measurement of 0.16 grams. Below the columns, another red circle encases a label indicating a width of approximately 12 millimetres marked by horizontal arrows. The right image presents a similar view, focusing on the clay columns and highlighting the area around them but without any measurement indications. Each column shows variances in texture, and both images are framed in black, creating a unified presentation.Development of gapping around model piles: aluminium (left) and CFRP (right)
The image consists of two close up photographs of clay structures depicting two columns, each partially embedded in a textured surface of clay. In the left image, a red arrow points horizontally with a label indicating a measurement of 0.16 grams. Below the columns, another red circle encases a label indicating a width of approximately 12 millimetres marked by horizontal arrows. The right image presents a similar view, focusing on the clay columns and highlighting the area around them but without any measurement indications. Each column shows variances in texture, and both images are framed in black, creating a unified presentation.Development of gapping around model piles: aluminium (left) and CFRP (right)
Pile failures due to buckling is a crucial factor that can limit the use of hollow FRPs in many applications, however, the buckling of FRP pipes can be controlled by the value of slenderness ratio (L/d) as well as using concrete-filled FRPs. Based on previous experimental studies (e.g. Abdulazeez et al., 2019; Feng et al., 2019; Prabhakar et al., 2019), long CFRP and GFRP pipe piles can behave with significant variation under axial and lateral movement. FRP pipes of smaller slenderness ratio can be compressed to crack at the ultimate compressive strain initiated by lateral deformation after buckling. However, the pipe piles with greater slenderness ratio may buckle in elastic zone followed by failure with oversize deformation.
Bending moment of pile shaft and pile cap deflection
The lateral capacity of a rigid pile is typically controlled by the resistance provided by the soil and its response along the pile shaft, which is dominated by the failure of soil adjacent to the pile. Alternatively, the failure mechanism of a flexible pile involves the formation of a plastic hinge within the pile, and its lateral capacity is influenced by both the flexural stiffness (EI) of the pile and the soil resistance along the deflected length of the pile. Therefore, it is essential to verify the straining actions occurring in piles during shaking which can be captured by mounting several strain gauges to record varying strains during shaking experiments. It should be noted that these gauges only measure strains at discrete points along the pile. Therefore, a curve fitting method must be employed to establish the straining pattern along the entire length of the pile. The bending moment envelope is defined by the absolute peak strain at each gauge during the excitation. This means that the result is not equivalent to actual bending moment diagram at the time step when the peak strain is recorded. As stated by the theory of elasticity and Hooke’s law, the generated moment in the pile section is a function of the recorded strain in the strain gauges (Timoshenko, 1940), referring to the following expression:
where Ep is the modulus of elasticity, Ip is defined as the cross-sectional moment of inertia, D is the outer diameter of the pile, and ε is defined as the recorded strain in the strain gauges.
The cross-sectional moment of inertia can be assessed using the following expression:
where r0 is the outer radius and ri is the inner radius of the pile.
Figure 10 illustrates bending moment envelopes along the pile shaft for both the FRP pile groups and aluminium piles at shaking events of K5 and K20. As expected, the maximum bending moment in all pile models occurred at the base. It is critical to explain that the trends can clearly illustrate no domination by inertial forces from the superstructure masses due to lightly loaded piles, therefore, pile response was influenced by the kinematic forces from the soil that resulted the larger stresses in the piles as the intensity of the earthquake increased.
The image features two graphs representing bending moment in newton millimetres against depth in millimetres. The left graph displays four curved lines for different material types, identified by specific markers, C F R P K 5 shown as a solid red line with diamond markers, G F R P K 5 shown as a dashed green line with square markers, and A L K 5 shown as a dashed blue line with triangular markers. The corresponding depth values are distributed from 0 to negative 900 millimetres on the Y-axis. The right graph similarly presents data for C F R P K 20, G F R P K 20, and A L K 20, using different labelled lines and markers. Both graphs are clearly labelled, with bending moment on the X-axis and depth on the Y-axis, allowing comparison of bending moment across varying depths for the specified materials.Recorded bending moment distribution along the piles at K5 and K20
The image features two graphs representing bending moment in newton millimetres against depth in millimetres. The left graph displays four curved lines for different material types, identified by specific markers, C F R P K 5 shown as a solid red line with diamond markers, G F R P K 5 shown as a dashed green line with square markers, and A L K 5 shown as a dashed blue line with triangular markers. The corresponding depth values are distributed from 0 to negative 900 millimetres on the Y-axis. The right graph similarly presents data for C F R P K 20, G F R P K 20, and A L K 20, using different labelled lines and markers. Both graphs are clearly labelled, with bending moment on the X-axis and depth on the Y-axis, allowing comparison of bending moment across varying depths for the specified materials.Recorded bending moment distribution along the piles at K5 and K20
The distribution of the moment amplitude along the FRP pile shafts demonstrate similar trend with slightly greater bending moments on CFRP piles compared to GFRP piles along the shaft due to higher flexural stiffness (EI). However., traditional aluminium piles have reached much larger moments close to the base up to the surface. This may suggest that the bending moment (i.e. especially close to the base) has been dominated by intensification of kinematic interaction through pile–soil models. To further validate the interaction effects, Figure 11 shows lateral deflections of the pile caps for all model piles during K20 shaking event. The lateral deflection indicates any movement at the cap; and the result revealed similar lateral deflections for the CFRP, and aluminium models pile caps, while GFRP experienced slightly less deflection and residual displacement as the result of strong shaking (K20). This can be great indicator of wave propagation through different materials and additionally it may be attributed to inertial interaction that impacted the response of the end-bearing aluminium group piles near the surface and its pile head.
The graph presents the displacement measured in metres of different types of pile caps, G F R P, C F R P, and A L, over a time span of 0 to 50 seconds. The horizontal X-axis represents time in seconds, while the vertical Y-axis indicates displacement values ranging from negative 0.1 to positive 0.1 metres. There are three distinct lines representing the displacement for each pile cap type. The G F R P data is represented by a dotted blue line, the C F R P data by a solid red line, and the A L data by a dashed green line. This visualisation allows comparison of the displacement behaviour of these different materials over the specified time period.Pile head/cap deflection at K20
The graph presents the displacement measured in metres of different types of pile caps, G F R P, C F R P, and A L, over a time span of 0 to 50 seconds. The horizontal X-axis represents time in seconds, while the vertical Y-axis indicates displacement values ranging from negative 0.1 to positive 0.1 metres. There are three distinct lines representing the displacement for each pile cap type. The G F R P data is represented by a dotted blue line, the C F R P data by a solid red line, and the A L data by a dashed green line. This visualisation allows comparison of the displacement behaviour of these different materials over the specified time period.Pile head/cap deflection at K20
Conclusions
A series of shake table tests were conducted to understand the seismic response of hollow FRP piles (carbon and GFRP) in cohesive soils. The SPI in soft clay deposits is a complex process involving inertial interaction between structure and pile foundation, kinematic interaction between piles and soils, soil amplification and the non-linear response of soils during strong earthquake motions. Results discussed in this paper have shown the following conclusions:
Soft clay deposits were shown to particularly amplify shaking, though surface motions were attenuated during very strong shakings due to soil non-linearity and stiffness degradation.
The earthquake characteristics, soil properties, and SPSI were found as critical factors that can impact the seismic excitation experienced by the pile cap and the structures.
The seismic response of the soil and kinematic pile–soil interaction deviated the response of FRP piles from that of free-field motions at a period of 0.05–1 s, corresponding to the frequency range of 1–20 Hz.
The frequency content of the amplified motions appeared to be consistent among the model pile groups. Therefore, the SPI has not deviated the frequency range of seismic motions recorded for both the pile systems and only amplified the input motion.
Comparing the response of the model piles, the result illustrated that CFRP piles have reached higher motion response in base which in turn induced larger bending moment and hence, lower response closer to surface. Alternatively, aluminium piles have reflected significant kinematic forces and seismic waves on the cap and structure. This can further underscore the importance of pile ductility and hence the kinematic pile–soil interaction on seismic response of pile foundations in seismic prone areas.
It is essential to note that the weight of the superstructure was not significantly large, therefore gapping may not solely be related on inertia, the impact of heavy superstructure load on pile gapping should be considered.
Higher flexibility of these FRP piles under seismic loading would be advantages for the design of pile foundations in liquefiable and soft soils. Higher kinematic interaction between existing conventional piles and soft ground materials could lead to significantly higher foundation motion and potential structural damage in seismic prone area. This low rigidity of FRP piles along with their higher durability in corrosive environment would make them a more feasible foundation option in these applications. However, there are a few inquiries which it needs further studies experimentally and numerically such as SPI effect within pile group size configuration and higher superstructure loads.
It is important to note that hollow FRP piles may not be suitable for all applications due to the risk of buckling failure. However, this issue can be addressed by modifying the pile’s cross-section and increasing the wall thickness. In addition, filling the FRP pile with concrete can significantly enhance its stiffness and overall mechanical properties.
Acknowledgements
This study was financially supported by Natural Sciences and Engineering Research Council of Canada (NSERC).
