Dynamic Responses of Block Type Quay Walls Under Cyclic Loading

The purpose of this research is to study the dynamic responses of gravity quay walls with block type consisting of “three blocks” experimentally. For this purpose, 1g shaking table tests were conducted under different cyclic loadings for two different saturated granular backfill materials (Soil 1 and Soil 2). In this study, Dn50 of Soil 1 and Soil 2 are selected as 2.2 cm and 1.0 cm, respectively. During the experiments, accelerations, soil pressures and displacements were measured for each block. Test results pointed out that Soil 2 caused more damage on structures. The measurements for displacement and tilting of each block were discussed in view of “acceptable level of damage in performance-based design” given in PIANC (2001). The result of the study showed that the definitions of damaged levels given in PIANC (2001) were reliable for using in performance-based methods for seismic design of block type quay walls.


Introduction
Quay walls are earth-retaining structures for the mooring of ships. Since the need for huge amount of investments and the large loads on the structure, which will increase in the future because of the trade actions, the design and construction of a quay wall becomes more important and complicated day by day. Block type wall is the simplest type of gravity quay wall, which consists of blocks of concrete or natural stone placed from the waterside on a foundation consisting of a layer of gravel or crushed stone on top of each other. Blocks maintain their stability through friction between themselves and between the bottom block and the seabed (Karakuş, 2007).
The probability of a large earthquake might be a rare event near a major city, but the economic and social impact of this natural disaster can be devastating (PIANC, 2001). In the past, it was reported that many gravity type structures were damaged in large earthquakes. Sumer et al. (2007) prepared an inventory including the observations of damage to marine structures caused by liquefaction in Kocaeli, Turkey earthquake on August 17, 1999. They reported that backfills behind quay walls and sheet-piled structures were liquefied, quay walls and sheet-piled structures had sea-wards movements, storage tanks were tilted and in some regions, seabed settlements were observed and also some of marine structures settled. Furthermore, in Derince Port, the block type quay wall was displaced seaward by 0.1−0.50 m and backfill settled by 0.5−1 m. In Tuzla Port, the block type quay wall moved seaward by 40 cm and settlement of backfill was about 10 cm. While liquefaction was observed on the backfill behind the block type quay wall in Derince Port, there was no observation of liquefaction on the backfill behind the block type quay wall in Tuzla Port. In 1989, in Kalamata Earthquake, block-type quay walls dislocated seaward (0.15±0.05 m) and tilted (4°−5°) (PIANC, 2001). In 1989 Chenoa earthquake, block type quay walls in Algiers Ports displaced seaward by 0.5 m and settled by 0.3 m (PIANC).
Although there are a lot of studies about dynamic response of gravity walls, mostly on caisson type, the number of studies about dynamic response of block type quay walls is limited in literature. Sadrekarimi et al. (2008) investigated both the static and dynamic behavior of hunchbacked gravity quay wall by using the 1g shaking table tests for various base accelerations on models with different subsoil relative densities. According to experiment results, it is understood that negative back-slope (the elevations below the break) reduces the lateral earth pressure, while positive back-slope (elevations above the break) increases the lateral earth pressure. They presented that relative density of seabed has a significant effect on the movement of the wall; when the seabed was softer, the wall moved more with bigger acceleration. Sadrekarimi (2011) reviewed the seismic performance of reduced-scale broken-back quay walls in 1g shaking table model experiments (Sadrekarimi, 2004) and compared it with the lateral displacements estimated using an improved sliding block approach. Therefore, implementing appropriate liquefaction remediation measures was recommended to improve the seismic performance of the quay wall. Cihan et al. (2015) experimentally and numerically investigated the stability of block type quay wall which consists of two concrete blocks. By using experiment results, the static friction coefficients between the rubble-block and block-block were determined and compared with the recommended friction coefficients in standards. PLAXIS V8.2 software program was used for numerical study to determine the material properties. Alielahi and Moghadam (2017) studied to assess the vulnerability of broken-back block quay walls during seismic events and they developed the fragility curves for block type quay walls. For this purpose, a numerical model was developed by FLAC software and this numerical model was validated with results of the shaking table tests that were conducted on two types of block type quay walls of Pars petrochemical port project. Fragility curves were produced for block type quay wall from the results of numerical model. Yuksel et al. (2017b) studied seismic response of hunched back quay wall experimentally and two different quay wall models were taken into consideration. Main difference between the models was the breaking point of the hunch. They found that model with lower breaking point exhibited a more stable performance in terms of displacement, settlement, and tilting. Dakoulas et al. (2018) investigated the seismic performance of the quay wall at the central wharf of Volos port in Greece by numerically using a comprehensive constitutive model for cyclic loading of granular materials. These granular materials were used by Dakoulas and Gazetas (2008) to understand the effects of the uncertainty regarding the spatial variability of the properties of the sandy gravel and rockfill materials at the backfill and foundation. The results showed that for constant average relative density, the effects of the frequency characteristics of the earthquake excitation were more important than the effects of its spatial variability. Therefore, it was revealed that a foundation of high relative density caused decline on settlements and rotation of the wall. It was found that negative excess pore water pressures developed just behind the wall generally.
The scope of this study is to investigate the dynamic response of block type quay wall (three blocks) considering basically the accelerations, soil pressures and displacements for different cyclic loading for different backfill materials (Soil 1 and Soil 2), using 1g shaking table tests to form the basis for the performance based design for block type quay walls under dynamic loads. In the studies mentioned before, the response of hunchbacked quay walls under earthquake was investigated. These quay walls consist of several concrete blocks which are not in the same dimensions. It should be investigated whether the soil pressures arising from backfill on these types of structures are different from the soil pressures on the vertical faced structures consisting of blocks of the same length. Although, the total heights of the structures in previous experimental studies in the literature (Sadrekarimi et al., 2008;Yuksel et al., 2017b) are similar with this study, each block's height is smaller than the block heights used in this study. The reason of using three blocks in this study is thought that the individual behaviors (displacement, acceleration, etc.) of the blocks can be investigated more precisely.

Experimental set-up
A series of 1g shaking table tests were carried out to investigate the dynamic response of block type quay walls at Hydraulics and Coastal and Harbor Lab., Civil Engineering Faculty at Yıldız Technical University. The one-degree of freedom 1g shaking table had deck dimensions of 400 cm× 100 cm×100 cm with a 4-ton load capacity. It was driven by a 100-kN capacity hydraulic actuator with operator controlling and PC software. The frequency range of shaking table was between 0 and 50 Hz, and the maximum horizontal displacement of shaking tank was ±20 mm. Shaking table was one dimensional in its motion, thus only longitudinal components of accelerations were obtained by omitting the transverse and vertical components. The experimental set-ups are shown in Fig. 1. The blocks were placed on the shaking table between dummies to simulate side effects as in prototype. Two different granular backfill materials (Soil 1 and Soil 2), which have different nominal diameters, were used as backfill materials. Physical characteristics of backfill materials and of blocks are presented in Table 1 and Table 2, respectively. The particle size distribution of Soil 1 and Soil 2 are plotted in Fig. 2.
Soil pressure cells (SP) were located on the rear face of the wall. The locations of SP1, SP2, SP3, SP4 were determined as 55 cm, 35 cm, 25 cm and 5 cm and these dimensions are given from top of the quay wall, respectively. SP2 and SP3 were placed on Block 2 to obtain variation of soil pressures on the middle of the wall. The locations of P1, P2, P3 from top surface of the wall were 2 cm, 22 cm, and 42 cm, respectively. Position sensors were placed near the top surface of each wall. Acc2, Acc3, Acc4 were placed at a depth of 5 cm, 25 cm and 45 cm from the top surface of the wall respectively. On the other hand, for all tests, one of the accelerometers (Acc1) was located on the outer face of model tank underneath the model to record the base accelerations of the cyclic loads.
The raining system was used to prepare the backfill behind the model wall to obtain the same relative density for each test. In Fig. 3, raining system and shaking table are shown. Relative density of the Soil 1 and Soil 2 were computed as 70%.
The following steps were followed to perform experiments.
(1) Frames were constructed within the 1g shaking table facility to divide the system into two parts to facilitate the use of only half of the 1g shaking table. The model box length was 4 m, height was 1 m, and width was 0.37 m. The pluviated length of the backfill was 1.2 m in the model box.
(2) The blocks were placed on the shaking table between dummies. Dummies were used to give the side effects from the adjacent blocks as under the prototype conditions.
(3) Soil pressure sensors were located on the blocks.
(4) Backfill materials (Soil 1 and Soil 2) were pluviated behind the blocks and dummies using raining system. Porosity, initial velocity of soil particles, deposition height and falling height are the major factors affecting the relative density of the soil particles prepared by raining method. Pluviation height and the gap of damper were chosen as 65 cm and 7 cm and were kept constant by lifting the sieve at each stage during backfilling (5) Accelerometers and displacement sensors were located on the blocks (6) The system was filled with water before starting the experiments and the absorbers were used to prevent the end effects of reflections caused by dynamic loading.
(7) The dynamic loading duration was selected as 30 s, long enough to observe the dynamic response of block(s), and it was kept constant in all tests.
Scale ratio was determined as 1/10. The soil density was kept the same for both the prototype and model to simplify scaling parameters in the 1 g model testing. In PIANC (2001), the weights of the backfills used in San Antonia Port, block type quay wall, Chile and in Kalamata Port, block type quay walls were 5-100 kg and 1-50 kg, respectively. In practice, it can be said that the weight of the backfill material can be between 1 kg and 100 kg. This means that nominal diameter of the backfill is 7 cm < D n50 < 34 cm. Scale factor of model in this study was 1/10. Therefore, D n50 of Soil 1 and Soil 2 were 22 cm and 10 cm, respectively and the block dimensions were 3 m×2 m×2.5 m in prototype.   Hulya Karakus CIHAN, Kubilay CIHAN China Ocean Eng., 2021, Vol. 35, No. 2, P. 281-290 283 The recommended scaling factors by Iai (1989) and being also used in this experiment were shown in Table 3. The experiments were conducted for two different models: Case 1 and Case 2. While Case 1 included three concrete blocks which had the same dimensions and coarser backfill material (Soil 1), Case 2 included three concrete blocks which had the same dimensions and finer backfill material (Soil 2). For each series, Case 1 and Case 2 conditions were tested under cyclic base motions in Table 4, and measured acceleration values versus frequencies are shown for Case 1 and Case 2.
During shaking tests, four type sensors, accelerometers (PCB Piezotronics, IMC 626B13), soil pressure sensors (Tokyo Sokki Kenkyujo, KDE-200-KPA), position transducers (UniMeasure, HX-PA series HX-PA-SS-L5M) and pore pressure sensors (DRUCK-PDCR81), were used to record the performances of models. IMI 626B13 ICP model accelerometers were used during the shaking table tests to record accelerations of tank and block(s). These accelerometers have 1000 mV/g sensitivity, ±5g measurement range, 0.2−6000 Hz frequency range, ±1% amplitude linearity. KDG-200-KPA type soil pressure sensors were used to measure soil pressure, and they are stainless-steel gauge which has outside diameter of 100 mm. Its weight and capacity are 1.2 kg and 200 kPa, respectively. Position transducers (UniMeasure, HX-PA series HX-PA-SS-L5M) measurement ranges are 0−500 mm, and they were used to measure instantaneous blocks displacements during the tests. Their wire rope diameters and weights are 0.4 mm and 0.9 kg, respectively. DRUCK-PDCR81 type pore pressure cells were used to obtain pore pressure measurements. Their dynamic measurement range is 0−100 kPa. An IMC system type Spartan-2 which had 32 analog inputs was used as the simultaneous data recorder. In models (Case 1 and Case 2), four accelerometers, four soil pressures sensors, four position transducers and two pore pressure sensors were used as shown in Fig. 1. While the sampling rate of accelerometer was chosen as 500 Hz, sampling rate of pore pressure sensor, position transducer and soil pressures sensor were chosen 10 Hz. To simulate the damaging effects of seismic waves with different frequencies and accelerations on the block type quay wall, almost sinusoidal waves with frequencies of 3 Hz, 4 Hz, 5 Hz and 6 Hz, were selected as the seismic loading. Similar input motions of sine waves were employed by Cihan et al. (2015) and Yuksel et al. (2017a). The cyclic-input base acceleration amplitude varied between 0.12g and 0.58g. Typical acceleration measurements with frequency of 4 Hz for Case 1 are presented as acceleration (g) versus time (second) for the accelerometers placed at the base of the set-up (Acc1) and at Block 1 (Acc2), Block 2 (Acc3), and Block 3 (Acc4) in Fig. 4 (Karakus, 2013).

Acceleration
According to experimental results, it was seen that increasing in frequency which also means that increasing in number of cycles of dynamic loading, caused an increment in acceleration measurements. Using Soil 1 (coarser) or Soil 2 (finer) backfill material does not cause significant difference in behavior of the material during seismic loading between 3 Hz and 5 Hz. In Fig. 5, amplification factors of acceleration for each block are shown for all backfill soil types. As seen in Fig. 5, the highest amplifications occurred on the top of the structures (Block 3) for both Case 1 and Case 2.

Excess pore pressure generation
The excess pore pressure has significant effects on soil pressure and horizontal displacements on structures under dynamic loading. Kim et al. (2005) stated that if the excess pore pressure increased, the backfill soil behaved increasingly like a fluid, thus the mobility of the soil increased. On the other hand, experimental, numerical, and analytical results showed that when permeability increased, the accumulation of excess pore pressure reduced. In this study, gravel type backfill materials (Soil 1 and Soil 2) were used and since gravels are more permeable, significant excess pore pressures usually do not generate for this kind of backfill. Thus, the effect of the excess pore pressure was neglected in 3.3 Soil pressure Talesnick et al. (2014) declared that sensor diameter relative to particle size was important on soil pressure measurement reliability. They suggested that particle size to sensor diameter ratio may be as six particles across the sensor diameter without losing reliability. In this study, particle size to sensor diameter ratios were approximately 4.5 and 11 for Case 1 and Case 2, respectively. Although, there was a possibility of some deviations in the soil pressure measurements for Case 1, it was not considered to be large enough to affect the test results.
As mentioned before, four soil pressure transducers were placed behind blocks, the points of which are shown in Fig. 1. The backfill pressure occurred at the base level of the models was calculated by linear extrapolation method. Total soil pressures were separated into a fluctuating component and a non-fluctuating component by using LOESS smoothing process (Matlab program was used). The name "loess" was derived from the term "locally weighted scatter plot smooth" as this method uses locally weighted linear regres-sion to smooth data. The span can be specified as a percentage of the total number of data points in the data set (http://www.mathworks.com). A span of 0.1 uses 10% of the data points and it was selected as 0.1. Total soil pressure and fluctuating and non-fluctuating components are shown together for Soil 1 with frequency of 5 Hz in Fig. 6 as an example (Karakus, 2013).
In order to define the vertical distribution of the fluctuating and non-fluctuating components at the moment that when the maximum total force occurred, the instantaneous soil pressure forces on the block type quay wall during cyclic motion were calculated and the moment of the maximum total force was determined. After that, fluctuating and non-fluctuating component of total soil pressures corresponding to the maximum total force were obtained from total soil pressure record. Figs. 7a and 7b show the vertical distribution of non-fluctuating components of total soil pressure for Case 1 and Case 2 for different cyclic loadings. As shown in Fig. 7, in the cyclic motions with the frequency of 3 Hz which had the maximum acceleration of 0.13g, vertic-   Case 2, in the cyclic motion with the frequency of 3 Hz which had the maximum acceleration of 0.12g, vertical distributions of non-fluctuating components were almost linear; however, the cyclic motion with frequency of 4 Hz and 5 Hz, (the maximum accelerations are 0.25g and 0.40g respectively) its variation occurred nonlinearly. The sharp drop of the total soil pressure was partly due to the arching effect at the lower points of the wall which are at the intersection between Block 1−Block 2 and Block 2− Block 3. Arching was defined by Paik and Salgado (2003) as a stress redistribution process where stress was transferred around a region of the soil mass, which then became subject to lower stresses. Paik and Salgado (2003) declared that the distribution of active earth pressure behind the wall was non-linear, and the earth pressure distribution differed depending on the mode of wall movement (tilting about the top or base of the wall or translation). With these results, it can be said that for coarser and finer backfill material, vertical distributions of non-fluctuating components should have nonlinearly shape with the increasing base acceleration. These results are consistent with some of previous studies (Hazarika et al., 2008;Dakoulas et al., 2018;Paik and Salgado, 2003).
In Figs. 8a and 8b, vertical distributions of fluctuating components occurred for Case 1 and Case 2 are given. In Case 1, the maximum fluctuating components were formed at the base under cyclic motion with the acceleration of 0.13g. However, the maximum fluctuating components were between 0.25 h and 0.35 h with the increasing frequencies (4 Hz, 5 Hz and 6 Hz) and acceleration values (0.27g, 0.38g and 0.58g). On the other hand, as seen in Fig. 8b In Fig. 9 and Fig. 10, the pressure distribution formed before shaking, after shaking and maximum total pressure during the shaking behind the wall are shown for Case 1 and Case 2, respectively. The pressures after shaking usually became larger than ones measured before shaking. Shapes of the pressure distributions were getting nonlinear by increasing acceleration of the cyclic motion due to increasing frequency. Similar trends formed for distribution of total soil pressures. It was thought that the reasons for this were resulted from the above-mentioned arching effect on backfill and densification of backfill.

Horizontal displacements and tilting of blocks
For gravity quay walls on firm foundations, typical failure modes during earthquakes are seaward displacements  and tilting (PIANC, 2001). Displacement and tilting measurements in the experiments were defined as described below. Position transducers measurements are shown in Fig. 11 for Soil 2 for the frequency of 5 Hz as an example (Karakus, 2013). As seen in Fig. 11, while Block 3, which was located on the top of the wall, had the maximum horizontal displacement, Block 1, which was located on the bottom of the wall, had the minimum horizontal displacement. According to test results, blocks responses under cyclic loading were similar in all tests.
If the seismic horizontal movement of the wall is char-acterized by the horizontal displacement at the wall base, Δx 1 , and at the wall top, Δx 2 , then the tilting of the upper block, α, is expressed as (Fig. 12) (Tiznado and Rodríguez-Roa, 2011): where H is the concrete block height. Summary of the horizontal displacement measurements and calculated tilting values with different frequencies for both Case 1 (coarser) and Case 2 (finer) are presented in Table 5. In PIANC (2001), the performance level of gravity  Hulya Karakus CIHAN, Kubilay CIHAN China Ocean Eng., 2021, Vol. 35, No. 2, P. 281-290 287 quay wall considering tilting is separated into four titles. If tilting (θ) is smaller than 3°, the minimum damage level (MD) occurs and when tilting is between 3° and 5°, damage level of gravity quay wall is defined as the controlled damage level (CD). In case larger tilting (5°−8°) damage level is accepted as extensive damage in near collapse (NC) and finally damage level is named as complete loss of structure (collapse (C)) if tilting is larger than 8°. In Fig. 13, tilting and damage levels which correspond to tilting under cyclic motion are shown for Case 1 and Case 2. As revealed by Fig. 13, tilting of Block 2 stayed in minimum damage level (MD) for Case 2, but Block 3 reached the controlled damage level (CD) when the base acceleration was 0.4g. In Case 1, minimum damage level occurred for Block 2 and Block 3 when the base acceleration was up to 0.4g. For acceleration value of 0.58g, while damage level of Block 2 was con-trolled damage level (CD), damage level of Block 3 was near collapse (NC). As seen from Table 5, horizontal displacement measurements increased while frequency increased for Case 1 and Case 2. In general, the calculated tilting degree and vertical displacement measurements increase while the frequency increases for Case 1 (coarser) and Case 2 (finer). The horizontal displacement measurements of the block(s) located at the top, were always larger than those of the block(s) located at the bottom.

Damage levels
Damage levels of structure considering displacements of the top block for three blocks were obtained by using d/H i . From experimental results, d and H i defined the blocks displacements and the height of blocks relative to base, respectively. In Fig. 14, variation of d/H i versus the base acceleration for Case 1 and Case 2 is shown. In PIANC (2001), the limit of performance for gravity wall is given as the ratio of the permanent displacement to the height of the gravity wall d/H i (%). According to this definition, the minimum damage (MD) occurs when d/H=0.015 and controlled damage (CD) occurs when d/H=0.015−0.05, near collapse (NC) occurs when d/H=0.05−0.10 and collapse (C) occurs when d/H>0.10.
As seen in Fig. 14, when the damage level of each block was investigated considering damage level defined in PI-ANC (2001), in Case 1, all blocks reached the upper limit of the minimum damage level while the base acceleration was Fig. 11. Horizontal displacement measurements of three blocks with the base acceleration of 0.38g for Soil 1. Fig. 12. Displacement and tilting for three blocks. around 0.20g and controlled damage level while the base acceleration was around 0.35g. In Case 2, all blocks reached the upper limit of the minimum damage level while the base acceleration was around 0.25g and collapse damage level while the base acceleration was around 0.40g. It was understood that the response of Case 2 during dynamic loading caused more damage than Case 1 especially after the base acceleration becomes larger than 0.3g in terms of the damage criteria given in PIANC (2001). In addition, it was observed during experiments that Case 2 settled down behind the structure more easily and this led to a more powerful push effect. Moreover, blocks were placed without any shear key between blocks and Case 2 could replace the space between the blocks and could enhance the slipping condition between the blocks. Acceptable damage level of the structure was determined by considering the most critical condition. According to experiment results, horizontal displacement was more critical than tilting, thus acceptable level of damage on the wall was obtained by considering critical condition.
Derince Port, located near Izmit, is the largest port in the area with about 1.5 km of waterfront structures and eight wharves. The Eastern Marmara Earthquake (Mw=7.4) struck the northwestern Turkey in 1999, including Izmit Bay area and eastern Marmara Sea region. The peak ground accelerations were obtained approximately 0.25g to 0.3g. Site measurements show that the block type quay wall moved seaward without any vertical displacement. However, 0.5 m lateral displacement towards the sea and 0.5−0.8 m settlement on the backfill behind the quay wall were observed (Yüksel et al., 2002 Tilting degree occurred at Derince Port, block type quay wall was calculated as α = 1.94° (arctan(0.50/14.75)).
PIANC (2001) gives the damage levels in terms of structural and operational levels. According to these definitions, the damage level for the Derince Port, block type quay wall falls into "controlled damage" as that for the structural level. In view of "operational damage" criteria, the block damage level was the short-term loss of serviceability. This result was also compatible with the site measurements such that even after the damage due to earthquake, the block type quay wall was still in use for berthing and mooring purposes, which was inconformity with the above given definitions.
This result is a good evidence of the reliability of the definitions of damage levels given in PIANC (2001) to be used in performance-based approaches for seismic design of block type quay walls.

Conclusions
In this study, seismic response of block type quay wall, which consisted of three concrete blocks, was investigated experimentally for two different backfill materials (coarse and fine). Based on the experiment results, it can be said that finer backfill material causes more damage (more displacement). For this reason, it is recommended to use coarser soil as backfill material especially in high seismic zones. If the damage parameters of each block are taken into consideration separately, the largest displacement and tilting are seen in the topmost block (Block 3). In Case 1 and Case 2, the damage level remains within the controlled damage level up to the base acceleration of 0.3g. By considering the damage levels caused by tilting and displacement, the damage levels corresponding to tilting or displacement may be different. Damage level of the structure has to be determined by considering the most critical condition related to damage level of tilting and horizontal displacements.
During the tests, total soil pressures on quay wall due to Hulya Karakus CIHAN, Kubilay CIHAN China Ocean Eng., 2021, Vol. 35, No. 2, P. 281-290 289 backfill were measured instantaneously at four points which had different elevations. It can be said that for coarser and finer backfill material, vertical distribution of non-fluctuating components has a nonlinear shape with increasing base acceleration. Also the shape of the total soil pressures distributions is becoming nonlinear by increasing frequency and also acceleration of cyclic motion due to arching effect on backfill and densification of backfill.