Experimental Studies on Extraction of Modified Suction Caisson (MSC) in Sand by Reverse Pumping Water

A suction caisson can be extracted by applying reverse pumping water, which cannot be regarded as the reverse process of installation because of the dramatically different soil-structure interaction behavior. Model tests were first carried out in this study to investigate the extraction behavior of the modified suction caisson (MSC) and the regular suction caisson (RSC) in sand by reverse pumping water. The effects of the installation ways (suction-assisted or jacking installation) and the reverse pumping rate on the variations of the over-pressure resulting form reverse pumping water were investigated. It was found that neither the RSC nor the MSC can be fully extracted from sand. When the maximum extraction displacement is obtained, the hydraulic gradient of the sand in the suction caisson reaches the critical value, leading to seepage failure. In addition, the maximum extraction displacement decreases with the increasing reverse pumping rate. Under the same reverse pumping rate, the final extraction displacements for the RSC and MSC installed by suction are lower than those for the RSC and MSC installed by jacking. The final extraction displacement of MSC is almost equal to that of the RSC with the same internal compartment length. Based on the force equilibrium, a method of estimating the maximum extraction displacement is proposed. It has been proved that the proposed method can rationally predict the maximum extraction displacement and the corresponding over-pressure.


Introduction
Generally speaking, offshore wind turbines are in operation for approximately 20 years. The world's first offshore wind farm Vindeby began operations off the southeastern coast of Denmark in 1991 (Lumbreras and Ramos, 2013), and in 2017 it were decommissioned and dismantled. The selection of the offshore wind farm site is greatly affected by some factors such as hydrology, climate and animal migration. It means although the ocean is so huge, it is not random to select the sites to construct offshore wind farms. Therefore, the existing foundation should be removed to update the offshore wind turbine with higher power, for example, from 3−5 MW to more than 7 MW (Lehane et al., 2014).
A suction caisson is a cylindrical structure that is openended, and closed by a lid at the upper end, which can be used as foundations of offshore structures, such as offshore oil and gas platforms (Senpere and Auvergne, 1982;Tjelta et al., 1990Tjelta et al., , 1992 or of offshore wind turbines (Koh et al., 2017;Det Norske Veritas, 2004;Zhang P. Y. et al., 2018). Compared with gravity base foundation and large diameter monopiles widely used as foundations for wind turbines, a suction caisson is characterized by more convenient installation and less steel consumption. In addition, another major advantage of the suction caisson lies in its ability to be extracted by reverse pumping water.
Suction caissons can readily penetrate into the seabed under a combination of the self-weight and suction by continually pumping the enclosed water out of the caisson (Tran et al., 2004). The suction caisson settles down to a certain depth under its own gravity and obtains the hydraulic seal between the caisson interior and the outside environ-ment. Then, pumping water out of the caisson compartment induces a net downward differential pressure (negative differential pressure, also called suction), which provides the downward driving force penetrates the caisson to the desired depth (Tran et al., 2005a). Previous studies show that the suction caisson can penetrate into fine, medium and coarse sand, as well as the soft clay and layered soil (Tran et al., 2005b). In addition, by applying reverse pumping water, the suction caisson can be extracted partially or fully from the soil (Lehane et al., 2014;Zhang Y. K. et al., 2018).
Previous studies focused on the penetration mechanism and the bearing capacity under the pull-out loading. However, the extraction behaviour of the suction caisson in sand has received little attention. Zhang et al. (2013) conducted installation and extraction tests on the behaviour of prototype suction caisson foundations in clay at Bohai Bay and found that the seepage induced the damage in soil plug when extracting the caisson by reverse pumping water. Lehane et al. (2014) proposed a method of predicting the pullout resistance of suction caisson in sand by using a series of centrifugal tests and numerical simulations. Zhang et al. (2017) carried out model tests to investigate the installation and extraction behavior of the suction caisson. When the suction caisson was extracted to a certain height, sand seepage failure occurred inside and outside the suction caisson, causing the soil seal failure to terminate the extraction. Cerfontaine et al. (2016) suggested that the pullout capacity should be equal to the sum of its dead weight and the vertical friction force of the caisson wall under undrained condition.
The modified suction caisson (MSC) is a novel type of foundation for offshore wind turbines, whose external structure is added around the lid of the regular suction caisson (RSC), which significantly increases its lateral and vertical bearing capacities. Model test and numerical simulation results of Li et al. (2014) showed that the lateral and vertical capacities of MSC were higher than those of the RSC under the same weight. In addition, the overturning movement of the MSC was in more effective control to meet the requirements of the offshore wind turbine operation. Li et al. (2018Li et al. ( , 2020 also found that the MSC had good performance of the suction-assisted installation and extraction by reverse pumping water in clay.
This paper aims to investigate the extraction behavior of the MSC in saturated fine sand compared with the RSC by model tests. The effects of penetration ways and reverse pumping rates on the extraction are discussed. Results can be used to guide the caisson retrieval for practical engineering.

Test models and devices
As shown in Fig. 1, the models of RSC and MSC are made of stainless steel, which have the vent connected to the vacuum pump that can pump water out to install the suction caisson or reversely pump water to extract. Both the internal compartment and the external structure of the model suction caissons have the beveled tip to reduce the tip resistance. A guide rod has been installed on the middle of caisson lid to guarantee the perpendicularity during penetrating and extracting. Dimensions and self-weights of the RSC and MSC are given in Table 1, and other parameters are shown in Fig. 2. As shown in Fig. 3, the stainless-steel circular test box is 1.5 m in diameter and 1 m in height. A hydraulic jack fixed on the test box was used to install the model suction caisson into the sand or pull the suction caisson out of the sand. During installation and extraction, the vertical displacement was measured by a Linear Variable Differential Transformer (LVDT) and a pressure transducer was used to record the pressure inside the internal compartment (Fig. 4).
The sand was collected from the beach in Fuzhou, and its parameters are given in Table 2 and grain size distribution curve is shown in Fig. 5.

Procedures of testing
The sand was stirred at the depth within 1.5 times the height of the caisson wall (Li et al., 2015). Then water was filled into the box until the water level reached 2 cm higher than sand surface. Meanwhile, the drain valve at the bottom of test box was opened in order to consolidate the sand. Each test was carried out at least 3 times and the sand should stand at least 24 h before conducting the test. Table 3 gives the model tests conducted in various ways.

Suction-assisted installation
During the suction-assisted installation, an upward seep-age makes the sand plug upheaval and loose. Additionally, a clear trend is reported that the sand tends to flow around the tip into the caisson (Zhang Y. K. et al., 2018). Once the soil plug surface touches the caisson lid, the penetration terminates. Table 4 lists the final sand plug height. The sand plug heights for the RSC and MSC increase with the increase of the installation pumping rate. Therefore, the installation pumping rate should be strictly controlled to guarantee the desired penetration depth. As shown in Fig. 6, under various installation pumping rates, the final sand plug height in the MSC is about 0.12−0.15 times the height of the internal compartment.
3.2 Extraction by reverse pumping water By reverse pumping water, the caisson can be extracted from the sand bed. Table 5 lists the extraction displacements of the RSC and MSC under different installation ways and reverse pumping rates. It can be seen that none of the suction caissons can be fully extracted from sand by reverse pumping water. The final extraction displacements are smaller than 0.5 times the suction caisson length. Further extraction can be achieved by applying the pull-out force using the hydraulic jack. For the prototype suction caisson, fully extraction can be achieved by using the combination of the pull-out force and reversing pumping. It should be noted that, there is a significant increase of the peak value of the over-pressure at the beginning of the MSC extraction.      Moreover, the final extraction displacement decreases gradually with the increase of the reverse pumping rate. As shown in Table 5, the final extraction displacement under the reverse pumping rate of 3 L/min is about 8% higher than that under the reverse pumping rate of 12 L/min. The in-stallation way also has a significant effect on the final extraction displacement. Under the same reverse pumping rate, the final extraction displacement of the suction caisson installed by jacking is obviously higher than that installed by suction. Fig. 7 gives the relationship between the extraction displacement and the time elapse. The extraction displacement increases linearly with time to the maximum value. However, when the reverse pumping rate is equal to or larger than a certain value (9 L/mm and 6 L/mm for the RSC and the MSC respectively), the extraction displacement reduces to a constant value after reaching the peak value, indicating that the suction caisson moves downward. The reason is that when the maximum extraction displacement achieves, seepage failure occurs in the suction caisson (Fig. 8a), and thus, the over-pressure dissipates rapidly. Therefore, the extraction force is smaller than the self-weight of the suction caisson, leading to the suction caisson fallback.

Variations of over-pressure
It is found that when the extraction displacement reaches the peak value, the seepage failure occurs (Zhang Y. K. et al., 2018). This over-pressure can be defined as the critical  HUANG Ling-xia et al. China Ocean Eng., 2021, Vol. 35, No. 2, P. 272-280 275 pressure when the seepage failure occurs. As shown in Table 6, the critical pressure increases with the increase of the pumping water rate. Fig. 9 gives the relationships between the over-pressure and the extraction displacement for the RSC installed by suction. The over-pressure−extraction displacement curves follow a similar trend consisting of three stages. The first stage is: the over-pressure increases rapidly without obvious extraction displacement. The over-pressure provides the upward extraction force. When the over-pressure exceeds a certain value, the extraction force is larger than the sum of the suction caisson self-weight and the frictional force between the suction caisson shaft and sand, leading to the start of the extraction. The extraction displacement corresponding to the peak value of the over-pressure decreases with the increase of the pumping rate. It indicates that the seepage induced by reverse pumping can reduce the extraction resistance. In the second stage, when the reverse pumping water rate is equal to or smaller than 9 L/s, the overpressure gradually decreases with the increasing extraction   . 9. Relationship between over-pressure and extraction displacement after suction penetration. displacement. When the reverse pumping water rate equals 12 L/s, the over-pressure stays constant. The third stage is the seepage failure stage. When the suction caisson is extracted to a certain height, seepage failure occurs around the suction caisson. In addition, a sharp reduction of the overpressure is observed and no further extraction displacement is achieved. Similar to the trend of RSCs, MSCs also have three stages during extraction. In the first stage, the maximum over-pressure is larger than that for the RSC. The external structure provides higher frictional forces. The final extraction displacement of the MSC is nearly equal to that of the RSC.
Although increasing the reverse pumping rate can increase the extraction rate, the negative correlation between the final extraction displacement and the reverse pumping rate still suggests that there seems to be an upper bound value for the over-pressure required during extraction. If seepage failure occurs, the open cavitation area around the suction caisson breaks the seal between the inside and outside of suction caisson, leading to the reductions of the over-pressure value and the extraction force. However, the excessively high over-pressure will induce the seepage failure. In some cases, it is observed that seepage failure occurs immediately for the suction caisson installed by suction when the over-pressure increases to 9−10 kPa. Fig. 10 shows the relationship between the over-pressure and the extraction displacement of the MSC and RSC installed by jacking force during extraction. Compared with the suction caissons installed by suction, the maximum over-pressure and the maximum extraction displacement are significantly increased for the suction caissons installed by jacking. It should be noted that the required pressure to extract and the final extraction displacement increase with the increasing self-weight of the suction caisson.

Estimation of maximum extraction displacement
The maximum extraction displacement appears to be closely related to the magnitude of the over-pressure. It is necessary to estimate the maximum extraction displacement (h L ) accurately. Once the hydraulic gradient of the surrounding soil reaches the critical value, seepage failure will occur, leading to the termination of the extraction. Fig. 11 shows the forces acting on the MSC and RSC. The extraction force (F L ), which is induced by over-pressure, can be calculated by where D i is the inner diameter of the suction caisson. Since the thickness of suction caisson wall is smaller than the outer diameter of suction caisson, it can be assumed that D i ≈D o , where D o is the outer diameter of the suction caisson. The suction caisson will be extracted when F L is larger than the sum of its self-weight and vertical frictional resistance. During RSC and MSC extraction, the frictional forces acting on the suction caisson wall decreases with the increasing extraction height. In this study, by considering the effect of soil plug height, the vertical force equilibrium can Fig. 10. Relationship between over-pressure and extraction displacement after jacking penetration.
HUANG Ling-xia et al. China Ocean Eng., 2021, Vol. 35, No. 2, P. 272-280 277 be expressed as: where F io and F ii are the outer and inner frictional forces of the caisson wall, and W is the self-weight of the suction caisson. The outer and inner frictional forces can be obtained by The outer and inner frictional forces are influenced by the soil plug height, extraction displacement, vertical effective stress in soil and the frictional coefficient of the interface between sand and caisson wall. For the MSC, Eq. (2) becomes where F so and F si are the outer and inner frictional forces of the external structure wall and can be obtained by According to Eqs. (2) and (4), the extraction resistance decreases with the increasing extraction displacement.
When a seepage failure occurs, the shortest seepage path from the caisson is 2L+t i . To predict the soil hydraulic gradient, the excess pore pressure P s on the soil plug surface should be obtained. Assuming that the position of the sand plug surface is unchanged during extraction, P s can be obtained by where P top is the over-pressure, and h L is the distance between the sand plug surface and the caisson lid. Based on the force equilibrium, Houlsby and Byrne (2005) proposed a method to calculate the required suction to penetrate the suction caisson in sand. We extend this method to estimate the maximum extraction displacement and the corresponding over-pressure. In addition, other assumptions are given as follows: (1)  In this study, is defined as the head loss in the caisson. Therefore, the excess pore pressure in the soil out of the suction caisson can be expressed by . The relationship between the over-pressure P s and buoyant uof sand can be expressed by where w is the unit weight of water, and is the buoyant unit weight of sand. When the critical hydraulic gradient is reached, according to Eq. (7), the critical excess pore pressure inside the caisson is: (8) α α α Eq. (8) indicates the effect of the factors such as head loss, embedded depth, and effective unit weight on the suction caisson extraction. Houlsby and Byrne (2005) suggested a function to express the relationship between and h/D and concluded that the value of is in the range of 0−5. Besides, is also influenced by the ratio of the hydraulic conductivities of the sand inside and outside the suction caisson: α α where k f =k i /k o , k i and k o are the hydraulic conductivities of sand inside and outside of the suction caisson. If the suction caisson is installed by jacking, k f =1. According to Eq. (9), the head loss inside the suction caisson is proportional to k f . In addition, 1 can be obtained by (10) Fig. 12 shows the over-pressures for the MSC and the RSC by Eqs. (6) and (8) under various installation ways. Model test results on the final extraction displacement and the correspoding over-pressure are also given in Fig. 12. It can be seen that the calculated results agree well with the test results.
For RSCs, when the maximum extraction displacement is reached, the hydraulic gradient of sand in the suction caisson is , and the hydraulic gradient of sand outside the suction caisson is . Therefore, when the seepage failure occurs, the critical excess pore pressure is α As shown in Fig. 12, the estimated results for the RSC and the MSC agree well with the test results when k f equals 0.92 and 0.54 respectively. Therefore, it can be concluded that the values of k f and is primarily effected by the installation way and the suction caisson type. In addition, the value of k f for the suction caisson installed by suction is smaller than that installed by jacking.

Conclusions
A series of model tests were performed to investigate the extraction behaviours of the RSC and the MSC by pumping water into the suction caissons. The following conclusions are obtained: (1) Neither the RSC nor the MSC can be fully extracted from sand, and the maximum extraction displacement of suction caissons gradually decreases with the increase of reverse pumping water rate. Under the same reverse pumping water rate, the maximum extraction displacement of MSC is almost equal to that of the RSC with the same internal compartment length. It was also found that the final extraction displacement of the suction caisson installed by jacking is larger than that installed by suction.
(2) The suction caisson extraction can be divided into three stages. At the first stage, the over-pressure increases rapidly without obvious extraction displacement. At the second stage, the suction caisson is extracted at a constant rate, and the over-pressure gradually decreases with further extraction. The third stage is the seepage failure stage. When the suction caisson is extracted to a certain height, seepage failure occurs around the suction caisson. In addition, a sharp reduction of the over-pressure is observed and no further extraction displacement is achieved.
(3) A method of estimating the maximum extraction displacement of the suction caisson is proposed by considering the influence of the installation ways, which has been proven more reasonable and highly accurate for predicting the maximum extraction displacement and the corresponding over-pressure. alternative to driving or drilling, Offshore Technology Conference, Offshore Technology Conference, Houston, Texas, USA, pp. 483-493. Tjelta, T.I., Aas, P.M., Hermstad. J. and Andenaes, E., 1990. The skirt piled Gullfaks C platform installation, Annual Offshore Technology Conference, Offshore Technology Conference, Houston, Texas, USA, pp. 453-462. Tjelta, T.I., Janbu, N. and Grande, L., 1992