Pile Installation Assessment of Offshore Wind Jacket Foundation in Completely Weathered Rock: A Case Study of the South China Sea

: The assessment of pile drivability is the premise of successful pile installation, and completely weathered (CW) rock may bring the risk of pile refusal. This paper presents a case of pile driving refusal that occurred in a CW gneiss layer. The physical and mechanical properties of the CW gneiss are analyzed based on the results of a laboratory test, a cone penetration test (CPT), and a standard penetration test (SPT). By comparing the properties of CW gneiss with those from intact gneiss, the CW gneiss layer in this study was found to be soil-like. Additionally, it is a sand-like layer according to CPT-based soil classiﬁcation systems. The soil resistance to driving (SRD) in pile driving was calculated with different methods for sand and compared with the SRD inferred from the pile-driving records. The results show that the modiﬁed Fugro’s method assuming a fully plugged condition may provide good pile drivability assessment in this study. It was found that it is necessary to determine the properties of the CW rock layer reasonably before conducting pile driving analysis, and a framework is proposed for analyzing the pile drivability in a CW rock layer based on this case study.


Introduction
Pile foundations have been used commonly for offshore wind turbines.With the rapid development of offshore wind farms, the requirements for pile foundations' capacity has been increased, including axial pile capacity, pullout capacity, and horizontal capacity [1][2][3][4].Meanwhile, with the increase in pile length, conditions of various soils including weathered rock may be encountered in construction, which brings a great risk of pile-driving refusal.
Pile refusal is the phenomenon in which the pile cannot be penetrated before reaching the design depth during pile installation.This mainly occurs in dense sand, hard clay, and weathered rock [5,6].The installation failure caused by pile refusal means a great cost in terms of time and economics [5].
Considering the consequences of pile refusal, it is necessary to accurately conduct pile drivability analysis.Wave equations are commonly used to conduct pile drivability analysis.During pile driving, the total soil resistance that affects the pile drivability consists of dynamic and static components.The dynamic component is affected by inertial and viscous rate effects [7,8].Meanwhile, the static component, the so-called soil resistance to driving (SRD), is different from the static axial capacity after long-term interruption due to the consolidation, stress equalization, and pile-aging effect [7].SRD is a key factor in pile drivability analysis, and the methods for calculating SRD is different for sand, clay, and rock.
The SRD in sand and clay can be determined based on the results from laboratory tests [9,10] or the cone penetration test (CPT) [1,7,11,12].Stevens et al. [10] proposed that SRD could be calculated with a friction angle for sand and with undrained shear strength for clay considering a cored and plugged pile condition.Semple and Gemeinhardt [9] proposed a stress history approach to assess SRD considering the over-consolidation ratio of soils.
With the wide application of CPT, CPT data are also used to determine the SRD of piles and the axial pile capacity, which is sometimes corrected to calculate SRD.Four CPT-based methods (ICP-05, UWA-05, Fugro-05, UWA-05, and NGI-05) for offshore pile capacity were included in the American Petroleum Institute's (API's) recommended practice (2GEO) [13][14][15][16].For the SRD of pile, based on the database of the pile for the offshore platform, Toolan and Fox [11] proposed a method to calculate SRD with the cone resistance of CPT.Considering the friction fatigue during pile driving, Alm and Hamre [12] proposed a CPT-based SRD method.Prendergast et al. [1] proposed three modified axial-capacity methods based on ICP-05, UWA-05, and Fugro-05 to assess the SRD of open-ended piles of sand.Considering soil plasticity, the over-consolidation ratio, the pile length, the pile slenderness, and the time between the pile driving and loading, Van Dijk [17] proposed a CPT-based method for long steel open-ended piles in clay.
For the SRD in rock, Stevens et al. [10] proposed that because the pile installation may break the rock into granular material, the shaft friction can be computed based on the friction angle as is the case in sand.The end bearing can be computed based on the friction angle as is the case in sand for a poor-to fair-quality rock.Meanwhile, for a more high-quality rock, the end bearing can be computed based on the unconfined compression strength (UCS).Brunning and Ishak [18] calculated the SRD in carbonate rock with Stevens's method and proposed that the unit end bearing is 2-3.5 times the unconfined compressive strength for the carbonate rock.Based on the onshore and offshore dynamic test results for low-medium-density chalk, Buckley et al. [19] studied the pile driving in low-to medium-density chalk and proposed a CPT-based ICP-05-like formula to calculate SRD considering the friction fatigue and the ratio of the pile diameter to the pile wall thickness.
It can be found that there is a great controversy about the methods of calculating SRD in rock.This could mainly be because of the definition of the rock itself.Howat [20] compared the shear strength profile of completely weathered granite with existing models of soil and rock behavior and found that the completely weathered granite is more soil-like than rock-like in its shear strength behavior.Gamon [21] raised an objection and proposed that "completely weathered granite" should be considered "extremely weathered granite" and pointed out that the finer textured material in the mass structure may be the residual soil.There are also other scholars who supported the claim that the completely weathered granite was more soil than rock, as observed from recent field investigations [22][23][24].Chen et al. [25] studied the uneven grain size distribution of completely weathered granitic gneiss and found that the weathered granitic gneiss's behavior was simultaneously that of sandy soil and clayey soil.Hawkins [26] suggested that highly weathered mudrock develops the mass-engineering properties of soil rather than those of intact rock, while completely weathered mudrock possesses no rock strength.
In this paper, a pile driving refusal in a completely weathered (CW) gneiss layer is studied based on a project in the South China Sea.The engineering properties of the CW gneiss were discussed based on results from laboratory tests, CPT, and SPT.The SRD was calculated with different methods using laboratory tests and CPT results and compared with that from back analysis according to pile-driving records.The framework used in this study could be helpful for the drivability analysis of an offshore pile installation in CW rock in practice.

Case Investigation
An offshore wind farm composed of 55 wind turbines was built about 20 km from the coast.The jacket substructure was used in this project, and it was supported by four steel piles (named pile A1, A2, B1, and B2).The layout of the piles and locations of sampling borehole (BH), SPT, and CPT are shown in Figure 1.

Case Investigation
An offshore wind farm composed of 55 wind turbines was built about 20 km from the coast.The jacket substructure was used in this project, and it was supported by four steel piles (named pile A1, A2, B1, and B2).The layout of the piles and locations of sampling borehole (BH), SPT, and CPT are shown in Figure 1.The results of CPT and SPT are shown in Figure 3.The final depth of CPT is 42.8 m, and it is 51.6 m for SPT.In Figure 3,  is the cone resistance,  is the unit sleeve friction resistance, and  is the pore water pressure.Figure 3

Case Investigation
An offshore wind farm composed of 55 wind turbines was built about 20 km from the coast.The jacket substructure was used in this project, and it was supported by four steel piles (named pile A1, A2, B1, and B2).The layout of the piles and locations of sampling borehole (BH), SPT, and CPT are shown in Figure 1.The results of CPT and SPT are shown in Figure 3.The final depth of CPT is 42.8 m, and it is 51.6 m for SPT.In Figure 3, q c is the cone resistance, f s is the unit sleeve friction resistance, and u is the pore water pressure.Figure 3 also shows the change in SPT-N value with depth.
The length of each pile is 61 m, its external diameter is 2.4 m, and the wall thickness at the pile toe is 55 mm.All piles were driven into the soil for about 23 m with a vibratory hammer at first, and then a hydraulic impact hammer (MHU1200) was used to drive the pile to the designed depth (53 m).The premature refusal occurred for all piles when the penetration depth was about 40 m; note that the CW gneiss layer appeared at 32.3 m.Take the pile B1 as an example: from 0 m to 23.25 m, the pile was driven using a vibratory hammer, a MHU1200 hydraulic hammer was used after 23.25 m, and the pile driving refusal occurred at 38.5 m.The length of each pile is 61 m, its external diameter is 2.4 m, and the wall thickness at the pile toe is 55 mm.All piles were driven into the soil for about 23 m with a vibratory hammer at first, and then a hydraulic impact hammer (MHU1200) was used to drive the pile to the designed depth (53 m).The premature refusal occurred for all piles when the penetration depth was about 40 m; note that the CW gneiss layer appeared at 32.3 m.Take the pile B1 as an example: from 0 m to 23.25 m, the pile was driven using a vibratory hammer, a MHU1200 hydraulic hammer was used after 23.25 m, and the pile driving refusal occurred at 38.5 m.
The pile-driving records of pile B1 under hydraulic hammer are shown in Figure 4. Considering the variance in energy per blow at different depths in the pile driving, the variation in energy with depth is also shown in Figure 4.It could be found that all the premature refusal occurred in the CW gneiss layer.It is also necessary to learn more about its engineering properties.It could be found that all the premature refusal occurred in the CW gneiss layer.It is also necessary to learn more about its engineering properties.It could be found that all the premature refusal occurred in the CW gneiss layer.It is also necessary to learn more about its engineering properties.The results of the laboratory tests showed that the water content of the CW gneiss varied from 12% to 24%, and the submerged unit weight varied from 7.1 kN/m 3 to 14.1 kN/m 3 (see Figure 2).Considering that the soil properties determined by either CPT or SPT tests are based on semi-empirical methods, the values of submerged unit weight, Young's modulus, and shear wave velocity were derived using CPT and SPT results, respectively [27][28][29][30][31][32][33][34][35][36][37].This could control the above values within a reasonable range.Additionally, their changes with depth are shown in Figure 6.The results of the laboratory tests showed that the water content of the CW gneiss varied from 12% to 24%, and the submerged unit weight varied from 7.1 kN/m 3 to 14.1 kN/m 3 (see Figure 2).Considering that the soil properties determined by either CPT or SPT tests are based on semi-empirical methods, the values of submerged unit weight, Young's modulus, and shear wave velocity were derived using CPT and SPT results, respectively [27][28][29][30][31][32][33][34][35][36][37].This could control the above values within a reasonable range.Additionally, their changes with depth are shown in Figure 6.

The Properties of the Gneiss Layers
It was found that the values obtained from SPT are close to but not completely consistent with those obtained from CPT.The reason may be an error from the empirical formula, or the spatial variability in engineering properties in the gneiss layer, though the location of SPT and CPT is only 32 m away.
It was found that the calculated submerged unit weight of the CW gneiss (6.2~13.2kN/m 3 ) (the value from laboratory test was 7.1~14.1 kN/m 3 ) was less than the empirical value of intact gneiss, which was 15~18 kN/m 3 based on Chinese engineering geotechnical manual (EGM).The calculated Young's modulus (11~113 MPa) almost falls within the empirical range of dense sand (25~100 Mpa) based on USACE engineer manuals (EM).The calculated shear wave velocity (300~580 m/s) almost falls into the empirical range of very dense soil and soft rock (360~600 m/s) according to Federal Emergency Management Agency (FEMA) P749.
As can be seen from the above index, the CW gneiss becomes more soil-like, which is consistent with the results of previous studies [20,22,24].Therefore, Robertson's CPT-based soil classification systems are used to find whether it is sand-like or clay-like [38,39].It was found that the values obtained from SPT are close to but not completely consistent with those obtained from CPT.The reason may be an error from the empirical formula, or the spatial variability in engineering properties in the gneiss layer, though the location of SPT and CPT is only 32 m away.
It was found that the calculated submerged unit weight of the CW gneiss (6.2~13.2kN/m 3 ) (the value from laboratory test was 7.1~14.1 kN/m 3 ) was less than the empirical value of intact gneiss, which was 15~18 kN/m 3 based on Chinese engineering geotechnical manual (EGM).The calculated Young's modulus (11~113 MPa) almost falls within Robertson [38] proposed a soil classification system based on normalized cone resistance (Q t ) and pore pressure parameter (B q ).
B q = ∆u/(q t − σ v0 ) where q t is the corrected cone resistance (q t = q c + 0.75u); u is pure water pressure; σ v0 is the total overburden stress; σ v0 is the effective overburden stress; ∆u is the excess pore water pressure (∆u = u 2 − u 0 ); u 2 is the pore pressure measured behind cone; u 0 is the in-situ pore pressure.The CW gneiss is classified as sand using the above method.In 2016, Robertson [39] recommended an updated soil classification system using behavior-based description.The CPT results of CW gneiss scattered in this zone are classified as sand-likedilative, transitional-dilative, and clay-like-dilative soils, but it is difficult to determine the type using the updated system above.
A method proposed by Niazi [40] is also used to classify the soil based on the classification index I c , which is defined by the corrected normalized cone resistance (Q tn ) and the normalized friction ratio (F r ).
(3) where p a is a reference stress (100 kPa); and n is the stress normalization exponent (<1.0).The variation in I c with depth is shown in Figure 7. From Figure 7, most points (more than 90%) along depths (32.3-42.8m) in the CW gneiss layer are classified as gravelly sands, sands, and sandy mixtures.Based on the above analysis, in this case, the soil-behavior type of CW gneiss could be sand-like.Because CW gneiss is defined as sand-like soil, the friction angles of the soil are determined with SPT and CPT data, respectively [28,[41][42][43], and the results are shown in Figure 8.It was found that the calculated friction angle of the CW gneiss layer is high, and the CPT-based friction angles are both less than those obtained based on SPT data.Because CW gneiss is defined as sand-like soil, the friction angles of the soil are determined with SPT and CPT data, respectively [28,[41][42][43], and the results are shown in Figure 8.It was found that the calculated friction angle of the CW gneiss layer is high, and the CPT-based friction angles are both less than those obtained based on SPT data.Because CW gneiss is defined as sand-like soil, the friction angles of the soil are determined with SPT and CPT data, respectively [28,[41][42][43], and the results are shown in Figure 8.It was found that the calculated friction angle of the CW gneiss layer is high, and the CPT-based friction angles are both less than those obtained based on SPT data.

The Methods Used to Calculate SRD
It is very important to decide the proper SRD in pile drivability analysis.The methods used to calculate SRD are different under different soil conditions.Considering that the method of calculating SRD in rock is controversial, and the risk of pile-driving refusal in rock is high, the SRD on completely weathered gneiss is calculated with different methods mentioned below.
Based on the properties of the CW gneiss layer, the following methods are used to obtain SRD, considering it as sand-like.The SRD is the sum of shaft resistance (Q s ) and end resistance (Q b ) during pile driving: where D is the pile external diameter; τ f is the local unit shaft friction at failure along the pile shaft; A b is the pile base area; and q b is the unit end bearing.For the cored pile, the τ f is the total (internal plus external) unit shaft friction, and the A b is the annual area.For the fully plugged pile, the τ f is the external unit shaft friction, and the A b is the total (annual plus plug) area.

Stevens's Method
Stevens et al. [10] proposed that unit end bearing and unit shaft friction of SRD in sand could be calculated using the static axial capacity with the API method [45], where N q is dimensionless bearing capacity factor; σ v0 is the effective overburden pressure; q b,max is the limiting end bearing values; K is the coefficient of lateral earth pressure; δ is friction angle between the pile and soil; and τ f ,max is the limiting shaft friction values.
It should be noted that Stevens et al. [10] proposed a lower bound (LB) and upper bound (UB) of SRD for both the cored and the fully plugged pile.For the cored pile, the LB adopted internal friction as half of the external friction, while the UB adopted an internal friction equal to the external friction.For the fully plugged pile, the LB is the sum of external friction and the base resistance considering the full base area (annal area plus plug area), while the UB increased the shaft friction by 30% and the base resistance by 50% compared with LB.

Alm and Hamre's Method
Alm and Hamre [12] proposed a CPT-based approach to calculate SRD in sand, q b = 0.15q c (q c /σ v0 ) 0.2 (10) k = (q c /σ v0 ) 0.5 /80 (12) where q c is the measured cone resistance; τ res and τ i are residual friction and initial friction; h is height above the pile tip; and p a is the absolute atmospheric pressure (100 kPa).Alm and Hamre [12] proposed that for the cored pile, the internal friction is equal to the external friction, but they are both reduced by 50%.This is equivalent to not considering the internal friction, as there is little difference between the internal friction and external friction for a large diameter and a thin wall pile.For the fully plugged pile, the SRD is the sum of the external friction and the base resistance considering the full base area.This assumption is also used by Prendergast et al. [1] (modified ICP and modified Fugro methods for sand).
For the modified Fugro-05 method, q b = 8.5q c,avg p a /q c,avg 0.5 A r 0.25 (19) τ f = 0.06q c σ v0 /p a 0.05 (h/R * ) −0.90 , for h/R * ≥ 4 (20) where q c,avg is the average q c ± 1.5D over the pile tip; R * is the equivalent pile radius 5 ); R is the external pile radius; and R i is the internal pile radius.
Van Dijk and Kolk [17] proposed an approach to calculate SRD in clay, q b = 0.7 q t,avg − σ v0 (22) where q t,avg is the average q t ± 1.5D over pile tip; σ v0 is the total overburden pressure; uL is the unit length to render the expression dimensionless; q n is the net cone resistance (= q t − σ v0 ); q t is the corrected cone resistance (q t = q c + 0.75u); and u is the pore water pressure.
It should be noted that in this study, for the layers above the CW gneiss layer, the modified Fugro method [1] (Equations ( 19)-( 21)) was used for calculating the SRD in sand layers, and Van Dijk's method [17] (Equations ( 22) and ( 23)) was used for calculating the SRD in clay layers.For the CW gneiss layer, the methods shown above (Equations ( 8)-( 21)) were all used for calculating SRD.

Inferred SRD
At the same time, the back analysis based on the pile driving records was performed with pile driving simulation software GRLWEAP, and the SRD known as the inferred SRD in this study was obtained.The quake and damping parameters used in the analysis are listed in Table 1.

Results and Discussion
A comparison of the inferred SRD and calculated SRD using methods for sand is shown in Figure 9, and it was found that the SRD calculated with the modified Fugro method [1] assuming a fully plugged condition is closer to the inferred SRD.For the plugged condition, the calculated SRD values using all methods except the modified ICP method are all larger than the inferred SRD, which could mean that the calculations are conservative.For the modified ICP method, the calculated SRD is about 30% smaller than the inferred SRD.
Based on the analysis mentioned above, a framework is proposed for analyzing the pile drivability in the CW rock layer.At first, we make a preliminary judgement on whether the CW rock is rock-like or soil-like by analyzing submerged unit weight (γ ), Young's modulus €, and shear wave velocity (V s ), which can be obtained from laboratory test, CPT test, or SPT.It should be determined whether it is soil-like, clay-like, or sand-like using Roberson's or Niazi's CPT-based systems for the classification of soil behavior type.For the sand-like layer, the modified Fugro method is recommended to conduct pile drivability analysis.The above process is shown in Figure 10.
It was found that the CPT results are very helpful in the analysis; therefore, it is suggested that the CPT test should be carried out for sites with CW rock layers.It is noted that the case study in this paper mainly studies the sand-like CW rock, whereas for rock-like and clay-like CW rock, much more research is needed in the future.

Conclusions
In this study, piles driven in the CW gneiss layer were studied based on a project in the South China sea.Pile installation in CW rock may bring about the risk of pile refusal.
Before pile drivability analysis in CW rock, it is necessary to analyze the properties of the CW rock first and then conduct analysis based on the properties.The case in this paper compared the CPT-based and SPT-based properties with those from intact gneiss and determined that the CW gneiss is soil-like.Further, the CW gneiss was determined to be sand-like using the CPT-based soil classification system.This could be helpful to understand the nature of CW gneiss.
Meanwhile, the calculation of SRD showed that the accuracy of pile drivability analysis can be significant improved after determining the properties of CW rock.Additionally, the results from the modified Fugro method are in good agreement with those inferred from pile driving records in practice.In light of this, a pile-driving analysis procedure on CW rock is proposed, which could provide a reference for pile installation on CW rock in the future.

Figure 1 .
Figure 1.Layout of piles and in situ tests.The water depth was 25 m.The soil profile obtained from the sampling borehole is shown in Figure 2. The change in engineering properties with depth, such as water content ( ) (based on ASTM D2216-19), submerged unit weight ( ) (based on ASTM D2216-19), liquid limit (LL), plastic limit (PL) (based on ASTM D4318-17), and undrained shear strength ( ) (from the vane shear test (VST) based on ASTM D2573-01) are also shown in Figure 2.
also shows the change in SPT-N value with depth.

Figure 1 .
Figure 1.Layout of piles and in situ tests.The water depth was 25 m.The soil profile obtained from the sampling borehole is shown in Figure 2. The change in engineering properties with depth, such as water content (ω c ) (based on ASTM D2216-19), submerged unit weight (γ ) (based on ASTM D2216-19), liquid limit (LL), plastic limit (PL) (based on ASTM D4318-17), and undrained shear strength (S u ) (from the vane shear test (VST) based on ASTM D2573-01) are also shown in Figure 2.

Figure 1 .
Figure 1.Layout of piles and in situ tests.The water depth was 25 m.The soil profile obtained from the sampling borehole is shown in Figure 2. The change in engineering properties with depth, such as water content ( ) (based on ASTM D2216-19), submerged unit weight ( ) (based on ASTM D2216-19), liquid limit (LL), plastic limit (PL) (based on ASTM D4318-17), and undrained shear strength ( ) (from the vane shear test (VST) based on ASTM D2573-01) are also shown in Figure 2.

Figure 2 .
Figure 2. Soil properties.The results of CPT and SPT are shown in Figure3.The final depth of CPT is 42.8 m, and it is 51.6 m for SPT.In Figure3,  is the cone resistance,  is the unit sleeve friction resistance, and  is the pore water pressure.Figure3also shows the change in SPT-N value with depth.

Figure 3 .
Figure 3. CPT and SPT results.The pile-driving records of pile B1 under hydraulic hammer are shown in Figure4.Considering the variance in energy per blow at different depths in the pile driving, the variation in energy with depth is also shown in Figure4.

Figure 5
Figure 5 is a photo of a sample from the gneiss layer from the site.It was found that the CW gneiss is gray-brown to brown-yellow, and the structure is basically damaged and still recognizable.Most of the core is weathered into soil and could be fragile when touched.Based on ASTM STP 984, the degree of weathering was divided into five categories: A-Micro Fresh State; B-Visually Fresh State; C-Stained State; D-Partly Decomposed State; and E-Completely Decomposed State.The CW gneiss layer is classified here as Grade E-Completely Decomposed State (from 32.3 m to 55.8 m).

Figure 5
Figure 5 is a photo of a sample from the gneiss layer from the site.It was found that the CW gneiss is gray-brown to brown-yellow, and the structure is basically damaged and still recognizable.Most of the core is weathered into soil and could be fragile when touched.Based on ASTM STP 984, the degree of weathering was divided into five categories: A-Micro Fresh State; B-Visually Fresh State; C-Stained State; D-Partly Decomposed State; and E-Completely Decomposed State.The CW gneiss layer is classified here as Grade E-Completely Decomposed State (from 32.3 m to 55.8 m).

J 15 Figure 7 .
Figure 7. Soil classification of CW gneiss based on CPT data from 32.3 m to 42.9 m.

Figure 7 .
Figure 7. Soil classification of CW gneiss based on CPT data from 32.3 m to 42.9 m.

Figure 7 .
Figure 7. Soil classification of CW gneiss based on CPT data from 32.3 m to 42.9 m.

Figure 9 .
Figure 9.Comparison of SRD between inferred and calculated values (for sand) (a) cored pile; (b) fully plugged pile.

Figure 9 .
Figure 9.Comparison of SRD between inferred and calculated values (for sand) (a) cored pile; (b) fully plugged pile.

Figure 9 .
Figure 9.Comparison of SRD between inferred and calculated values (for sand) (a) cored pile; (b) fully plugged pile.

Table 1 .
Quake and damping values used in SRD prediction.