Application of PISA method for monopile design in layered soil

The foundation design for offshore wind turbines (OWTs) is critical to ensuring their structural integrity and long-term stability. Large-diameter monopiles are widely used for OWTs owing to their proven effectiveness. The conventional design approach, derived from the oil and gas industry’s p-y model, has limitations when applied to monopiles for OWTs. Consequently, a novel design method based on the PISA joint industry research project was developed, which extends beyond the traditional lateral soil resistance (p-y curves) and incorporates additional soil reaction components from shaft friction, pile base shear, and base moment. Unlike the simplistic rule-based conventional approaches, the proposed method necessitates site-specific 3D finite element (FE) simulations for precise calibration of soil reaction curves. Although successfully applied in European offshore wind farms, the PISA method’s performance in regions with distinct seabed compositions, such as China, remains to be verified. Therefore, this paper presents a comprehensive case study conducted at the Xiangshan wind farm, a representative site characterized by layered soil. The study employed in-situ investigations (cone penetration test, borehole) and laboratory tests (direct simple shear, bender element) to delineate the seabed soil profile and derive essential soil parameters for accurate foundation design. To facilitate the design, an Abaqus-based design tool was developed, which automated the extraction of soil reaction curves from the 3D FE simulations, calibration of PISA springs, and execution of one-dimensional beam-spring analyses employing Timoshenko beam theory. Comparative analyses were performed on the computed responses of piles, encompassing load-deflection curves, pile deflections, and bending moment profiles. The outcomes from the 3D FE simulations, uncalibrated rule-based PISA model, and calibrated PISA model were compared to validate the PISA method for monopile design in layered soil conditions.


Introduction
Recent decades have witnessed rapid developments in offshore wind market.By the end of 2022, a total of 64.3 GW of global offshore wind capacity was in operation, accounting for 7.1% of the global wind power installation.Additionally, more than 380 GW of offshore wind capacity is predicted to be added in the next ten years [1].For a typical offshore wind farm, the cost of substructure systems can be up to 30%-40% of the whole project, making it one of the main factors influencing the project's economic feasibility and a key driver of cost optimization [2].Currently, most offshore wind farms are 1337 (2024) 012061 IOP Publishing doi:10.1088/1755-1315/1337/1/012061 2 installed close to shore with water depths of less than 50 m.Among all the foundation types, monopiles are the most widely used for their proven effectiveness, figuring in more than 80% of offshore wind turbine installations in Europe [3].
Monopile foundations are subjected to lateral loads from offshore environmental forces, such as wind, waves, and currents.Therefore, the lateral loading response of a monopile is the main consideration in foundation design.As offshore wind turbines (OWTs) are tall slender dynamicsensitive systems, accurately predicting their foundation stiffness and corresponding dynamic behavior, e.g., natural frequency and fatigue loads is critical to avoid any resonance with environmental loads [4].Normally, the p-y approach is used in the design of monopiles, whereby the monopile foundation is idealized as beam, while the soil is represented by a series of non-interacting, non-linear springs along the pile length.The most widely used p-y formulation for piles is that recommended by the American Petroleum Institute [5].This formulation, which was based primarily on results from a small number of lateral tests conducted in the field on flexible piles, has been successfully used for the design of small diameter offshore piles for many years.However, its reliability has been questioned by both academic researchers and industry engineers, when using it for the design of monopile foundations for OWTs (Klinkvort 2012).Some researchers [6] contest that inaccuracies arise in the predictions obtained with the API p-y model because this approach ignores the contributions along the shaft and at the base of the monopile.Correspondingly, a new design model was proposed by the JIP Pile Soil Analysis (PISA) project, wherein besides the lateral soil springs (i.e., p-y curves), extra springs along the shaft (i.e., distributed rotational springs) and at the pile base (i.e., shear force and base moment springs) were also included to more accurately capture the monopile response [6].Unlike the simplistic rule-based design common in conventional approaches, the PISA model needs to be calibrated against site-specific 3D Finite Element (FE) simulations.In recent years, the PISA model has been successfully applied in many European offshore wind farms.However, the seabed in Europe is mainly uniformly dense sand, as opposed to the highly varied layered soil in China.Therefore, the performance of the PISA design approach in complex layered soil remains to be verified.
Therefore, to evaluate the performance of PISA design approach in layered soil, an example case study of monopile design based on one of the typical offshore wind farms in China was performed.The in-situ site investigations (i.e., cone penetration test (CPT) and borehole sampling) and laboratory soil tests (i.e., physical and mechanical tests) were performed at the given monopile location.The first part of this paper presents the interpretation of the geotechnical data and proposed design soil parameters for monopile response calculation.In the second part, the response of the monopile under ultimate limited state (ULS load) was computed using a 3D finite element model in Abaqus.The computed results were then used to calibrate the PISA spring parameters.Finally, beam-spring simulations using both default and site-calibrated PISA spring parameters were performed to investigate the applicability of the PISA model in layered soils.

Geotechnical data interpretation and design soil parameters
For the monopile foundation in this case study, detailed in-situ site investigations and advanced laboratory tests were performed to obtain the physical and mechanical properties of the soil at this location.This section presents the interpreted results and proposed design soil parameters.

Site investigation results
An in-situ CPT was performed at the design monopile location with a maximum penetration of 86 m.The data pertaining to CPT tip resistance (qc), sleeve friction (fs) and pore water pressure (u2) along the whole penetration length is presented in figure 1 a.As shown in the figure, a small qc and fs values can be observed in the shallow (20-m) soil with high u2 values.This indicates a soft clay type soil in the shallow (20-m) layer.The same results were also observed between approximately 37 to 47 m.On the contrary, the CPT data in the range of 20 to 37 m and below 47 m exhibited very high qc and fs values with small u2 values (even negative).This suggests that the soil in these layers was coarse sand type soil.To quantify the soil stratigraphy more accurately, normalized CPT parameters (i.e., Qt, Fr, and Bq) were calculated and presented in figure 1 -m) layer, implying a soft clay layer, which was also confirmed as a clay-like soil from the soil classification method in [7].A similar response was also identified in the range of approximately 37 to 47 m.For other depth ranges, such as 20 to 37 m, and below 47 m, high Qt (> 20) and negative Bq values were noticed, suggesting a sand-like soil with some dilatancy.

Mechanical properties and design parameter profiles
To further quantify the mechanical properties of the soils at the location and provide inputs on design soil parameters for monopile response calculation, a comprehensive lab testing was performed using the borehole samples.The mechanical properties of the soil from the lab tests are presented in figure 2. As shown in the figure, advanced triaxial and direct simple shear tests were performed to characterize the strength of the soil.In the same figure, the undrained shear strength, calculated from the CPT using an Nkt value of 18 is also included.An excellent consistency between the lab testing results and those derived from CPT data can be observed, despite the smaller strength from the lab tests at a depth of approximately 40 m.This is due to the over-consolidation history of the in-situ soil, which was not considered in the lab testing.Two bender element tests were also performed to quantify the small strain stiffness of the soil.The values calculated from the CPT-based empirical relationships were also included in the same figure.As shown in figure 2, the small strain shear modulus measured from the bender element tests are very close to those derived from CPT data, implying the reliability of the CPT-based correlations.For the sand layers, considering the high fine/clay content and design load state (i.e., ULS) in this study, an undrained loading condition was assumed and the undrained shear strength chart in [8] was used to determine the shear strength of the sand layers, which can be calculated from empirical charts with relative density and fine content (figure.10.1 in [8]).The relative density of sand in this study was approximately 60% based on the CPT-correlation from Jamiolkowski et al. [9], while the sample fine contents were 46%, 27%, and 44% in depth ranges of 19-30 m, 30-40 m, and 50-60 m, respectively.The calculated values are shown in figure 2. Finally, design profiles of soil undrained shear strength and small strain stiffness were proposed for the site, as shown in figure 2. The nonlinear stress-strain response of the soil was modeled using the NGI-ADP model [10].The model features a non-linear hardening law that links the current yield stress with the current plastic shear strain.The hardening law parameters can be calibrated against site-specific stress-strain response measured in soil laboratory tests.The model is therefore suitable for both undrained ultimate capacity and deformation analyses.For the clay and sand soil layers at this site, it was found that their physical properties were very similar to the representative soil units in the HDEC database, i.e., South Sea Clay 2 and Jiangsu silty sand [11,12].Therefore, the stress-strain curves of South Sea Clay 2 and Jiangsu silty sand were adopted to represent those of clay and silty sand soil layers at the site, respectively.The calibrated NGI-ADP parameters for these soils, together with a comparison between the measured data in the lab tests and computed results from the NGI-ADP mode, are represented in figure 3.

FEM monopile-soil system model
The steel monopile at the site had a diameter of 8.7 m (the part below the seabed) and a total depth of 80.7 m with an embedded length of 54.4 m.The wall thickness varied along the pile with a range of 75 to 98 mm.In this study, the FE monopile model, as shown in figure 4, had a plane dimension of 15D by 15D (where D is the pile diameter) and a depth of 100 m; these dimensions were large enough to eliminate any boundary effect.A mesh sensitivity study was also performed using two FE models with different mesh densities, as shown in figure 4. The model pile had a stick-up height of 30.1 m, which was calculated based on the design loads.The steel pipe monopile was simulated by an elastic solid cylinder with the same section bending stiffness (i.e., EI, where E is the Young's modulus, and I is the sectional moment inertia) as the monopile.Its Young's modulus and Poisson's ratio were 210 GPa and 0.3, respectively.(Note: more soil layers can be observed in this figure than those proposed in figure 2. This is mainly because artificial layers were created at the depth, where the pile wall thickness changed, to create a uniform model mesh especially at the interface of pile and soil.)

Monopile response
Figure 5 presents the computed load-deflection response of the monopile at the mudline from the FE models with different mesh densities.Here, results from a parallel calculation using NGI's internal monopile design tool "SUMO" were also included [13].As shown in the figure, the two mesh density models yielded almost the exact same monopile response, suggesting that the model mesh adopted in this exercise was adequate to provide accurate predictions.Additionally, it can be seen that the computed response from "SUMO" was also consistent with those from the 3D FE simulation in Abaqus.The maximum difference in the computed monopile response from the two design tools was less than 1%.However, it should be noted that the computation time of "SUMO" was only approximately 2 mins -significantly less than that of the advanced 3D FE simulation in Abaqus (2 hours).The PISA design method (as shown in figure 6) considers the soil resistance from shaft friction, base shear, and base moment in addition to the lateral soil resistance in the traditional p-y approach.In PISA JIP, explicit model parameters are also provided which were calibrated against the FE simulations, validated against the field tests at Dunkirk testing site.However, it should be noted that PISA JIP recommends that the designer perform a few 3D FE monopile simulations at the design site and calibrate the PISA spring parameters based on the simulation results for that design site.The calibrated PISA model can then be used for monopile design and system dynamic analysis at the calibrated site.In this study, an automatic calibration procedure was implemented which would directly read the extracted results from the ABAQUS simulations and calibrate the PISA spring parameters based on the results for this site.Figure 7 shows the computed monopile response from the ABAQUS simulation, beam-spring simulation using the default PISA spring parameters, and calibrated PISA based on the FE results at Therefore, it can be concluded that the PISA design approach could satisfactorily capture the monopile response in the layered soil, although site-based 3D FE simulations were required to calibrate the model spring parameters.Using the default PISA spring parameters can significantly overpredict the foundation response, leading to an unsafe foundation design.

Conclusions
This study presented an example case study of a monopile design in layered soil from a typical offshore wind farm in China.Detailed interpretations of in-situ data and advanced lab testing data were provided to define the soil stratigraphy and characterize the mechanical properties of clay and sand layers at the site.Based on the geotechnical data interpretation, design soil parameter profiles were proposed for the monopile response calculation.Then, an advanced 3D FEM simulation of the monopile foundation at the site was performed in Abaqus.The PISA spring parameters were then calibrated against the 3D FE simulation results and used in the corresponding beam-spring analyses.It

Figure 1 .
Figure 1.CPT data at the monopole site (HDEC, 2023a): (a) qc, fs, and u2 profiles; (b) Qt, Fr, and Bq profiles.In the same figure, the soil classification results interpreted from the classification chart in [7] using the normalized CPT parameters are also plotted alongside the CPT data.As show in the figure, low Qt (< 10), and Fr (< 1) values, but high Bq values of approximately 0.5 were observed in the shallow (20-m) layer, implying a soft clay layer, which was also confirmed as a clay-like soil from the soil classification method in[7].A similar response was also identified in the range of approximately 37 to 47 m.For other depth ranges, such as 20 to 37 m, and below 47 m, high Qt (> 20) and negative Bq values were noticed, suggesting a sand-like soil with some dilatancy.

Figure 2 .
Figure 2. Design soil strength and stiffness parameter profiles.The nonlinear stress-strain response of the soil was modeled using the NGI-ADP model[10].The model features a non-linear hardening law that links the current yield stress with the current plastic shear strain.The hardening law parameters can be calibrated against site-specific stress-strain response measured in soil laboratory tests.The model is therefore suitable for both undrained ultimate capacity and deformation analyses.For the clay and sand soil layers at this site, it was found that their physical properties were very similar to the representative soil units in the HDEC database, i.e., South Sea Clay 2 and Jiangsu silty sand[11,12].Therefore, the stress-strain curves of South Sea Clay 2 and Jiangsu silty sand were adopted to represent those of clay and silty sand soil layers at the site, respectively.The calibrated NGI-ADP parameters for these soils, together with a comparison between the measured data in the lab tests and computed results from the NGI-ADP mode, are represented in figure3.

Figure 6 .
Figure 6.Illustration of the PISA design method.Figure7shows the computed monopile response from the ABAQUS simulation, beam-spring simulation using the default PISA spring parameters, and calibrated PISA based on the FE results at

Figure 7 .
Figure 7.Comparison of computed monopile responses between ABAQUS and calibrated PISA springs: (a) pile response at mudline; (b) pile response in the soil.Therefore, it can be concluded that the PISA design approach could satisfactorily capture the monopile response in the layered soil, although site-based 3D FE simulations were required to calibrate the model spring parameters.Using the default PISA spring parameters can significantly overpredict the foundation response, leading to an unsafe foundation design.