Skip to content

DOMAIN MAP 3: Centrifuge Modelling, Physical Testing, and Experimental Geotechnics

Synthesised: 2026-04-17 | Source: 45 batch summary files (batch01_agent1 through batch09_agent5) | ~1,200 papers scanned

Domain at a glance

mindmap
  root((D3 · Centrifuge<br/>1 200 papers))
    Scaling laws
      1-g stress mismatch
      Ng centrifuge stress match
      Model-scale corrections
      Time scaling conflict
      Pore-fluid scaling
    Research groups
      KAIST 100g / 70g
      Cambridge Schofield
      UWA Oh Randolph
      ETH Weber Martakis
      Oxford Houlsby Byrne
      Aalborg Nielsen
      Delft Gavin
    Model sand character
      UWA superfine silica
      SNU silica
      Fontainebleau
      Chow 2019 UWA SF sand
    Foundations tested
      Monopile many studies
      Suction caisson Cox 2014
      Tripod bucket DJ Kim 2014
      Gravity base fewer
      Skirted / spudcan
    Scour experiments
      Local scour most common
      Global scour fewer
      Uniform S/D simplification
      Asymmetric scour gap
      Backfill after scour gap
    Cyclic protocols
      LeBlanc 8k-60k cycles
      Cox 12k cycles
      Barari self-healing
      Garala kinematic
    Standards
      ISSMGE centrifuge
      ASCE physical modelling
      DNV test requirements
    Relevance to PhD
      J1 centrifuge tripod 70g
      J3 saturation + backfill 70g
      Data feeds J2 J5 E A

Centrifuge test protocol schematic

flowchart TB
    Prep[Sample preparation<br/>SNU silica sand<br/>target Dr]:::p
    Mount[Model mounting<br/>tripod bucket<br/>instrumentation]:::m
    Spin[Spin-up to target g<br/>70g for tripod<br/>stress equilibrium]:::s
    Base[Baseline measurement<br/>impulse release<br/>frequency extraction]:::b
    Scour[Scour stage<br/>S/D = 0.2 0.4 0.58<br/>excavation at 1g or Ng]:::sc
    Test[Dynamic test<br/>impulse · sine sweep<br/>square-wave pulse]:::t
    Backfill[Backfill stage<br/>optional<br/>recovery measurement]:::f
    Analyse[Data analysis<br/>f1 f2 damping<br/>LVDT displacement]:::a

    Prep --> Mount --> Spin --> Base --> Scour --> Test --> Backfill --> Analyse

    classDef p fill:#f5f5f5,stroke:#666
    classDef m fill:#e1f5fe,stroke:#0277bd
    classDef s fill:#e8eaf6,stroke:#3f51b5
    classDef b fill:#e8f5e9,stroke:#2e7d32
    classDef sc fill:#ffebee,stroke:#c62828
    classDef t fill:#fff3e0,stroke:#e65100
    classDef f fill:#f3e5f5,stroke:#7b1fa2
    classDef a fill:#fff8e1,stroke:#f9a825

Research groups contributing to tripod-bucket evidence

%%{init: {"theme":"base","themeVariables":{"fontSize":"15px","fontFamily":"Inter, system-ui, -apple-system, sans-serif","primaryTextColor":"#1a1a1a","lineColor":"#888"},"flowchart":{"nodeSpacing":22,"rankSpacing":42,"padding":14,"useMaxWidth":true}}}%%
flowchart TB
    subgraph Groups ["&nbsp;<b>Research groups</b>&nbsp;"]
        direction LR
        KAIST["<b>KAIST</b><br/>100g · 70g"]:::kr
        UWA["<b>UWA</b><br/>drum + beam"]:::au
        Camb["<b>Cambridge</b><br/>Schofield 5g"]:::uk
        ETH["<b>ETH Zurich</b><br/>beam"]:::ch
        Oxf["<b>Oxford</b><br/>beam + swing"]:::uk
    end

    subgraph Output ["&nbsp;<b>Experimental datasets produced</b>&nbsp;"]
        direction LR
        Mono["<b>Monopile data</b>"]:::out
        Caisson["<b>Suction caisson data</b>"]:::out
        Tripod["<b>Tripod bucket</b><br/>🔴 THIS PROGRAMME"]:::hot
    end

    KAIST ==> Tripod
    KAIST --> Mono
    UWA --> Mono
    UWA --> Caisson
    Camb --> Mono
    Camb --> Caisson
    ETH --> Mono
    Oxf --> Caisson
    Oxf --> Mono

    classDef kr fill:#fff3e0,stroke:#e65100,stroke-width:2.5px,color:#e65100
    classDef au fill:#e3f2fd,stroke:#1976d2,stroke-width:2px
    classDef uk fill:#f3e5f5,stroke:#6a1b9a,stroke-width:2px
    classDef ch fill:#e8f5e9,stroke:#2e7d32,stroke-width:2px
    classDef out fill:#fafafa,stroke:#666,stroke-width:2px
    classDef hot fill:#ffebee,stroke:#c62828,stroke-width:3px,color:#b71c1c

    style Groups fill:#fafafa,stroke:#bdbdbd,stroke-dasharray:5 5
    style Output fill:#fff8e1,stroke:#ffe082,stroke-dasharray:5 5

1. ESTABLISHED KNOWLEDGE

1.1 Centrifuge Modelling as the Validation Standard

Centrifuge testing is universally accepted as the gold standard for stress-level-accurate physical modelling of offshore foundations. Across the entire literature corpus (1988-2025), every numerical study of suction buckets, monopiles, or scour effects that claims validation relies on centrifuge data. The method appears in over 80 papers spanning six research groups: KAIST (Kim, Seong, Choo, Jeong), UWA (Chow, Fu, Gourvenec, Cassidy), Cambridge (Madabhushi, Garala, Kariyawasam, Morley), Oxford (Byrne, Villalobos, Houlsby, Bienen, McAdam), ETH Zurich (Taeseri, Martakis), and TU Delft (Senders). Typical acceleration levels range from 20g to 100g, with 50g-80g being the most common for offshore wind applications.

Key established facts:

  • Suction caissons can be installed by suction in both sand and clay with drastic reduction in penetration force versus jacking (Villalobos 2006, Byrne 2000, Houlsby et al. 2005/2006 field trials).
  • Tripod bucket foundations yield at half the moment capacity of monopod buckets but at only 20% of the monopod rotation angle (DJ Kim et al. 2014, centrifuge at 100g in KAIST silty sand).
  • Cyclic loading below the monotonic yield moment produces negligible permanent deformation for suction caissons (DJ Kim 2014, Byrne 2000, Cox et al. 2014). Once exceeded, cumulative plastic rotation follows a power law with cycle count.
  • Rotational stiffness of suction caissons increases logarithmically with number of cycles, while accumulated rotation follows a power law (Cox et al. 2014, 12,000 cycles at 1:200 scale).
  • Pile stiffness increases with moderate cyclic loading in sand, contradicting the common degradation assumption (LeBlanc et al. 2010, 8,000-60,000 1g cycles; Almeida 2024, centrifuge showing lateral capacity doubling after 4,800 cycles in quartz sand).
  • Installation method (wished-in-place vs driven vs jacked vs suction) significantly alters subsequent lateral response (Bienen 2021, Fan 2021, CEL large-deformation FEA validated against centrifuge).
  • Suction installation produces large soil-plug heave along interior skirt, while jacking produces negligible displacement (JH Kim & DS Kim 2019, half-bucket centrifuge with PIV).
  • Partially drained behaviour during cyclic loading lies between drained and undrained extremes, and assuming either limit alone is unconservative (Hsu 1998, Villalobos 2006, Stapelfeldt 2020).

1.2 Scour Experiments and Protocols

  • Scour is the No. 1 cause of US bridge collapse (~22/year, Yao & Briaud 2013) and a primary threat to OWT monopile integrity.
  • Horseshoe vortex is the primary scouring mechanism around monopiles; Froude-number-based empirical equations predict equilibrium S/D (Qi & Gao 2014).
  • Cyclic pile movement creates deeper and wider scour holes than static predictions via a multi-stage storm-backfill mechanism (Al-Hammadi & Simons 2019, two-scale flume experiments).
  • Local scour and global scour produce materially different structural effects: local scour preserves some overconsolidation effects in surrounding soil, while global scour removes overburden uniformly (Chortis et al. 2020, centrifuge p-y curves; Ciancimino et al. 2022, hybrid 1g-Ng approach with 3D-printed scour holes).
  • Scour reduces lateral capacity of monopiles (Li et al. 2020/2021, centrifuge) and suction buckets (Chen et al. 2018, 1g model tests in sand; Cheng et al. 2024, 3D FEM in clay). Moment capacity decreases approximately linearly with scour depth for all scour geometries (Li et al. 2021).
  • Scour protection (rock armour, width 2D, 30 kPa overburden) reduces lateral displacement by up to 41% (Q. Li et al. 2024, centrifuge + FE).
  • Eco-friendly scour protection using artificial seaweed mats reduces flow velocity around suction buckets (Kang & Kwon 2025, physical model + field verification). Geotextile sand containers are a viable alternative to rock armour (Hoyme 2018, LCA study). OWF scour protection rock creates artificial reef habitat (Coolen et al. 2019, Tonk et al. 2020 at Borssele V).

1.3 Sand Properties and Characterisation

  • UWA superfine silica sand (Chow 2019) is the international reference sand for centrifuge programmes, with full characterisation of permeability, shear strength (DSS, triaxial, direct shear), and stress-dilatancy parameters.
  • Fraction E sand (Cambridge) and Hostun HN31 sand (French/European programmes) serve as secondary standard sands.
  • The Hardin equation and modified hyperbolic G/G0 degradation curve (Oztoprak & Bolton 2013, 454 lab tests, 3860 data points) provide the accepted framework for small-strain stiffness. Three parameters are required: elastic threshold strain, reference strain, and curvature coefficient.
  • In-flight CPT and bender element measurements are standardised for centrifuge use (KOCED standards 2019-2023; JH Kim et al. 2016/2017 at KAIST; Lee et al. 2018 seismic CPT). Empirical Vs-qc correlations exist for normally consolidated uncemented silica sand (JH Kim et al. 2017).
  • Particle size effects on miniature cone tip resistance are negligible for 7-13 mm diameter cones (JH Kim et al. 2016). Tip resistance decreases with g-level above critical depth.
  • Carbonate sand shows greater stiffness degradation under cyclic lateral loading than quartz sand (Almeida 2024, centrifuge). Calcareous silt penetrometer resistance varies by factor 4.1 between drained and undrained conditions (Chow et al. 2019, centrifuge T-bar/piezocone).

1.4 Combined V-H-M Loading Framework

The coupled treatment of vertical, horizontal, and moment loads as a single failure envelope is the universally accepted framework for offshore foundation design, established by the Oxford/Cambridge/UWA lineage: Butterfield & Gottardi (1994, 3D cigar-shaped envelope on sand), Martin (1994), Cassidy (1999), Byrne (2000), Villalobos (2006), Bienen (2007), Fu (2017), and Gourvenec (2007, 2008). Work-hardening plasticity provides the most successful constitutive framework. Standardised sign conventions were adopted from Butterfield et al. (1997).

2. ACTIVE FRONTIERS (2023-2025)

2.1 Scour + Dynamics Coupling for OWT Foundations

The frontier is coupling scour morphology with structural dynamic response for real-time monitoring. Key advances:

  • Natural-frequency-based scour quantification: Weil et al. (2023) demonstrated digital-twin + automated OMA for quantifiable (not just detectable) scour assessment. Smith et al. (2023) used hierarchical Bayesian inference across a wind farm population to detect scour as an anomaly. Abdelhak et al. (2024) demonstrated unsupervised novelty detection from acceleration data.
  • Centrifuge validation of scour-frequency sensitivity: Kim et al. (2025) measured max 5.3% frequency reduction at scour depth = 0.6D for tripod suction buckets; multi-footing arrangement redistributes stiffness loss. Kariyawasam et al. (2020, Cambridge centrifuge) found up to 40% frequency variation for 30% embedment loss.
  • Coupled 3D-to-1D frameworks: Kim et al. (2026, submitted) mapped 3D FEM scour results to 1D BNWF models, finding 48% embedment loss causes only 6.1% frequency drop but crosses 1P boundary at 4.5 m scour.
  • Scour under seismic loading: Xu & Madabhushi et al. (2024) compared rock-berm scour protection with plate foundation under seismic liquefaction in dynamic centrifuge tests. Zhang, Jia et al. (2025) showed scour protection slightly increases 1st natural frequency under earthquakes.

2.2 Silent Piling and Novel Foundation Installation

Screw piles installed by rotary jacking at low reaction force are an active frontier for noise-free offshore installation. Cerfontaine et al. (2021/2023) demonstrated feasibility via centrifuge + DEM, showing that advancement ratio controls performance and over-flighting (AR < 1.0) enhances cyclic uplift capacity (Wang et al. 2023, centrifuge + DEM).

2.3 Physics-Informed Digital Twins for Foundation Monitoring

Branlard et al. (2024) validated a physics-based Kalman-filter digital twin for the TetraSpar FOWT at 10-15% error on fatigue loads versus field measurements. Zou et al. (2022) achieved the first in-situ validation of GPLFM for below-mudline virtual strain sensing at Westermeerwind Park. Simpson et al. (2024) applied Bayesian model updating to reduce uncertainty in OWT foundation stiffness from monitoring data.

2.4 Data-Driven Macroelement Models

Lopez et al. (2025) developed an ANN-based macroelement predicting suction bucket response in 3D V-H-M space with excellent generalisation. Jin et al. (2025) proposed a novel hypoplastic macroelement incorporating a single scour parameter, showing that scour depth dominates capacity loss while failure envelope shape remains unchanged.

2.5 Real-Time Hybrid Simulation

Al-Subaihawi et al. (2024) demonstrated a coupled aero-hydro-geotech RTHS framework linking OpenFAST with a physical soil model for real-time nonlinear foundation interaction. Song et al. (2020) conceptualised the partitioning of OWT into physical (foundation) and numerical (tower+turbine) substructures.

3. CONTESTED CLAIMS

3.1 p-y Curve Adequacy for Large-Diameter Monopiles

Standard API/DNV p-y curves remain the industry default but are consistently found inadequate for large-diameter (>4 m) monopiles. Haiderali et al. (2023) showed p-y is inaccurate for XL/XXL monopiles in clay. Choo & Kim (2015) found API/Reese overestimates initial stiffness from centrifuge tests. PISA-based p-y curves (Sastre Jurado et al. 2022, McAdam et al. 2023) yield substantially better agreement with monitoring data. However, p-y curves remain entrenched in classification society codes, creating a theory-practice gap.

3.2 Local versus Global Scour Equivalence

The common design simplification of treating local scour as equivalent to removing soil uniformly (global scour) is contested. Qi et al. (2016, centrifuge 1:250) showed local scour yields stiffer p-y response than global scour at the same depth below scour base due to overconsolidation effects. Ciancimino et al. (2022) found local scour reduces moment capacity by up to 38% versus 48% for general scour. The appropriate effective-depth correction remains unresolved.

3.3 Drainage Condition During Storm Loading

Villalobos (2006) found undrained moment capacity under low vertical loads was "very small" for suction caissons, while Byrne (2000) showed meaningful tensile resistance under partially drained conditions. Nielsen et al. (2017) found that under partly drained conditions at 1 Hz, two-way loading produces larger rotations than one-way, contrary to drained findings. The transition between drained and undrained behaviour remains difficult to predict at prototype scale.

3.4 Constitutive Model Selection for Cyclic SSI

No single constitutive model satisfies all cyclic loading requirements. Orakci et al. (2024) compared SANISAND-MS and PM4SAND for monopile ratcheting with different accuracy and calibration complexity. Cudny & Truty (2020) addressed the overshooting problem in HSS via HS-Brick, while Wang (2019) highlighted principal stress rotation effects. Sorum et al. (2022) showed macro-element vs p-y model choice causes materially different fatigue predictions.

3.5 Dimensional Analysis Sufficiency for Scaling

Classical Buckingham Pi analysis is challenged by Atar et al. (2021) and Davey & Ochoa-Cabrero (2023), who propose finite similitude theory offering countably infinite scaling alternatives. Meanwhile, Buckingham Pi continues to yield high predictive accuracy in specific applications (<20% error for proppant distribution, >98% for heat transfer, R2=0.85 for bridge scour depth via CNN, Huynh 2025). The disagreement is domain-specific: scale effects in geomechanics (stress-level dependence, grain size effects) make classical scaling more problematic than in fluid mechanics.

4. VERIFIED GAPS

4.1 Combined Multi-Hazard Loading on OWT Foundations

Individual hazards (scour, seismic, cyclic wind-wave, liquefaction, corrosion) are studied in isolation across >50 papers. No study addresses the full coupled problem of scour + cyclic fatigue + seismic + corrosion under long-term operation. Jia et al. (2024) combined seismic + scour for buckets; Kontoni & Farghaly (2023) addressed wind + wave + earthquake with TMDs; Cao et al. (2024) studied scour + fatigue. The full multi-hazard envelope remains unexplored.

4.2 Prototype-Scale Validation of Centrifuge Results

Universally acknowledged since the earliest papers (Byrne 2000, Villalobos 2006). Field trials at Bothkennar (Houlsby et al. 2005, clay) and Luce Bay (Houlsby et al. 2006, sand) remain among the very few prototype-scale suction caisson datasets. Shonberg et al. (2017) reported monitoring data from the first full-scale suction bucket jacket at Borkum Riffgrund 1. Systematic field-scale validation campaigns bridging centrifuge to prototype remain scarce.

4.3 Long-Term Cyclic Degradation Under Realistic Loading

Most experimental studies apply hundreds to thousands of uniform cycles. Real OWT foundations experience millions of low-amplitude variable-direction cycles over 25+ years. Lombardi et al. (2013) reached 172,000 cycles; LeBlanc et al. (2010) 60,000; Cox et al. (2014) 12,000. Nikitas (2020, thesis) explored millions of cycles at 1g but no centrifuge study has achieved this. Zorzi (2020) models cyclic accumulation probabilistically but notes difficulty representing variable amplitude and direction.

4.4 Carbonate Sand Behaviour Under Cyclic Lateral Loading

Almeida (2024) is one of very few centrifuge studies on carbonate sand cyclic response. Tropical/subtropical OWT deployment requires understanding of carbonate sand crushability, cementation, and post-cyclic degradation, which remains poorly characterised compared to quartz sand.

4.5 CPT Interpretation for Silty/Transitional Soils

Fioravante et al. (2022) explicitly states no robust CPTU interpretation methodology exists for silty soils, which dominate many offshore wind sites. Calibration chamber work is ongoing but guidelines do not yet exist. Lehane et al. (2023) developed centrifuge CPT scaling corrections but only for clean sands.

4.6 Vibration-Based Scour Monitoring Validated on OWT Monopiles

All mature vibration-based scour detection systems target bridge piers (Prendergast et al. 2013-2017, Kariyawasam 2020, Kawabe & Kim 2023). Field deployment on OWT monopiles or suction buckets has not been validated, despite numerical demonstrations (Jawalageri 2022, Abdelhak 2024, Kim et al. 2025/2026).

4.7 Backfill Characterisation After Scour Events

No systematic experimental programme characterises the mechanical properties of naturally backfilled scour holes. Al-Hammadi & Simons (2019) documented multi-stage storm-backfill mechanisms in flume experiments, but the resulting soil properties (density, stiffness) of backfilled material versus virgin ground are unquantified.

5. QUANTITATIVE BENCHMARKS

5.1 Foundation Geometry and Capacity

Parameter Value Source
Tripod bucket dimensions (prototype) D=6.5 m, L=8.0 m, c-c=26.9 m DJ Kim 2014 (100g centrifuge)
Tripod yield rotation vs monopod 20% of monopod rotation DJ Kim 2014
Tripod yield moment vs monopod 50% of monopod moment DJ Kim 2014
Hybrid bucket capacity improvement 1.91x vertical, 1.82x combined vs conventional JH Kim et al. 2020 (centrifuge)
Cyclic stiffness increase (suction caissons) Logarithmic with cycle count Cox et al. 2014 (12,000 cycles)
Lateral capacity doubling after cycling ~2x after 4,800 cycles (quartz sand) Almeida 2024 (centrifuge)
Stiffness increase (monopile, first 500 cycles) Exponential growth Almeida 2024
Clustered bucket lateral-moment improvement Up to 58% over monopod Choo et al. 2014 (centrifuge)
Fast loading rate capacity augmentation 1.5x (due to dilation-induced suction) Choo et al. 2021 (70g, silty sand)
Suction caisson critical uplift displacement ~0.02D (mono), ~0.01D (group) Zhu et al. 2019 (centrifuge, clay)

5.2 Scour Effects

Parameter Value Source
Frequency reduction at 0.6D scour (tripod bucket) Max 5.3% Kim et al. 2025 (centrifuge)
Frequency variation for 30% embedment loss (bridge) Up to 40% Kariyawasam et al. 2020 (centrifuge)
Moment capacity reduction from local scour Up to 38% Ciancimino et al. 2022
Moment capacity reduction from general scour Up to 48% Ciancimino et al. 2022
Scour protection lateral displacement reduction Up to 41% (width 2D, 30 kPa) Q. Li et al. 2024
Fatigue life reduction from scour (1.3D depth) ~24% Cao et al. 2024
Fatigue life increase from TMD (1% mass ratio) ~65% Cao et al. 2024
Bridge collapses/year from scour (US) ~22 Yao & Briaud 2013
Natural frequency at scour=4.5 m crossing 1P Threshold for tripod bucket Kim et al. 2026 (submitted)

5.3 Sand Properties

Parameter Value Source
G/G0 degradation database 454 tests, factor-1.13 accuracy for 3860 points Oztoprak & Bolton 2013
Vs-qc correlation for NC silica sand Empirical (centrifuge validated) JH Kim et al. 2017
Penetrometer drained/undrained ratio (calcareous silt) 4.1x Chow et al. 2019
Consolidation-induced strength gain (calcareous silt) 2.5x Chow et al. 2019
Liquefaction resistance increase with age ~40% after 400 years Towhata et al. 2016

5.4 Design and Cost

Parameter Value Source
Foundation cost share of OWT 15-40% of total Houlsby & Byrne 2000; Kim 2014 (25-33%)
Monopile market share for OWT 77% Abdelhak 2024
Monitored natural frequency vs design 5-15% higher than predicted Stuyts et al. 2022/2023
Fatigue damage variation from foundation model choice Up to 22% (Aasen 2017); up to 180% (Katsikogiannis 2019)
Soil damping ratio range for monopile OWT 0.17-1.3% of critical Rezaei et al. 2018
LFFD contribution to total fatigue Up to 65% Sadeghi et al. 2023
Dense sand crust for rocking isolation z/B = 1 sufficient Tsatsis & Anastasopoulos 2015

5.5 Centrifuge Testing Standards (KOCED/Korean)

Standard Content
KOCED 0005-7383 (2019) Air pluviation method for sand ground preparation
KOCED 0013-7411 (2021) Compaction method for sand models
KOCED 0021-xxxx (2020) Miniature CPT for dry sand evaluation
KOCED 0029 (2020) Bender element Vs measurement in centrifuge
KOCED 0048-7586 (2023) Debris flow inclined flume measurement

Synthesised from 45 batch summary files covering ~1200 papers (1902-2026). Focus: centrifuge modelling, physical model testing, scour experiments, sand properties, backfill/scour protection, and soil characterisation.