MASTER KNOWLEDGE MAP

Dissertation: Scour Assessment of Offshore Wind Turbine Tripod Suction Bucket Foundations Synthesised from: 5 domain maps covering 1,952 papers (1981--2026) Generated: 2026-04-17

The map in one mindmap

mindmap
  root((Master Knowledge Map<br/>1 952 papers<br/>5 domains))
    Field consensus
      V-H-M envelopes govern
      SSI controls OWT frequency
      Monitored f 5-15% above design
      Scour lowers frequency
      Global scour > local
      Centrifuge is gold standard
      Cyclic stiffens sand
      PISA > API p-y
      Fatigue varies ±180%
      OMA SSI/FDD dominant
    Open debates
      Frequency vs mode shape
      Local vs general scour rule
      Physics vs data-driven DT
      Constitutive for cyclic SSI
      10-min DLC vs long-term fatigue
    Verified gaps
      G1 scour + cyclic [CRITICAL]
      G2 bucket scour [CRITICAL]
      G3 field OWT SHM [CRITICAL]
      G4 closed-loop DT
      G5 probabilistic capacity
      G6 silty mixed soils
      G7 prototype validation
      G8 EOV benchmark
      G9 SHM encoder
      G10 unified framework
    Portfolio firsts
      Centrifuge tripod scour [J1]
      Field tripod scour [V1]
      Probabilistic tripod [J5]
      EOV benchmark [V2]
      Cross-soil encoder [E]
      Bayesian loop [A]
    11 papers
      Theory
        J11
      Numerical
        J2
        Op3
      Experimental
        J1
        J3
      Probabilistic
        J5
      Field
        V1
      SHM theory
        V2
      ML
        E
        B
      Decision
        A

Dissertation architecture — four stacked layers

Each paper produces something the next consumes. A weakness at any layer propagates upward.

flowchart TB
    subgraph L1 [" LAYER 1 · MECHANICS "]
        J11[J11<br/>Vesic cavity<br/>generalisation]
        J1[J1<br/>Centrifuge<br/>scour-frequency]
        J3[J3<br/>Centrifuge<br/>saturation + backfill]
    end

    subgraph L2 [" LAYER 2 · MODELLING "]
        J2[J2<br/>3D FE-calibrated<br/>Winkler]
        J5[J5<br/>Monte Carlo<br/>1 794 realisations]
        Op3[Op³<br/>Open-source<br/>pipeline]
    end

    subgraph L3 [" LAYER 3 · REPRESENTATION "]
        E[E<br/>Physics-informed<br/>encoder]
    end

    subgraph L4 [" LAYER 4 · MONITORING & DECISION "]
        V1[V1<br/>32-month field<br/>monitoring]
        V2[V2<br/>State-function<br/>EOV benchmark]
        B[B<br/>Buckingham-Pi<br/>feature ranking]
        A[A<br/>Bayesian<br/>decision loop]
    end

    J11 -- spring shape --> J2
    J2 -- PL-1 coefficients --> V1 & A & Op3
    J3 -- validation --> J2
    Op3 -- runs --> J5
    J5 -- capacity distribution --> A
    V1 -- detection threshold --> A
    B -- feature ranking --> V2 & A
    V2 -- compensation --> V1 & B
    J5 -- database --> E
    E -- latent state --> A

    style L1 fill:#f3e5f5,stroke:#7b1fa2
    style L2 fill:#e3f2fd,stroke:#1565c0
    style L3 fill:#fff8e1,stroke:#f9a825
    style L4 fill:#e8f5e9,stroke:#2e7d32

Portfolio × verified-gaps matrix

Filled dots = direct contribution. J1 and J11 inherit G2 (bucket scour) as their primary gap; G1 (scour + cyclic) and G4 (closed-loop DT) are covered by a single paper each — the dissertation's thinnest coverage.

%%{init: {"theme":"base","themeVariables":{"fontSize":"15px","fontFamily":"Inter, system-ui, -apple-system, sans-serif","primaryTextColor":"#1a1a1a","lineColor":"#888"},"flowchart":{"nodeSpacing":20,"rankSpacing":35,"padding":12,"useMaxWidth":true}}}%%
flowchart TB
    subgraph Papers ["&nbsp;<b>11 portfolio papers</b>&nbsp;"]
        direction LR
        J1((J1)):::pub
        J2((J2)):::pap
        J3((J3)):::pap
        J5((J5)):::pap
        J11((J11)):::pap
        V1((V1)):::pap
        V2((V2)):::pap
        B((B)):::pap
        A((A)):::pap
        E((E)):::pap
        Op3((Op3)):::pap
    end

    subgraph Gaps ["&nbsp;<b>10 verified gaps</b>&nbsp;"]
        direction TB
        G1["<b>G1</b> · scour + cyclic · 🔴 CRITICAL"]:::crit
        G2["<b>G2</b> · bucket scour · 🔴 CRITICAL"]:::crit
        G3["<b>G3</b> · field OWT SHM · 🔴 CRITICAL"]:::crit
        G4["<b>G4</b> · closed-loop DT · 🟠 HIGH"]:::high
        G5["<b>G5</b> · probabilistic capacity · 🟠 HIGH"]:::high
        G6["<b>G6</b> · silty/mixed soils · 🟠 HIGH"]:::high
        G7["<b>G7</b> · prototype validation · 🟠 HIGH"]:::high
        G8["<b>G8</b> · EOV benchmark · 🟣 MED"]:::med
        G9["<b>G9</b> · SHM encoder · 🟣 MED"]:::med
        G10["<b>G10</b> · unified framework · 🟣 MED"]:::med
    end

    J1 --> G2
    J1 --> G6
    J2 --> G2
    J3 --> G1
    J3 --> G2
    J3 --> G6
    J5 --> G2
    J5 --> G5
    J11 --> G2
    V1 --> G3
    V1 --> G7
    V2 --> G3
    V2 --> G8
    B --> G8
    A --> G4
    A --> G5
    E --> G9
    Op3 --> G10

    click J1 "../papers/J1" "Open Paper J1"
    click J2 "../papers/J2/" "Open Paper J2 (full manuscript)"
    click J3 "../papers/J3/" "Open Paper J3 (full manuscript)"
    click J5 "../papers/J5/" "Open Paper J5 (full manuscript)"
    click J11 "../papers/J11/" "Open Paper J11 (full manuscript)"
    click V1 "../papers/V/" "Open Paper V (V1/V2 consolidated)"
    click V2 "../papers/V/" "Open Paper V (V1/V2 consolidated)"
    click B "../papers/B/" "Open Paper B (full manuscript)"
    click A "../papers/A/" "Open Paper A (full manuscript)"
    click E "../papers/E/" "Open Paper E (full manuscript)"
    click Op3 "../papers/Op3/" "Open Paper Op3 (full manuscript)"

    classDef crit fill:#ffebee,stroke:#c62828,stroke-width:2.5px,color:#b71c1c
    classDef high fill:#fff3e0,stroke:#ef6c00,stroke-width:2px,color:#e65100
    classDef med fill:#f3e5f5,stroke:#7b1fa2,stroke-width:2px,color:#4a148c
    classDef pap fill:#fff,stroke:#444,stroke-width:1.5px
    classDef pub fill:#fff,stroke:#2e7d32,stroke-width:3px,color:#1b5e20

    style Papers fill:#fafafa,stroke:#bdbdbd,stroke-dasharray:5 5
    style Gaps fill:#fff8f0,stroke:#ffccbc,stroke-dasharray:5 5

Coverage by gap

%%{init: {"theme": "base"}}%%
pie showData
    title Papers per gap (primary contributions)
    "G2 bucket scour (centre of gravity)" : 5
    "G3 field OWT SHM" : 2
    "G6 silty/mixed soils" : 2
    "G5 probabilistic capacity" : 2
    "G8 EOV benchmark" : 2
    "G1 scour + cyclic" : 1
    "G4 closed-loop DT" : 1
    "G7 prototype validation" : 1
    "G9 SHM encoder" : 1
    "G10 unified framework" : 1

Field debates map

Where the field has not converged. Each debate is mapped to which of the 11 papers contributes evidence.

%%{init: {"theme":"base","themeVariables":{"fontSize":"15px","fontFamily":"Inter, system-ui, -apple-system, sans-serif","primaryTextColor":"#1a1a1a","lineColor":"#666"},"flowchart":{"nodeSpacing":25,"rankSpacing":38,"padding":14,"useMaxWidth":true}}}%%
flowchart TB
    subgraph D1 ["&nbsp;<b>Debate 1 · frequency vs mode shape</b>&nbsp;"]
        direction TB
        S1a[Prendergast · Li<br/>Weijtjens · Kim]:::side -.-> F1["<b>Frequency side</b>"]:::pos
        S1b[Kariyawasam<br/>Malekjafarian]:::side -.-> M1["<b>Mode-shape side</b>"]:::pos
        V1_1((V1)):::paper -- contributes --> F1
        V2_1((V2)):::paper -- contributes --> F1
    end

    subgraph D2 ["&nbsp;<b>Debate 2 · local vs general scour</b>&nbsp;"]
        direction TB
        S2a["DNV 1.3D<br/>uniform rule"]:::side -.-> U2["<b>Uniform side</b>"]:::pos
        S2b[Ciancimino · Qi]:::side -.-> Diff2["<b>Differentiated side</b>"]:::pos
        J2_1((J2)):::paper -- evidence --> Diff2
        J3_1((J3)):::paper -- evidence --> Diff2
    end

    subgraph D3 ["&nbsp;<b>Debate 3 · physics vs data-driven DT</b>&nbsp;"]
        direction TB
        S3a[Ritto · Cross<br/>Branlard]:::side -.-> Phy3["<b>Physics-first</b>"]:::pos
        S3b[Zhong · Zhang<br/>Stadtmann]:::side -.-> Data3["<b>Data-first</b>"]:::pos
        S3c[Bull 2025]:::side -.-> Ens3["<b>Ensemble</b>"]:::pos
        A_1((A)):::paper -- hybrid --> Ens3
        E_1((E)):::paper -- physics-informed --> Phy3
    end

    D1 ==> D2 ==> D3

    classDef side fill:#fff,stroke:#666,color:#333
    classDef pos fill:#ffecb3,stroke:#f9a825,stroke-width:2px,color:#e65100
    classDef paper fill:#fff,stroke:#2e7d32,stroke-width:2px,color:#1b5e20

    style D1 fill:#e3f2fd,stroke:#1976d2,stroke-dasharray:5 5
    style D2 fill:#fff3e0,stroke:#f57c00,stroke-dasharray:5 5
    style D3 fill:#f3e5f5,stroke:#7b1fa2,stroke-dasharray:5 5

See the numbered debate sections below for sides, resolving evidence, and domains involved.

The ten verified gaps, severity-ranked

%%{init: {"theme":"base","themeVariables":{"fontSize":"15px","fontFamily":"Inter, system-ui, -apple-system, sans-serif","primaryTextColor":"#1a1a1a","lineColor":"#666"},"flowchart":{"nodeSpacing":18,"rankSpacing":25,"padding":14,"useMaxWidth":true}}}%%
flowchart TB
    subgraph Crit ["&nbsp;🔴 <b>CRITICAL severity</b>&nbsp;"]
        G1["<b>G1 · scour + cyclic loading coupling</b>"]:::c
        G2["<b>G2 · scour on suction buckets</b>"]:::c
        G3["<b>G3 · long-term field OWT scour monitoring</b>"]:::c
    end

    subgraph High ["&nbsp;🟠 <b>HIGH severity</b>&nbsp;"]
        G4["<b>G4 · integrated SHM-DT closed loop</b>"]:::h
        G5["<b>G5 · probabilistic capacity for tripod buckets</b>"]:::h
        G6["<b>G6 · scour in cohesive and mixed soils</b>"]:::h
        G7["<b>G7 · prototype-scale validation</b>"]:::h
    end

    subgraph Med ["&nbsp;🟣 <b>MEDIUM severity</b>&nbsp;"]
        G8["<b>G8 · EOV compensation benchmark</b>"]:::m
        G9["<b>G9 · autoencoder for OWT structural state</b>"]:::m
        G10["<b>G10 · unified software framework</b>"]:::m
    end

    Crit ==> High ==> Med

    classDef c fill:#ffebee,stroke:#c62828,color:#b71c1c,stroke-width:3px
    classDef h fill:#fff3e0,stroke:#ef6c00,color:#e65100,stroke-width:2px
    classDef m fill:#f3e5f5,stroke:#7b1fa2,color:#4a148c,stroke-width:2px

    style Crit fill:#fff5f5,stroke:#ffcdd2,stroke-dasharray:5 5
    style High fill:#fffaf0,stroke:#ffe0b2,stroke-dasharray:5 5
    style Med fill:#faf5ff,stroke:#e1bee7,stroke-dasharray:5 5

Six "firsts" the portfolio contributes

%%{init: {"theme":"base","themeVariables":{"fontSize":"15px","fontFamily":"Inter, system-ui, -apple-system, sans-serif","primaryTextColor":"#1a1a1a","lineColor":"#666"},"flowchart":{"nodeSpacing":20,"rankSpacing":30,"padding":14,"useMaxWidth":true}}}%%
flowchart TB
    J1([J1]):::p --> F1["<b>First centrifuge scour-frequency dataset</b><br/>for tripod buckets"]:::f --> G2f["<b>G2</b> bucket scour"]:::g
    V1([V1]):::p --> F2["<b>First field scour-frequency dataset</b><br/>on operational OWT tripod"]:::f --> G3f["<b>G3</b> field OWT SHM"]:::g
    J5([J5]):::p --> F3["<b>First probabilistic capacity ensemble</b><br/>for scoured tripods"]:::f --> G5f["<b>G5</b> probabilistic capacity"]:::g
    V2([V2]):::p --> F4["<b>First EOV-method benchmark</b><br/>on OWT data"]:::f --> G8f["<b>G8</b> EOV benchmark"]:::g
    E([E]):::p --> F5["<b>First physics-informed encoder</b><br/>for cross-soil structural state"]:::f --> G9f["<b>G9</b> SHM encoder"]:::g
    A([A]):::p --> F6["<b>First closed-loop Bayesian decision</b><br/>for scour maintenance"]:::f --> G4f["<b>G4</b> closed-loop DT"]:::g

    classDef p fill:#fff,stroke:#2e7d32,stroke-width:2.5px,color:#1b5e20
    classDef f fill:#f1f8e9,stroke:#558b2f,stroke-width:2px,color:#1b5e20
    classDef g fill:#fff3e0,stroke:#e65100,stroke-width:2px,color:#e65100

---

Paper Index

ID	Short Title	Topic
J1	Kim et al. 2025 (OE, published)	Scour impacts on natural frequency of tripod suction buckets -- centrifuge
J2	Kim et al. (OE submitted)	3D FE-calibrated Winkler model for scour-frequency degradation
J3	Kim et al. (OE submitted)	Centrifuge: saturation and backfill effects on scour response
J5	Kim et al. (MC/OptumGX)	Probabilistic 3D limit analysis -- 1,794 realisations
J11	Kim et al. (Vesic dissipation)	Generalised cavity expansion for tripod buckets
V1	Kim et al. (JCSHM submitted)	32-month field vibration-based scour monitoring
V2	Kim et al. (SHM-SAGE)	EOV compensation method comparison for scour monitoring
E	Kim et al. (Encoder)	Physics-informed encoder for structural-state representation
A	Kim et al. (Decision)	Bayesian scour assessment fusing encoder with field monitoring
B	Kim et al. (Buckingham Pi)	Systematic evaluation of 64 scour-detection features
Op3	Kim et al. (Op3 framework)	Open-source Python coupling OptumGX + OpenSeesPy + OpenFAST

Note: 11 papers identified. Op3 is a software/framework paper supporting the other 10; the coverage matrix below treats all 11.

---

1. THE FIELD'S CONSENSUS

These claims appear in 3 or more of the 5 domain maps with no credible dissent.

1. Combined V-H-M loading governs offshore foundation design. The failure-envelope framework -- not classical superposition -- is the accepted method. Confirmed in D1, D3, D5 (Oxford-Cambridge-UWA lineage: Butterfield 1994 through Fu 2017).

2. Soil-structure interaction controls OWT natural frequency; fixed-base assumptions are unconservative. Errors of 5--30% result from ignoring SSI. Confirmed in D1, D2, D4, D5 (Arany 2016, Stuyts 2022, Zaaijer 2006).

3. Monitored natural frequencies are systematically 5--15% higher than design predictions. This reflects conservative API/DNV p-y stiffness assumptions. Confirmed in D1, D2, D4, D5 (Stuyts 2022/2023, McAdam 2023).

4. Scour degrades foundation performance: reduces stiffness, lowers frequency, accelerates fatigue. Unanimous across monopiles, suction buckets, and bridge piers. Confirmed in all 5 domains.

5. Global scour is more detrimental than local scour for the same nominal depth. Local scour preserves overconsolidation effects. Confirmed in D1, D2, D3, D5 (Li 2020, Qi 2016, Ciancimino 2022).

6. Centrifuge testing is the gold-standard validation method for offshore geotechnics. Universally relied upon across six major research groups. Confirmed in D1, D3, D5.

7. Cyclic loading stiffens sand (contradicting API degradation assumptions). LeBlanc 2010 and Cox 2014 are the anchoring studies. Confirmed in D1, D2, D3, D5.

8. API/DNV p-y curves are inadequate for large-diameter monopiles; PISA is superior. Confirmed in D1, D2, D3, D4, D5.

9. Foundation model choice causes up to 180% variation in predicted fatigue damage. The most consequential modelling uncertainty for OWT lifetime. Confirmed in D1, D2, D4, D5 (Katsikogiannis 2019).

10. Operational modal analysis is the accepted field technique for continuous OWT monitoring. SSI and FDD are dominant algorithms. Confirmed in D2, D4, D5.

---

2. THE FIELD'S FIVE BIGGEST DEBATES

Debate 1: Frequency vs. mode shape as primary scour indicator

• Side A (frequency): Simple, single-sensor, well-validated relationship between frequency drop and scour depth. Championed by Prendergast, Li, Weijtjens, Kim.
• Side B (mode shape): Field frequency variability (EOV) can exceed scour-induced shifts, making frequency unreliable without compensation. Mode-shape ratios are more sensitive and partially self-compensating for temperature. Championed by Kariyawasam, Malekjafarian, Khan, Jawalageri.
• Resolving evidence: A field deployment with co-located scour bathymetry and multi-sensor modal extraction over multiple years would settle this. V1 and V2 contribute frequency-side evidence; no equivalent mode-shape field dataset exists for OWT.
• Domains: D1, D2, D5.

Debate 2: Local vs. general scour treatment in design codes

• Side A (conservative simplification): Treat all scour as uniform mudline lowering (DNV 1.3D rule). Simple, conservative for most cases.
• Side B (differentiated treatment): Local and general scour produce fundamentally different failure mechanisms (38% vs. 48% capacity loss per Ciancimino 2022) and different p-y responses (Qi 2016). A single rule is either non-conservative or wasteful depending on geometry.
• Resolving evidence: Standardised correction factors validated against field scour surveys and centrifuge data for multiple foundation types. J2 and J3 provide tripod-specific data that did not previously exist.
• Domains: D1, D2, D3, D5.

Debate 3: Physics-based vs. data-driven digital twins

• Side A (physics-first): Physics models generalise beyond training data, are interpretable, and transfer across sites. Grey-box models with GP kernels recommended. Championed by Ritto, Cross, Branlard.
• Side B (data-first): Deep learning surrogates achieve 500x speedup and high within-domain accuracy. Physics models require calibration that may be unavailable. Championed by Zhong, Zhang, Stadtmann.
• Side C (ensemble): Bull (2025) argues for probabilistic ensembles of both, not an either-or choice.
• Resolving evidence: Head-to-head comparison on the same OWT with field truth over multiple years. Paper A implicitly adopts a hybrid approach; no head-to-head benchmark exists.
• Domains: D2, D4, D5.

Debate 4: Constitutive model selection for cyclic SSI in sand

• Side A: SANISAND-MS captures fabric evolution but is calibration-intensive.
• Side B: PM4SAND is simpler but less accurate for ratcheting.
• Side C: Macro-element models bypass constitutive complexity but lack code acceptance.
• Resolving evidence: Boundary-value-problem-level validation against field-monitored cyclic response -- not just element tests. No such dataset exists.
• Domains: D1, D3, D4, D5.

Debate 5: Adequacy of the 10-minute DLC framework for fatigue

• Side A (status quo): DNV/IEC 10-minute simulations with short-term statistics suffice; correction factors handle long-term effects.
• Side B (reform): Sadeghi (2023) shows up to 65% of fatigue damage comes from low-frequency dynamics (periods > 1 day), invisible in 10-minute windows. Fundamental data segmentation reform may be needed.
• Resolving evidence: Multi-year continuous fatigue monitoring with simultaneous 10-minute DLC predictions and long-term load reconstruction. V1's 32-month dataset could contribute if fatigue-specific analysis is added.
• Domains: D2, D4, D5.

---

3. THE FIELD'S TEN VERIFIED GAPS

Ranked by consequence for the dissertation topic, with confirmation across domain maps.

#	Gap	Confirming Domains	Severity
G1	Combined scour + cyclic loading on OWT foundations. Scour studies ignore cycling; cyclic studies ignore scour. No coupled experimental or numerical investigation exists.	D1, D2, D3, D4, D5	Critical
G2	Scour effects on suction bucket/caisson foundations. Research overwhelmingly targets monopiles and bridge piers. Only ~5 papers address bucket-specific scour.	D1, D2, D3, D5	Critical
G3	Long-term field validation of frequency-based scour monitoring for OWT. All prior work is laboratory, centrifuge, or bridge-only. No operational OWT system has been published.	D1, D2, D3, D4, D5	Critical
G4	Integrated scour-SHM-digital-twin closed loop. Digital twins, scour monitoring, and reliability updating exist independently but have never been connected for OWT.	D2, D4, D5	High
G5	Probabilistic scour capacity assessment for tripod suction buckets. All published assessments use deterministic single-point soil profiles. Population-scale uncertainty quantification is absent.	D1, D4	High
G6	Scour in cohesive and mixed (silty) soils. Most scour studies use clean sand. Silty sand (Yellow Sea conditions) and clay scour are poorly characterised.	D1, D2, D3, D5	High
G7	Prototype-scale validation of centrifuge-derived design methods. Universally called for since the 1990s, still lacking for suction buckets beyond Bothkennar/Luce Bay/Borkum Riffgrund 1.	D1, D3, D5	High
G8	EOV compensation methods benchmarked for OWT scour monitoring. Multiple methods exist (regression, PCA, GP, Bayesian) but no systematic comparison on the same OWT dataset.	D2, D4	Medium
G9	Autoencoder/encoder-based representation learning for OWT structural state. Autoencoders are mature in other domains but zero papers apply them to geotechnical or OWT SHM sensor data.	D4 (explicit), D2	Medium
G10	Unified software framework for coupled geotechnical-structural-aeroelastic scour analysis. Each tool chain (FEM, p-y, aeroelastic) operates in isolation; no open-source pipeline exists.	D1, D4, D5	Medium

---

4. THE PhD's COVERAGE MAP

Matrix: G1--G10 gaps vs. 11 papers. Filled cells indicate direct contribution.

Gap	J1	J2	J3	J5	J11	V1	V2	E	A	B	Op3
G1: Scour + cyclic coupling			X
G2: Scour on suction buckets	X	X	X	X	X
G3: Field scour monitoring for OWT						X	X
G4: Scour-SHM-DT closed loop									X
G5: Probabilistic scour capacity				X					X
G6: Scour in silty/mixed soils	X		X
G7: Prototype-scale validation						X
G8: EOV compensation benchmark							X			X
G9: Encoder for OWT state								X	X
G10: Unified software framework											X

Gaps with strong coverage

• G2 (scour on suction buckets): addressed by 5 papers -- the dissertation's centre of gravity.
• G3 (field monitoring): V1 + V2 provide the first published operational OWT scour monitoring dataset.

Gaps with single-paper coverage

• G1 (scour + cyclic coupling): only J3 (centrifuge with backfill). The coupled millions-of-cycles regime remains open.
• G4 (closed-loop DT): only Paper A, which fuses encoder + field data for decisions but does not implement a real-time operational loop.
• G7 (prototype validation): V1's field data partially addresses this but is monitoring-focused, not a design-method validation campaign.
• G10 (unified framework): only Op3.

Gaps not fully closed

• G1 remains partially open: J3 addresses backfill after scour but not progressive scour under millions of cycles.
• G6 is addressed in silty sand (J1, J3 at KAIST) but not in clay or mixed transitional soils.

Papers vs. gaps coverage check

Every paper addresses at least one verified gap. J11 (Vesic generalisation) addresses only G2; its contribution is foundational theory rather than gap-filling, which is appropriate for a mechanics paper but should be framed accordingly.

---

5. OFFENSIVE GAP FRAMINGS

One sentence per paper, framing the gap as an indictment of the field.

J1: The field spent three decades perfecting scour prediction for monopiles while leaving suction-bucket tripods -- a foundation type already deployed at Borkum Riffgrund 1 -- without a single centrifuge dataset linking scour depth to natural frequency.

J2: Every published scour-stiffness study treats foundations as isolated columns, yet tripod suction buckets redistribute load through three footings whose interaction makes monopile-derived Winkler models fundamentally wrong.

J3: The field's scour experiments universally prepare pristine sand beds and excavate once, ignoring the reality that every storm backfills the scour hole with loose sediment whose stiffness nobody has measured.

J5: Probabilistic methods have been standard in structural reliability for 40 years, yet the entire literature on tripod suction-bucket scour consists of deterministic point estimates that conceal a 30% capacity spread.

J11: Vesic's cavity expansion theory has been applied unchanged since 1972, despite the obvious geometric violation that a tripod bucket's three-footed stress field bears no resemblance to a single expanding sphere.

V1: The community has published dozens of numerical demonstrations that natural frequency tracks scour, yet not one study recorded both simultaneously on an operational offshore wind turbine until this 32-month campaign.

V2: Four competing EOV compensation methods are advocated in parallel literatures with no shared benchmark, leaving practitioners to choose by reputation rather than evidence.

E: Every SHM paper treats each turbine as a unique snowflake requiring site-specific model calibration, when a physics-informed encoder trained on CPT-conditioned features can generalise across soil types without retraining.

A: The field builds increasingly sophisticated monitoring systems and capacity models in parallel but never closes the loop to answer the only question that matters: should I inspect this foundation today or next year?

B: Hundreds of papers propose vibration features for damage detection, yet no study has systematically ranked which features actually detect scour versus which merely respond to wind and tide.

Op3: Geotechnical engineers run OptumGX, structural engineers run OpenSees, and wind engineers run OpenFAST -- on three separate computers, with manual CSV hand-offs -- because no open-source pipeline connects them.

---

6. ULTIMATE RESEARCH QUESTIONS

Each question is one that the literature, verified against the master knowledge map, cannot currently answer.

Paper	Question
J1	How does scour depth map to natural frequency reduction for a tripod suction-bucket foundation in sand, and at what depth does the 1P resonance boundary become violated?
J2	Can a 3D-FE-calibrated 1D Winkler model reproduce the scour-frequency relationship for tripod buckets with sufficient accuracy to replace full 3D analysis in design?
J3	Does natural backfill after a scour event recover the foundation's dynamic stiffness, or does the loose backfill material create a false sense of safety?
J5	What is the full probability distribution of horizontal capacity and natural frequency for a tripod suction-bucket foundation across the plausible range of soil conditions and scour depths?
J11	How must Vesic's rigidity index be modified to account for the dissipation-weighted, three-dimensional stress state around a tripod suction-bucket foundation?
V1	Can vibration-based monitoring detect scour on an operational offshore wind turbine tripod, and what is the minimum detectable scour depth under real environmental noise?
V2	Which environmental compensation method -- state-function decomposition, cointegration, PCA, or Gaussian process regression -- yields the lowest residual variance on a common OWT scour-monitoring dataset?
E	Can a single physics-informed encoder represent the structural state of suction-bucket foundations across multiple soil types without site-specific retraining?
A	What is the cost-optimal inspection interval for a scoured tripod foundation when capacity uncertainty from an encoder is fused with field-monitored frequency data in a Bayesian decision framework?
B	Among the combinatorial space of vibration-derived scour features, which subset is simultaneously sensitive to scour and insensitive to environmental variability across different soil conditions?
Op3	Can an open-source pipeline coupling 3D limit analysis, nonlinear structural FEM, and aeroelastic simulation produce consistent scour assessments without manual data translation?

---

CROSS-CUTTING SYNTHESIS

The 11 papers form a vertically integrated stack. The bottom layer is mechanics (J11 theory, J5 probabilistic capacity, J1/J3 centrifuge evidence). The middle layer is modelling (J2 calibrated Winkler, Op3 framework, E encoder). The top layer is monitoring and decision (V1/V2 field SHM, B feature selection, A Bayesian decision). This architecture means each paper's output feeds the next, but it also means a weakness at any layer propagates upward.

The single largest remaining vulnerability is G1: coupled scour + long-term cyclic loading. J3 addresses backfill but not the millions-of-cycles regime. If cyclic loading fundamentally alters scour morphology or post-scour stiffness recovery -- as Al-Hammadi (2019) suggests for monopiles -- then the static-scour assumption underlying J1, J2, J5, and V1 could underestimate real degradation. This is the most important item for future work.

The second vulnerability is generalisability beyond sand. J1 and J3 test in KAIST silty sand; J5 models soft clay. But no paper bridges the gap between these two soil types, and cohesive-soil scour mechanics remain fundamentally different. Paper E's cross-soil encoder is the closest attempt, but its training data is synthetic.

What the PhD contributes that did not exist before: - The first centrifuge dataset linking scour to frequency for tripod suction buckets (J1). - The first field dataset linking scour to frequency for an operational OWT tripod (V1). - The first probabilistic capacity ensemble for scoured tripod foundations (J5). - The first systematic benchmark of EOV compensation methods on OWT data (V2). - The first physics-informed encoder for cross-soil structural-state representation (E). - The first closed-loop Bayesian decision framework connecting scour monitoring to maintenance scheduling (A).

These six "firsts" map onto gaps G2, G3, G5, G8, G9, and G4 respectively, confirming that the dissertation addresses the field's most consequential voids.

---

~3,200 words. Synthesised from Domain Maps 1--5 (45 batch summaries each, ~1,952 unique papers total).