Skip to content

Updates

Daily log of research and writing activity on the PhD portfolio. Newest first.

2026-04-18 · Fri · site day

The site you are reading went live. Everything from the dissertation programme that previously lived in scattered .qmd files, _review/ renders, .bib files, and shell scripts got consolidated into one browseable structure.

Content populated from the dissertation work of prior days:

  • Papers tab — ten full first-author manuscripts, each on its own page with figures, tables, and references. Tables and display-math equations now render correctly after fixing two rendering bugs: md_in_html required markdown="1" on <div id="tbl-*"> wrappers, and <span id="eq-*">\[…\]</span> wrappers were blocking arithmatex from processing display math. Resolved in f742533.
  • Reviews tab — point-by-point trackers for J2 R2 (Ocean Engineering), J3 R1 (Ocean Engineering), and the MSSP rejection → V resubmission consolidation.
  • Literature Review tab — the 1,952-paper hierarchical synthesis, now with mermaid mindmaps, 5 domain maps, 45 batch summaries, and the master knowledge map as browseable pages.
  • Workflow tab — methodology and per-paper prototypes for claim-driven writing: claim-driven base, J2 / J3 / J5 / V methodology prototypes, J2 / J3 results prototypes, and the V&V commission 6 equivalence-proof derivation.
  • Status tab — publication pipeline with vertical timeline, 20-week cadence, effort estimates, triage ladder, and a dedicated full-page Gantt chart view.
  • Presentation tab — dissertation defence deck (reveal.js), renovated for the new V1/V2-consolidated portfolio.
  • About tab — background and contact.

Site infrastructure:

  • Logo + favicon, site meta for social cards (generated on CI).
  • Scroll-progress bar (top), save-as-PDF button (top-right), navigation.instant for SPA-style page swaps.
  • Dark-mode CSS overrides so every mermaid diagram remains readable in slate palette.
  • Custom 404, tags system, git-revision-date last-updated footer.
  • Home and About pages redesigned for simplicity — no grid-card clutter.
  • Every paper portfolio mermaid diagram made TB (vertical) with ≥15 px fonts for readability.
  • :material-*: icon shortcodes now render as SVG (the pymdownx.emoji extension was missing its Material twemoji config).

2026-04-17 · Thu · synthesis day

Focus on compressing the literature corpus and sharpening the dissertation positioning.

  • Dissertation gap analysis — per-paper defensive / offensive / constructive gap framing, portfolio arc, and the rule that every paper needs both a defensive and an offensive gap. Full note: 2026-04-17 — Dissertation gap analysis: ten papers, one spine.
  • Literature synthesis — hierarchical divide-and-conquer over 1,952 papers:
    • Round 1 · 9 batches × 5 agents read ~40 papers each → 45 batch summaries
    • Round 2 · 5 domain agents → geotechnical, SHM, centrifuge, ML/decision, offshore wind scour
    • Round 3 · 1 master knowledge map with 10 consensus claims, 5 debates, 10 verified gaps, 11 × 10 coverage matrix, and one-sentence offensive framing per paper
    • Round 4 · per-paper enhanced gap claims (completed for J1–J3, V)
  • Paper E archived — merged into Paper A (2026-04-17). Central claim was circular: capacity was never independently measured. Paper A now self-contained with encoder architecture + validation + fusion + decision.

2026-04-16 · Wed · portfolio reframing

Hard decisions on scope and strategy.

  • Paper B reframed — from "success story" to honest audit titled "What Works and What Doesn't for Tripod Detection." Three definitive answers over vague coverage; reviewer-defensibility up.
  • Paper B field result — the +55.8 % measured coherence shift on 32-month Gunsan record. Transforms the paper from projected claim into measured field detection.
  • Paper V2 refocused — dropped the Johansen-cointegration equivalence proof (inapplicable on parked-state data, features are stationary, not I(1)). Reframed to empirical four-method benchmark (state-function / cointegration / PCA / GPR).
  • V1 + V2 → V consolidation (decided 2026-04-18 after re-examining the MSSP rejection) — both reviewer objections addressable in a single stronger paper targeting Engineering Structures.
  • Paper count: 11 → 10 active in the defense window.

Before 2026-04-16 · background

The work catalogued above rests on five years of prior research activity:

  • Centrifuge programme at KAIST — 22 tests at 70 g across 5 soil series on 1:70-scale tripod suction bucket models, 53-channel instrumentation.
  • Field monitoring at Gunsan — 32 months of continuous vibration data on an operational 4.2 MW offshore wind turbine, 22,616 windows processed.
  • Numerical pipeline (Op³) — 3D OptumGX limit analysis (177 parametric runs) coupled to 1D OpenSeesPy dynamics, reproducing centrifuge frequencies within 1.19 % mean error. Published on PyPI with 140 tests, 39 benchmarks, and a CI/CD pipeline.
  • Monte Carlo ensemble — 1,794 real simulations on the Op³ framework, no synthetic augmentation. CPT-conditioned encoder achieves Pearson r = 0.989 on capacity prediction, 0.76 on unseen soil types.

For the underlying scientific contributions, see the Papers tab. For the publication pipeline, see Status.

This log is maintained by hand. For the authoritative change history, see GitHub commits.