Reconstructing Balclutha

​

The research question

Can a historical ship record become a vessel that people can inspect, take apart, reassemble, challenge, and correct without turning gaps in the record into false certainty? That question led to several smaller ones:

• Can we keep the source drawing, our interpretation, the generated geometry, and the review decision separate?
• Can a person trace any visible part back to its evidence and see what remains uncertain?
• Can the vessel be decomposed and recomposed without losing identities or relationships?
• Can the system stop safely when a file, unit, source, or claim is unsupported?
• Can another researcher reproduce the bounded result from a clean starting point?

​

Why Balclutha

Balclutha is a three-masted, steel-hulled sailing cargo ship built in 1886. She is historically important, physically complex, and unusually well documented. The Historic American Engineering Record survey held by the Library of Congress includes 71 measured drawings and 240 data pages, alongside photographs and caption records. That combination made Balclutha a strong research case. It gave us evidence across the hull, decks, internal structure, openings, outfitting, and rig. It also contained the kinds of contradictions, later modifications, missing details, and temporal gaps that real engineering research must handle.

​

The sources we admitted

We worked from the public HAER CA-54 record and its associated survey material. Our closed source inventory covers all 71 measured drawing sheets and all 240 pages in the survey volume. We do not silently merge records from different moments in the vessel’s life. The best-documented 1988 HAER survey configuration is the primary visual assembly. The original construction record, later modifications, the separate 1998 port-side total-station survey, and the 2020 LiDAR-derived hull expansion remain labeled as distinct evidence states. This matters because two records can both be useful without describing exactly the same side, surface, date, or physical reference.

​

What we reconstructed

The result is a complete visual research assembly with four inspectable layers:

1. Hull envelope hypothesis: the vessel’s overall visual hull.
2. Internal structural hypothesis: the visible structural system inside that envelope.
3. Above-deck and outfitting hypothesis: deckhouses, openings, equipment, and related features.
4. Rigging silhouette hypothesis: the masts, yards, and rig families needed to read the vessel as a complete sailing ship.

The governed vessel hierarchy contains 65 component identities and 49 named evidence gaps: 114 selectable identities connected by 136 declared relationships. The visual assembly contains 307 imported solids. Forty-four identities own independently selectable geometry; the remaining 70 stay selectable as evidence records, questions, or unresolved envelopes rather than receiving invented part geometry. At Frames 60-62, a smaller bounded reconstruction remains nested inside the whole-vessel assembly. It contains 32 components and 18 relationships and gives us a more tightly controlled place to test source interpretation, geometry intake, decomposition, and recomposition.

​

What the Workbench proves

The Workbench proves that a research vessel can be more than a picture. A reader can:

• browse the complete vessel hierarchy;
• select reconstructed geometry or a named unresolved identity;
• isolate, hide, and restore parts and families;
• inspect cutaway and exploded presentations;
• move from vessel to system, assembly, and part;
• review evidence state, assumptions, measurements, and correction questions;
• recompose the governed inventory without losing its 114 identities or 136 relationships.

An unresolved selection does not disappear. It shows the evidence envelope and the question that must be answered before stronger geometry can be claimed.

​

What remains hypothetical or unresolved

The vessel is visually complete because unresolved regions are still represented and selectable. That does not mean every region is equally certain. The remaining limits include:

• exact as-built geometry for the full vessel;
• present-condition geometry;
• complete plate, connection, scantling, and material schedules;
• cross-projection equivalence where drawings do not independently establish it;
• hydrostatics, structural adequacy, class compliance, and production readiness;
• several local boundaries, including the next Frame 62-64 evidence gate;
• historical changes whose original, surveyed, modified, and later states cannot yet be reconciled.

Where evidence is missing, we keep a named envelope, proxy, or question. We do not fill the space with silent certainty.

​

How we tested repeatability and failure

We tested more than whether the vessel looked plausible on screen. For the bounded Frames 60-62 case, we restored declared sources and controls into a clean environment, rebuilt the research chain, and compared independent primary and replay results. We also tested both a supported STEP path and a deliberately malformed STEP path. The supported path remained bounded; the malformed file stopped before it could emit geometry, measurements, authority, or downstream work. For the complete visual assembly, independent review checked inventory closure, selections, omissions, rights boundaries, and authority boundaries. Mutation tests deliberately changed or removed inputs and required the process to fail closed instead of quietly accepting the change. This does not prove that the whole ship is physically exact. It proves that the research record is inspectable, its bounded replay is reproducible, and known failure paths do not silently become accepted engineering output.

​

The authority boundary

This work can support research, review, discussion, and correction. It does not provide:

• naval-architect approval;
• structural-engineering approval;
• class or regulatory approval;
• measurement authority for unverified geometry;
• a production model, fabrication package, or work release;
• permission to treat every historical source as commercially redistributable.

The AI-assisted process can organize evidence, propose interpretations, generate bounded geometry, and expose contradictions. Accountable people still decide whether an interpretation is correct and whether it is fit for a real engineering purpose.

​

How architects and researchers can challenge it

The preferred correction is specific and reviewable:

1. Select the identity in the Workbench.
2. Read its source, temporal state, assumption, and correction question.
3. Identify the drawing, report, survey, photograph, or direct observation that changes the record.
4. State which boundary, dimension, relationship, or time-state is affected.
5. Keep the correction as a review proposal until an accountable person accepts or supersedes it.

That means a disagreement becomes useful evidence rather than an untracked visual edit.

​

What comes next

Our next geometric gate is the Frame 62-64 block. It remains blocked until the missing restoration and accountable boundary evidence are acquired and verified. We will not use the complete visual vessel as permission to skip that gate. Beyond that block, the larger goal is a vessel whose systems, assemblies, and parts can be inspected, decomposed, corrected, and recomposed while keeping the source and uncertainty attached. The complete visual ship is the research container for that work. It is not the end of verification.