Parts Book Data Model: From PDF Pages to Structured Catalog Data

Every parts book starts as a useful document and ends as a liability. The diagrams were drawn when the product launched. The parts lists were accurate that year. Five years later, a third of the part numbers have been superseded, the manufacturer has made two mid-serial engineering changes, and the parts book still shows the original numbers. The technician trusts the diagram, orders the old part, and receives a “discontinued — no stock” response.

The solution is not to redraw the diagrams every time a part number changes. It is to separate the layers so the parts book can stay visually stable while the data behind it stays current.

The four layers

1. Visual layer — the exploded-view diagram

The diagram shows each component in its spatial relationship to the rest of the assembly. It is expensive to produce (drawn from CAD models or by technical illustrators) and almost never updated after initial publication. The visual layer is static, and that is acceptable — the physical arrangement of parts rarely changes even when part numbers do.

2. Positional layer — callout numbers

Each component in the diagram has a callout or position number that links the visual location to a row in the parts list. Position 14 in the diagram maps to row 14 in the table. This layer is also stable: the positions do not change unless the assembly itself is redesigned.

3. Identity layer — part numbers and supersession

This is where the data goes stale. The parts list maps each position to a part number, a description, and a quantity. When the manufacturer supersedes a part number, the mapping changes — but the position in the diagram does not. If the identity layer is stored as static text (a cell in a PDF table), it cannot follow the change. If it is stored as a reference to a live product identity that carries supersession chains, the parts book updates automatically.

4. Applicability layer — which machines this assembly covers

The same assembly may appear in multiple equipment models, but with different parts at certain positions depending on the model variant or serial range. The applicability data — make, model, serial-range start, serial-range end — determines which version of the parts list applies to a specific machine.

What a PDF collapses

A PDF parts-book page contains all four layers flattened into one:

The structured data model

Separating the layers into a data model means each layer can update independently:

The critical relationship is position → part identity, not position → static part number. When a part is superseded, the position’s reference resolves through the chain to the current part. The diagram stays the same. The position stays the same. The part number the customer sees is always current.

Handling mid-serial overrides

For assemblies that serve multiple equipment variants, a position may point to different parts depending on the serial range:

This is an applicability-scoped override on the position. The diagram still shows position 14 in the same location. The part at position 14 depends on the customer’s serial number. Without this structure, the catalog cannot answer “which pump fits my machine?”

From PDF to data

The conversion from a PDF parts book to structured data involves:

1. Extract the visual layer

   Isolate the exploded-view image from each page. Store it as a referenced asset, not an embedded bitmap.

2. Extract the positional layer

   Map each callout number in the diagram to its coordinates (for interactive overlays) and to the corresponding row in the parts list.

3. Extract the identity layer

   Parse the parts-list table: position number, part number, description, quantity. Resolve each part number against the current product identity to check for supersessions.

4. Extract the applicability layer

   Parse the header or cover page for equipment model, serial range, and production year. Structure it as a typed fitment record.

5. Link the layers

   Connect positions to part identities (not static strings) and scope them by applicability. The result is a parts book that can update its part numbers without touching the diagram.

For the detailed workflow, see the PDF parts-book conversion playbook.

Where Claro fits

Claro provides the identity layer that makes the parts book live. Each position in the structured parts book points to a product identity in the Product Identity & Evidence Graph. When a part is superseded, the position resolves to the new part automatically. When a customer enters their serial number, the applicability layer selects the correct version of the parts list. The diagram stays static; the data behind it stays current.

Related

FAQ

What are the layers of a parts book? +

A parts book has four layers: the visual layer (exploded-view diagrams), the positional layer (callout numbers that link diagram locations to parts-list rows), the identity layer (part numbers, descriptions, supersession chains) and the applicability layer (which machines and serial ranges each assembly covers). A PDF collapses all four into a flat image plus a table.

Why does a parts book go stale? +

Because the visual layer (the exploded-view diagram) is drawn once and rarely redrawn, while the identity layer (part numbers) changes whenever a manufacturer supersedes a part. The diagram still shows position 14, but the part number at position 14 may have changed twice since the diagram was published.

How do you keep a parts book current without redrawing every diagram? +

By separating the layers. The diagram stays as it is. The position number stays as it is. The part number at each position is a reference to a live product identity that follows supersession chains. When a part is superseded, the parts book updates automatically because the position points to the identity, not to a static part number.