Converting legacy 2D drawings to 3D models: scope, cost drivers and pitfalls

Most manufacturers that have been around for a few decades have the same archive: paper prints in a cabinet, scanned PDFs, AutoCAD files from three CAD managers ago. The parts are still made. Then a customer asks for a STEP file, a supplier needs a model to program, or the company moves to 3D and wants the archive to follow. This article is about scoping that work honestly.

What makes a conversion slow?

The size of the part matters less than you would expect. A 2-metre weldment drawn cleanly is faster to model than a small cast housing with three partial views and a note that says “see old revision”.

Cost driver	Why it costs time
Missing or partial views	The modeler has to infer geometry, then confirm it
Conflicting dimensions	Two views say different things; someone must decide
Dimensions not to scale (“NTS”)	The drawing cannot be measured to fill gaps
Many variants on one drawing (tabulated)	Each variant is a configuration to build and check
Unclear revision history	Which version is actually being made?
Complex surfaces (castings, forgings)	Draft, fillets and parting lines must be reconstructed
Poor scans	Reading the drawing itself takes time

When you send a sample, a good supplier will tell you which of these apply. That list is more useful than a price per drawing.

Scoping in three steps

1. Inventory. Count the drawings and tag each with what it is: part or assembly, still active or obsolete, format (paper, PDF, DWG), and how many variants it covers. Converting obsolete parts is usually wasted money.

2. Sample. Pick five to ten drawings that represent the range: an easy one, a typical one, the worst one. Have them converted first. This gives a real time per complexity group and shows how many questions each drawing generates.

3. Group and price. Sort the remaining drawings into complexity groups using what the sample taught. Price by group, with a rule for drawings that turn out harder than expected.

Conflicts: the part nobody budgets for

Turning a drawing into a solid model is the most thorough check that drawing has ever had. A 3D model cannot hold two values for the same edge, so every inconsistency surfaces: a hole located from two references that do not agree, a section that does not match the plan view, a chamfer shown in one view only.

Agree before you start how these are handled:

• the modeler never resolves a conflict silently;
• each conflict is logged with the drawing, the views involved and a proposed resolution;
• your engineer decides, ideally by checking a real part when possible;
• the decision is recorded on the new drawing’s revision note.

Which model do you want at the end?

“A 3D model” can mean very different things, and the difference shows up in the price:

Model type	Good for	Not good for
Dumb solid (STEP-like)	Quoting, CAM, visualization	Making changes later
Parametric model, simple feature tree	Most production parts; easy edits	Large variant families
Parametric with configurations or design table	Families of similar parts	Teams that do not use configurations

For parts that are still evolving, ask for a parametric model with a feature tree a colleague can follow: sketches fully defined, features named, no hidden dependencies on external references.

Keep the old drawing linked

The original drawing is evidence. Keep it attached to the new model (in your PDM, or at least in the same folder with the same part number), and note on the new drawing which legacy drawing and revision it replaces. When a question comes up in five years, someone will want to see what the old print said.

Paper, PDF or DWG: does the source format matter?

Yes, but less than the content. A DWG with real geometry lets the modeler pick up profiles directly, which helps with complex outlines. A scanned PDF has to be read and retyped, which is slower and adds a risk of transcription errors. Paper prints add a scanning step and sometimes faded dimensions. In all three cases, though, the time-consuming part is interpreting the drawing, not reading it, so a clean scan of a well-drawn print can be faster than a messy DWG.

What to check on the first converted batch

When the sample comes back, review it as you would review any new drawing, and also check:

• Feature tree quality: can someone on your team change a dimension without the model breaking?
• Conflict log: were conflicts logged and resolved with you, or quietly decided?
• Drawing standard: does the new drawing follow your current standard, and are changes from the old drawing listed?
• Part numbers and revisions: does the new revision scheme make clear that this drawing supersedes the legacy one?

Fix any problem in the process now. It will be repeated on every drawing in the archive otherwise.

Choosing what not to convert

Not every drawing deserves a model. Good candidates for leaving as they are: obsolete parts, parts bought as standard catalogue items, and simple parts that are never changed and are made from the PDF without trouble. For those, a clean re-scan and a correct entry in the document index may be all you need.

A realistic plan for an archive

1. Start with the product family that is quoted or reordered most often.
2. Convert a sample, measure time and questions, adjust the groups.
3. Convert the family, release it, and use it as the template for the next one.
4. Leave obsolete parts alone unless someone asks for them.

Conversion done this way pays back on every RFQ, supplier question and engineering change after it. Done as a bulk “scan and model everything” job, it often produces models nobody trusts.