Reverse Engineering: Reconstructing Existing Parts

A part breaks, the supplier no longer exists, and the last drawing dates back to the 90s — or never existed at all. Or: you are holding an injection-molded part, want to improve it, but nobody knows the design data. Reverse engineering solves exactly this problem: turning the physical part back into a production-ready CAD model.

When Reverse Engineering Makes Sense

• Spare part supply – The OEM no longer delivers, but the machine has to keep running
• Legacy products – Old products need to be modernized or remanufactured without existing data
• Further development – An existing part should be improved (lighter, stiffer, cheaper)
• Process change – A machined part should be converted to injection molding (or vice versa)
• Documentation gaps – The design only existed with an external contractor who never delivered the data

The Workflow in 4 Steps

1. Capture – The part is measured: via 3D scanner, CT scanner or tactile probing. Result: a point cloud or an STL mesh.
2. Data preparation – The mesh is cleaned: closing holes, filtering noise, aligning to meaningful reference planes.
3. Reconstruction – A parametric CAD model is built from the mesh: planar faces, radii, holes and sketches — not a blunt copy of the scan surface.
4. Validation – Manufacture a prototype, check fits, measure against the original. Only then is the model production-ready.

Measurement Methods Compared

Optical 3D Scanner (Structured Light / Laser)

The standard for most cases. Accuracy: ±0.02–0.05 mm. Captures external geometry quickly and completely. Limits: deep holes, undercuts and internal features are hard or impossible to reach. Cost: EUR 300–800 per part from a service provider.

CT Scanner (Computed Tomography)

X-rays the part and also delivers internal geometry: channels, cavities, wall thicknesses. Accuracy: ±0.01–0.05 mm. Cost: EUR 500–1,500 per scan. Worth it for complex molded parts with cooling channels, foam cores or hidden features.

Tactile Measurement (CMM, Measuring Arm)

Coordinate measuring machines probe individual points. Accuracy: ±0.001–0.005 mm. The right approach for critical fit dimensions, bearing seats and hole patterns. Downside: point by point, not full-surface. In practice, the combination wins: scanner for the shape, CMM for the fits.

STL Is Not CAD: The Most Important Point

The most common misconception in reverse engineering: a scan produces a triangle mesh (STL) — thousands of small triangles approximating the surface. This file is nearly worthless for manufacturing. No mold maker and no CNC shop can work with it.

Only a parametric model is production-ready: planar faces where the original had planar faces, exact radii, round holes at defined positions, sketches with dimensions. That means real engineering work — the mesh serves as a reference, not as the result.

Tolerances: The Underestimated Problem

The measured part is not the nominal part. It contains all manufacturing deviations, wear and — for plastic parts — shrinkage and warpage. Anyone who adopts the scanned dimension 1:1 builds the original's defects into the new design.

In practice this means:

• A measured radius of 4.7 mm was probably 5.0 mm nominally — round values to plausible design dimensions
• Wall thicknesses on molded parts vary due to warpage — average across several locations
• Check hole patterns against standard dimensions and against function (Does the screw fit? Does the mating part fit?)
• Whenever possible: measure multiple samples and use the average

Legal Considerations

Reverse engineering your own products or spare parts for your own machines is usually unproblematic. It gets more critical with third-party products: registered designs, patents and copyright can prohibit remanufacturing someone else's parts. Technical capability does not protect against injunctions.

Rule of thumb: For internal maintenance and your own legacy products, generally safe. For selling reproductions of third-party parts: get legal advice first.

Cost and Timeline

Realistic ranges for reconstruction work:

• Simple part (machined, few features): EUR 500–2,000, 1–2 weeks
• Injection-molded enclosure with ribs, clips and connectors: EUR 2,000–6,000, 3–6 weeks
• Assemblies (5–15 parts with fits between them): from EUR 5,000, depending on complexity

The decisive cost driver is not the scan but the parametric reconstruction — the more freeform surfaces and fits, the more effort.

Common Mistakes

1. Scanning only– Handing the STL off as a "CAD model". The manufacturer cannot use it, and the project stalls.
2. Copying scan errors – Warpage, shrinkage and wear of the original get copied along. Without engineering rework, the part remains a replica with the same weaknesses.
3. Ignoring material – The original was POM, the reproduction becomes ABS — and neither shrinkage nor strength match anymore. Identify the material (e.g. IR spectroscopy at a lab, approx. EUR 100–300).
4. Skipping validation – Without a prototype and fit check, the first production run regularly fails.

Case Study: Gearbox Cover from the 80s

A machine builder needed spare gearbox covers for equipment whose manufacturer had gone out of business 20 years ago. Approach: 3D-scan of the original for the outer shape, CMM measurement of the four bearing seats and the sealing face, parametric reconstruction in CAD. Measured hole spacing was rationalized back to the metric grid, wear marks on the sealing face corrected. Result: CNC-machined replacement cover, fit on first try. Total cost: approx. EUR 3,500 including prototype — versus a six-figure new investment in the equipment.

Anton Steenken

B.Eng. · Hardware R&D Engineer · Founder of engineer your idea