Scan-to-CAD and Reverse Engineering Services

If you want the plain-language background first, start with our explainer on what reverse engineering is, then come back here for the service scope, deliverables, and process.

3D scanning for reverse engineering

The capture step is where reverse engineering succeeds or fails, because every downstream model is only as good as the point cloud behind it. For parts and objects we scan with a Creaform MetraSCAN 3D, a handheld optical metrology scanner built for parts, tight tolerances, and QC. It produces a high-density mesh or point cloud of the object at metrology-grade accuracy, capturing fine features, edges, and complex surfaces from any angle as we move around the part. For large installed equipment and building-scale conditions that cannot move to a lab, we use a tripod-based terrestrial scanner (Trimble X7) instead, giving dense, dimensionally consistent coverage from multiple positions that we register into one aligned dataset.

Handheld metrology scanning suits parts and objects from small machined components up to larger assemblies such as a weldment, a pump body, a casting, a bracket, or a machine frame, capturing tight tolerances and fine detail at close range. Large fixed conditions you cannot move to a lab are instead handled with terrestrial scanning. Transparent surfaces, highly reflective finishes, and deep internal cavities are harder for any optical method, and we will tell you up front when a part is a poor fit for optical scanning rather than promise a result the data cannot support. Capture, alignment, and the handoff into modeling are covered step by step in our guide on how to scan an object into CAD.

After scanning we register the individual setups into a single coordinate system and clean the data: removing noise, stray points, and anything outside the object of interest. The registered point cloud is the deliverable everything else is built from, and it can be exported on its own in E57, RCP, or RCS for clients who want to model in their own environment.

Scan-to-CAD: from point cloud to usable geometry

Scan-to-CAD is the reconstruction step that converts the registered point cloud into the geometry your team actually works in. A point cloud is millions of measured points; it is not a surface, a solid, or a drawing. Turning it into editable CAD is interpretation work, and it is delivered through our modeling partners who specialize in the target software and discipline.

How far the reconstruction goes depends on what you need the file for. Three broad routes exist, and many projects combine them:

The split between as-is and design intent is the single most important scoping decision, and it is explained in depth in our comparison of scan-to-CAD reverse engineering versus product design. Choosing the wrong one means paying for detail you cannot use, or receiving a clean model that no longer matches the part on your floor.

Output options: parametric CAD, mesh, and 2D drawings

There is no single correct deliverable. The right output is the one that matches how the file will be used, and reverse engineering routinely ends in more than one format.

STEP and IGES are the neutral exchange formats most CAD systems read, which makes them the safe choice when you do not know which package the receiving engineer uses. Native files preserve the editable feature tree but tie you to a specific application. For freeform shapes, an STL or OBJ mesh is often more honest than forcing the geometry into parametric features it was never designed around.

As-is documentation versus design-intent reconstruction

These two intents sound similar and are frequently confused, so it is worth being precise. As-is documentation reproduces the object as measured. If a frame is twisted, a flange is out of plane, or a casting has drifted from its nominal shape, the model shows it. This is what you want for clash detection, replacement parts that must mate with existing hardware, and any record of the actual condition.

Design-intent reconstruction looks past the measured surface to the geometry the object was meant to have. A hole the scan reads at 19.92 millimeters becomes a clean 20 millimeter hole; a slightly bowed face becomes a true plane. This produces a tidy, editable CAD model suited to modification and re-manufacture, but it deliberately departs from the exact as-scanned dimensions. Knowing which one you need before modeling starts is the difference between a useful file and an expensive surprise.

Tolerance and accuracy expectations

Accuracy in reverse engineering is the sum of several stages, not a single scanner number. The scanner has its own ranging accuracy, registration adds a small alignment uncertainty across setups, and the modeling step introduces a fitting tolerance when surfaces are reconstructed into CAD. The realistic accuracy of the final model is governed by the weakest link in that chain, not by the scanner alone.

Handheld metrology scanning is built for tight tolerances and is well matched to the dimensional reverse engineering of parts, mechanical components, and assemblies, capturing fine features at metrology-grade accuracy. Even so, the final model still reflects registration and modeling tolerances, and we are explicit that we do not certify a part the way a coordinate measuring machine in a controlled inspection routine would. We scope the tolerance target before capture so the scan density, fixturing, and modeling approach match the requirement, and we are explicit about what the data can and cannot guarantee. We do not perform licensed land surveying, and we do not certify manufacturing tolerances or assume mechanical design responsibility unless that is separately and formally scoped with a qualified partner.

The field-to-CAD process

A reverse engineering project moves through a predictable sequence:

1. Scope. Define the object, the target output and format, the as-is versus design-intent intent, the tolerance expectation, and the software the file lands in.
2. Field capture. Scan the object with the handheld MetraSCAN 3D, moving around the part to cover all required surfaces and features, with reference targets where they help alignment; large installed conditions are captured with terrestrial scanning instead.
3. Registration and cleanup. Align the setups into one coordinate system, remove noise, and isolate the object. This is done in house.
4. CAD reconstruction. Our modeling partners build the parametric solid, surface model, mesh, or drawings to the agreed scope.
5. QA and handoff. Check the model against the point cloud, document where interpretation was required, and deliver native and neutral files with notes.

The honest division of labor is worth stating plainly: field scanning and registration are our core, in-house strength, and CAD reconstruction is delivered through a vetted partner network. That keeps the measurement reliable and the modeling in expert hands.

What affects cost

Reverse engineering does not have a single dollar-per-part or dollar-per-square-foot rate, and there is no neutral industry benchmark for price. Standards such as the USIBD Level of Accuracy framework define how accurate a deliverable is, not what it should cost. Pricing comes from two separable parts of the work.

Field scanning is billed by the day or the half day, because what we are really selling is time on site and the data that comes from it. A typical scan day runs in the range of 3,200 to 5,000 dollars and can cover a great deal of capture; complex access, many small setups, or restricted environments slow that down. CAD reconstruction is priced by complexity, not by area: a simple as-is shell is inexpensive, while a fully parametric, tolerance-driven model of an intricate assembly is the expensive end of the range. The drivers are object complexity, the chosen output, the accuracy target, the number of features to reconstruct, and how clean the surfaces are.

For a structured walk-through of exactly what to send so a quote reflects the real scope, see our reverse engineering quote checklist, which lists the target output, photos, key dimensions, tolerance expectations, and access details that make pricing accurate instead of padded.