STEP to Parametric in NX: 4 Routes and One Acceptance Test

STEP to Parametric in NX: The Short Version

STEP to parametric means turning a history-free body from a STEP import or legacy migration into a native NX part with a feature history designers can edit. Four routes exist: Synchronous Modeling, NX’s recognition helpers, manual remodeling and an AI plus NXOpen rebuild. inteliScience has rebuilt flanges and screws via the AI route; automated comparison with the source decides success.

Route	Result	Fits
1. Synchronous Modeling	Face edits on the imported body, no history	Parts that rarely change
2. NX recognition helpers	Better edits and CAM features, no modeling history	Imported data you keep as a body
3. Manual remodeling	A full history, built by a designer	Variant masters, a few key parts
4. AI + NXOpen rebuild	History from a class template, comparison report per part	Recurring classes such as flanges and screws

The Problem: A Migrated Body Carries Geometry, Not Design Intent

A DACH machine builder’s CAD team asked us a precise question: can a new NX part with a sensible feature history be built from migrated 3D data that has none? Every change to such a part starts with a body that has faces and volume, but no sketch, no expression and no feature order.

The gap is structural. Exchange standards traditionally carry shape. NIST researchers proposed in 2007 to transfer construction history, parameters and design features through ISO 10303 Parts 55, 108 and 111, and tested it with a prototype translator (NIST). The STEP files in a migrated archive carry no history, so NX imports a body.

What to validate after a migration is covered in our article on validating CAD data after a migration to NX. A STEP to parametric project starts where that checklist ends: the data is clean, but it does not behave like native NX data when someone has to change it.

The team’s own idea was the most useful part: rebuild, compare the result automatically with the source, and measure the success rate. We agreed. Hence this post: four routes, one acceptance test.

Three Routes Without AI: Edit, Recognize, Remodel

Before any AI, NX gives you two ways to work with a history-free body, and a designer gives you a third. For a good share of an archive, one of them is the right answer.

Route 1: Keep the Body, Edit It With Synchronous Modeling

Synchronous Modeling edits faces without a history. Siemens’ Designcenter NX tips (28 April 2026) name Move Face, Resize Hole, Replace Face, Delete Face, Resize Pattern and Radiate Face, and call the approach particularly valuable on imported geometry or models without feature history (Siemens Designcenter blog). When NX misreads a chamfer or blend on foreign geometry, Label Chamfer and Label Notch Blend tell it what those faces are, so a moved face keeps a constant chamfer instead of extending the slope (TU Darmstadt NX tutorial).

For parts that change once or twice, this is the right route and cheaper than any rebuild. It gives you no design intent: no expression PCD to change, no sketch to reconstrain. That is also why our AI geometry manipulation jobs flag feature-less regions instead of guessing.

Route 2: What NX Recognizes, and What It Does Not

We found no Modeling command that converts an imported body into a feature history in the Siemens documentation and community material we checked in September 2026. If your release has one, test it first. NX does offer recognition helpers:

Resize Pattern loads a whole pattern from its faces, and the labelling commands mark chamfers and blends.
Select Similar Faces finds faces with similar radius, curvature or orientation and, according to Siemens, learns from the user’s selections (NX Design blog, 24 February 2026).
NX CAM’s Find Features with Parametric Recognition classifies machining features such as holes, also on imported files without a feature tree (Swoosh Technologies). The output is machining features for CAM, not modeling history.

Solid Edge, by contrast, ships Recognize Holes (ST5) and Recognize Hole Pattern (ST6) for imported geometry (Solid Edge blog). In NX, recognition feeds editing and machining: an input to route 4, not a replacement.

Route 3: Remodel by Hand Against the Imported Body

A designer rebuilds the part with the imported body on a reference layer, following your modeling rules. It gives the best history, because a person decides the design intent, and it does not scale to an archive. We recommend it for future variant masters and for whatever the automated route flags.

Route 4: Rebuild the Feature History With AI and NXOpen

Route 4 automates route 3 for recurring part classes. We have rebuilt simple parts such as flanges and screws with a feature history this way. For very complex geometry, success is not guaranteed, and we say so before a project starts.

Our AI application works in a running NX session through NX tools: it reads the work state, creates sketches and builds geometry, and each step stays an editable feature in the Part Navigator (see how AI agents read and build geometry in NX). The model matters less than the rules: a hosted model or an open-weight model on your own hardware. Models and data can run in the EU, in Germany or on-premise, depending on your data protection requirements.

What cannot be automated is the specification. Design rules and the desired history structure must be fixed per class before the first run; otherwise “sensible history” means whatever the pipeline happened to build.

The Acceptance Test: Rebuilt Part Against Source Geometry

A rebuild is accepted when the body matches the source within agreed tolerances and the feature tree matches the template. A perfect body with a messy history misses the point; a clean history with a missing chamfer is a wrong part.

Check	Catches	Interactive in NX	Batch (NXOpen)	Example criterion
Solid validity	Open or extra bodies, tiny faces	Examine Geometry	`Body.IsSolidBody`, body count	Exactly one solid
Volume and area	Missing or extra material	Measure	`MeasureBodies.Volume`, `Area`	Delta at most 0.1 percent
Centroid	Features on the wrong side	Measure	`MeasureBodies.Centroid`	Offset at most 0.05 mm
Bounding box	Wrong size or orientation	Measure	UF `AskBoundingBox`	Within modeling tolerance
Face and edge count	Missing chamfers, split faces	Compare Body	`Body.GetFaces()`, `GetEdges()`	Equal, or difference explained
Surface deviation	What scalar checks average out	Compare Body, Deviation Gauge	`DeviationGaugeBuilder`	Maximum at most 0.02 mm
History structure	Wrong order, unnamed expressions	Part Navigator	`Feature.FeatureType`, `GetExpressions()`	Matches the template

The criteria are starting values for a discussion, not benchmarks; derive yours from your modeling tolerance and the accuracy of the source data.

Compare Body colors faces that are missing, new or changed between a control and a sample body, and its Deviation option shows how far each face moved (NX Design blog, July 2023). Deviation Gauge measures target against reference objects with inner and outer tolerances, and its NXOpen builder exposes the same settings (NX Open reference). The single Measure dialog, since NX 12.0.1, returns body volume, area and mass (Swoosh Technologies), and Examine Geometry flags tiny objects (Applied CAx).

Why so many checks? Volume misses a hole on the wrong side; the centroid catches it. A symmetric flange whose bolt pattern is rotated by half a pitch fools both, and only surface deviation or Compare Body shows it. Our clear recommendation: never accept a rebuild on mass properties alone.

The scalar checks take a few lines. This read-only journal assumes the source body on layer 250 and the rebuilt body on layer 1 of the rebuilt part:

# NX journal (Python): scalar acceptance checks, rebuilt body vs. imported source body.
# Convention for this example: source body on layer 250, rebuilt body on layer 1.
# Read-only: writes to the listing window, changes nothing in the part.
import math
import NXOpen

SOURCE_LAYER, REBUILT_LAYER = 250, 1

session = NXOpen.Session.GetSession()
work_part = session.Parts.Work
lw = session.ListingWindow
lw.Open()

uc = work_part.UnitCollection
units = [uc.GetBase("Area"), uc.GetBase("Volume"), uc.GetBase("Mass"), uc.GetBase("Length")]

def solid_on_layer(layer):
    solids = [b for b in work_part.Bodies if b.IsSolidBody and b.Layer == layer]
    if len(solids) != 1:
        raise RuntimeError(f"layer {layer}: expected 1 solid, found {len(solids)}")
    return solids[0]

def metrics(body):
    mp = work_part.MeasureManager.NewMassProperties(units, 0.999, [body])
    c = mp.Centroid
    return {"volume": mp.Volume, "area": mp.Area, "centroid": (c.X, c.Y, c.Z),
            "faces": len(body.GetFaces()), "edges": len(body.GetEdges())}

src = metrics(solid_on_layer(SOURCE_LAYER))
new = metrics(solid_on_layer(REBUILT_LAYER))

shift = math.sqrt(sum((a - b) ** 2 for a, b in zip(src["centroid"], new["centroid"])))
lw.WriteLine(f"volume delta   {abs(new['volume'] - src['volume']) / src['volume']:.5f}")
lw.WriteLine(f"area delta     {abs(new['area'] - src['area']) / src['area']:.5f}")
lw.WriteLine(f"centroid shift {shift:.4f}")
lw.WriteLine(f"faces {src['faces']} -> {new['faces']}, edges {src['edges']} -> {new['edges']}")

# Feature history of the rebuilt part in creation order, for the template check
for feat in sorted(work_part.Features, key=lambda f: f.Timestamp):
    lw.WriteLine(f"{feat.Timestamp:3d}  {feat.FeatureType:<20} {feat.GetFeatureName()}")

Every member is in the NXOpen reference: MeasureManager (Siemens suggests 0.999 as accuracy for a more exact measurement), MeasureBodies, Body and Feature.

Report the success rate per class, never as one number for the batch:

Attempted, rebuilt, passed geometry, passed history, accepted after review.
Success rate: accepted divided by attempted.
Failure codes (classification, extraction, rebuild, geometry, history), because each points to a different fix.

This runs as a pilot. What’s missing for production: Templates for every part class, a review queue with a release gate, Teamcenter storage next to the source, GDPR-compliant hosting (EU, Germany or on-premise), monitoring, and a re-run after each NX upgrade. We close the gap from “works on twenty flanges” to “runs over the archive” with AI-assisted geometry creation on NXOpen: parametric output, confirm gate, logged runs.

STEP to Parametric by Part Class: Which Route When

The route depends on the part class and on how often the part changes. We start pilots with this table; your data will move some rows.

Part class	Recommended route	Why
Standard parts (screws, nuts, pins)	Library part where one exists, otherwise template rebuild	Library parts carry attributes; screws already rebuilt
Rotational parts (flanges, shafts, bushings)	AI + NXOpen rebuild with a class template	Clean profiles and hole patterns; flanges already rebuilt
Prismatic parts (plates, brackets, blocks)	AI + NXOpen rebuild, tested in the pilot	Works when the body decomposes into known features
Sheet metal parts	Your release’s solid-to-sheet-metal conversion first	Dedicated tooling, such as Sheet Metal from Solid in Designcenter X Essentials
Castings, forgings, freeform parts	Keep the body, edit synchronously	Automatic rebuild success is not guaranteed
Archive parts that will not change	Keep the body, validate only	A history nobody edits is cost without value
Future variant masters	Manual remodel as a master template	Design intent is a person’s decision

We’d advise against rebuilding the whole archive. Rebuild the classes that change often, keep the rest as validated bodies, and let the pilot numbers draw the line.

A Batch Pilot Plan, and What to Prepare

A STEP to parametric pilot answers one question per class: what share of parts passes the acceptance test without manual work? At inteliScience, NX and NXOpen specialists delivering from Romania for DACH engineering teams, we run it on your data in this order:

Sample about 50 migrated parts in three or four classes, plus 5 that your designers consider hard.
Fix history templates and acceptance criteria per class before the first run.
Run the batch and report per class, with failure codes.
Fix templates and rules, never individual parts. Rerun with 10 extra parts from outside the sample, so the pipeline is not tuned to it.
Review every failed part and a random share of passed ones, so the automated test is itself checked.
Decide per class: automatic rebuild, synchronous edits, manual remodel, or leave as is. Repeat the acceptance run after each NX upgrade; the next one, Designcenter 2612, went through Siemens’ Early Access Program in September 2026 (Siemens).

What to prepare (the full list for a joint session, with roles and access questions, is our NX AI workshop checklist):

Sample parts as STEP or migrated NX files, with the source system where known.
Drawings where they exist: they show nominal sizes and which dimensions carry intent.
Your modeling rules, wherever they live (often a Confluence wiki): feature order, sketch rules, expression names.
The desired history template per class, with seed part, layers and attributes.
Acceptance tolerances, and who signs off a class.
Where rebuilt parts are stored in Teamcenter and how the link to the source is kept.

If you would rather hand the rebuild over, our STEP file remodeling service has NX design engineers rebuild supplier STEP models as clean native NX parts, checked against the source geometry.

Frequently Asked Questions

Can Siemens NX turn a STEP file into a parametric part automatically?

Not with a documented Modeling command we could find. NX offers Synchronous Modeling for direct edits, recognition helpers for patterns, chamfers and blends, and feature recognition in NX CAM. A STEP to parametric rebuild with a real history is either manual remodeling or an NXOpen pipeline driven by class templates.

Synchronous Modeling vs. rebuilding the feature history: which is better?

It depends on how often the part changes. For a part edited once or twice, Synchronous Modeling on the imported body is faster and cheaper. For parts that change repeatedly or become variant masters, a rebuilt history with named expressions pays off, because the next designer edits a model instead of a body.

How much does it cost to rebuild migrated parts with a feature history?

The effort scales with the number of part classes and templates rather than the part count, because one template serves its whole class. That is why we start with a pilot: it measures success rate and review time per class before you commit to the archive.

How do you check that a rebuilt part matches the original?

Automatically, against the source body: solid validity, volume and area, centroid, bounding box, face and edge count, surface deviation within a tolerance, and a check that the feature tree matches the template. Compare Body and Deviation Gauge show the same comparison interactively for the parts an engineer reviews.

Does an AI rebuild work on complex parts such as castings?

Not reliably. Flanges and screws have been rebuilt with a history; for complex geometry with drafts, variable blends or freeform surfaces, success is not guaranteed. Keep those as validated bodies, edit them synchronously, or have a designer remodel them if they become variant masters.

Next Step

Bring twenty migrated parts from two classes and your modeling rules to a first call. We will assess which route fits each class and what a pilot with an automated comparison would measure. Talk to our engineers about rebuilding your migrated parts

References

Siemens Designcenter blog, “Designcenter NX, Tips and Tricks: Synchronous Modeling” (28 April 2026): https://blogs.sw.siemens.com/designcenter/designcenternx-synchronous-modeling/ (retrieved 2026-09-29)
TU Darmstadt, NX CAD tutorial, “Special characteristics chamfer and blend” (Label Chamfer, Label Notch Blend): https://www.iim.maschinenbau.tu-darmstadt.de/CAD-Tutorial/8_synchronous_modeling/1_feature_explanation/5_special_characteristics_chamfer_and_blend/index.en.html (retrieved 2026-09-29)
Siemens NX Design blog, “Designcenter NX, How To: Select Similar Faces” (24 February 2026): https://blogs.sw.siemens.com/nx-design/designcenter-nx-how-to-select-similar-faces/ (retrieved 2026-09-29)
Swoosh Technologies, “Feature Recognition and Grouping in Feature Based Machining” (6 March 2020): https://www.swooshtech.com/2020/03/06/feature-based-machining/ (retrieved 2026-09-29)
Siemens Solid Edge blog, “A new perspective on recognizing holes and hole patterns in Solid Edge” (11 September 2018): https://blogs.sw.siemens.com/solidedge/a-new-perspective-on-recognizing-holes-and-hole-patterns-in-solid-edge/ (retrieved 2026-09-29)
Siemens NX Design blog, “Compare Body, NX Tips and Tricks” (7 July 2023): https://blogs.sw.siemens.com/nx-design/compare-body-nx-tips-and-tricks/ (retrieved 2026-09-29)
NX Open C++ Reference, NXOpen::GeometricAnalysis::DeviationGaugeBuilder: https://www.ugapi.com/doc/NXOpen/a03653.html (retrieved 2026-09-29)
Swoosh Technologies, “How to use the Measurement Command in NX” (17 January 2020): https://www.swooshtech.com/2020/01/17/siemens-nx-measuring/ (retrieved 2026-09-29)
Applied CAx, “Examine Geometry and Heal Geometry, NX CAD Tips and Tricks” (14 July 2017): https://www.appliedcax.com/resources/nx-cad/examine-geometry-heal-geometry-nx-cad-tips-and-tricks/ (retrieved 2026-09-29)
Siemens NXOpen .NET Reference, NXOpen.MeasureManager (NewMassProperties): https://docs.sw.siemens.com/documentation/external/PL20191127135844554/en-US/nx_api_sc/nx/1926/nx_api_sc/en_US/nxopen_net/a06041.html (retrieved 2026-09-29)
NX Open C++ Reference, NXOpen::MeasureBodies, NXOpen::Body and NXOpen::Features::Feature: https://www.ugapi.com/doc/nxopen/a03988.html, https://www.ugapi.com/doc/NXOpen/a00707.html, https://www.ugapi.com/doc/NXOpen/a02827.html (retrieved 2026-09-29)
Kim, Pratt, Iyer, Sriram, “Standardized Data Exchange of CAD Models with Design Intent,” NIST (2007): https://www.nist.gov/publications/standardized-data-exchange-cad-models-design-intent (retrieved 2026-09-29)
Siemens NX Design blog, “Designcenter X: NX Essentials, How-to: Sheet Metal Best in Class Commands” (30 April 2026): https://blogs.sw.siemens.com/nx-design/designcenter-x-nx-essentials-how-to-sheet-metal-best-in-class-commands/ (retrieved 2026-09-29)
Siemens, “Welcome to the September 2026 Early Access Program for Designcenter” (Designcenter 2612, 31 August to 25 September 2026): https://resources.sw.siemens.com/en-US/presentation-welcome-to-designcenter-early-access-program/ (retrieved 2026-09-29)

STEP to Parametric in NX: 4 Routes and One Acceptance Test

STEP to Parametric in NX: The Short Version

The Problem: A Migrated Body Carries Geometry, Not Design Intent

Three Routes Without AI: Edit, Recognize, Remodel

Route 1: Keep the Body, Edit It With Synchronous Modeling

Route 2: What NX Recognizes, and What It Does Not

Route 3: Remodel by Hand Against the Imported Body

Route 4: Rebuild the Feature History With AI and NXOpen

The Acceptance Test: Rebuilt Part Against Source Geometry

STEP to Parametric by Part Class: Which Route When

A Batch Pilot Plan, and What to Prepare

Frequently Asked Questions

Can Siemens NX turn a STEP file into a parametric part automatically?

Synchronous Modeling vs. rebuilding the feature history: which is better?

How much does it cost to rebuild migrated parts with a feature history?

How do you check that a rebuilt part matches the original?

Does an AI rebuild work on complex parts such as castings?

Next Step

References

Company

Services

STEP to Parametric in NX: 4 Routes and One Acceptance Test

STEP to Parametric in NX: The Short Version

The Problem: A Migrated Body Carries Geometry, Not Design Intent

Three Routes Without AI: Edit, Recognize, Remodel

Route 1: Keep the Body, Edit It With Synchronous Modeling

Route 2: What NX Recognizes, and What It Does Not

Route 3: Remodel by Hand Against the Imported Body

Route 4: Rebuild the Feature History With AI and NXOpen

The Acceptance Test: Rebuilt Part Against Source Geometry

STEP to Parametric by Part Class: Which Route When

A Batch Pilot Plan, and What to Prepare

Frequently Asked Questions

Can Siemens NX turn a STEP file into a parametric part automatically?

Synchronous Modeling vs. rebuilding the feature history: which is better?

How much does it cost to rebuild migrated parts with a feature history?

How do you check that a rebuilt part matches the original?

Does an AI rebuild work on complex parts such as castings?

Next Step

Related Topics

References