Contacts
Get in touch
Close

Contacts

Romania, Cluj-Napoca – 400157
Str. Ploiești Nr. 9

Romania, Vama – 447350
Str. Fermei Nr. 14

+40 774 463 429

info@inteliscience.net

NX MBD with AI

NX MBD with AI

NX MBD with AI: Associative PMI on the 3D Model, Placed in Your Live Session

Model-based definition in NX (NX MBD) stands or falls with the PMI on the model. An AI agent can place part of it today: dimensions that NX measures itself, attached to real faces and shown in the right model views. Datums, GD&T and tolerances stay with your engineers, and we say so before a project starts.

Short answer

NX MBD, model-based definition in Siemens NX, means the 3D model carries the product and manufacturing information (PMI), such as dimensions, datum features, feature control frames and notes, as semantic objects attached to its geometry, so the model rather than a 2D drawing is the reference for manufacturing and inspection. With AI, an agent creates part of that PMI in a running NX session through typed NXOpen tools: today associative linear and radial PMI dimensions, organized in model views, with every value measured by NX rather than typed in by the AI.

inteliScience, a Siemens NX, NXOpen and Teamcenter specialist delivering nearshore from Romania, builds these PMI tools and the MBD method around them: your guideline as the reference, NX’s own MBD tools where they fit, and the agent where no rule exists yet.

This page is about where AI fits into NX MBD and what an agent can put on the model in a live session today. Whether the model or the drawing is your master is a separate decision, covered in MBD or 2D drawing as the master. Preparing the routine PMI of whole part families from rules is Rule-Based PMI Creation; authoring PMI for one part in a conversation is the MBD & PMI Authoring Assistant; checking PMI before release is MBD/PMI compliance automation. The agent-based pages belong to our family of NX AI assistants and chatbots.

NX MBD with AI: the short version

  • What NX MBD is: PMI on the 3D model as semantic objects, organized in model views, so the model is the reference for manufacturing and inspection.
  • What the agent places today: associative linear and radial PMI dimensions on annotation planes the tool sets explicitly, PMI shown in the right model views, and existing PMI moved or re-attached. Hole and thread callouts are built and in testing.
  • What stays with engineers: datum schemes, feature control frames and tolerances. Our 3D PMI tools have no tolerance input, so the AI cannot invent one.
  • Where it fits next to NX: the MBD Logic Editor for rules you can write as logic, PMI Advisor and Check-Mate for checking, the agent for requests no rule covers yet.
  • What it is not: a 2D-to-3D migration service, a drawing generator or an automatic release.

What NX MBD puts on the model

In NX, PMI is created on the 3D part in the Modeling or Gateway application. Each PMI object exists twice: as a semantic object, listed under the PMI node of the MBD Navigator, and as a display instance in a model view. The semantic object is what downstream systems read; the display decides who sees it in which view.

The PMI types NX offers include dimensions (rapid, linear, radial, angular, chamfer, thickness, arc length and ordinate), datum feature symbols, datum targets, feature control frames, notes, symbols and PMI section views. PMI hole and thread callouts are associated with the hole feature rather than with its faces. For the hand-over, NX maps semantic PMI to STEP AP242, so a CMM program or a supplier whose software reads AP242 gets the same definition. A complete MBD model is a lot of objects, placed on the right faces, in the right views, to a standard. That volume is where automation pays.

NX’s own MBD tools, and where an AI agent fits

Siemens NX already automates parts of MBD. Before an agent is added, it is worth knowing which job each tool does best.

Siemens NX MBD tools compared with an AI agent that places PMI: what each does and what it is best at
Tool What it does Best at
MBD Logic Editor Rules, built as logic, that select geometry and create PMI automatically, for example: find the planar faces, measure the parallel pairs, group them, create linear dimensions Rules you can write down completely, applied the same way every time
PMI hole and thread callouts Callouts associated with the hole feature, with their content set by preferences Holes and threads, often the most repetitive PMI on a machined part
PMI Advisor and Check-Mate PMI Advisor checks PMI against configurable ISO or ASME rules, with pass and fail tags in the graphics window; Check-Mate adds GD&T validity tests Catching errors before release
AI agent with typed PMI tools (this page) Places PMI in the live session from a request, through NX’s own PMI builders, with the value measured by NX Requests no rule covers yet, parts outside your rule families, and the gaps between rules

Our rule of thumb: if a rule fits completely into the Logic Editor, write it there. A language model adds nothing to a rule that already runs as logic, and we build rule-based PMI too. The agent earns its place where rules run out: an annotation an engineer asks for in a sentence, a part outside the families your rules cover, or the gap between the dimensions a rule creates and the complete set the model needs. Why the guideline matters more than the model is covered in what worked with AI-generated PMI in NX.

What the agent puts on the model today

The PMI catalog is smaller than the modeling catalog, on purpose. Every type in it is created through NX’s own PMI builders; the status column says which ones are verified in a running session.

PMI types an AI agent can place on the 3D model in Siemens NX through typed NXOpen tools, with their status and how the tool creates them
PMI Status How the tool does it
Linear PMI dimensions Working in the live session The agent names two faces or edges; NX measures the distance on an annotation plane the tool sets explicitly
Radial PMI dimensions Working in the live session Attached to the circular edge or cylindrical face; NX measures the radius
Model views and PMI display Working in the live session PMI assigned to model views: one view with all PMI plus views per process step, a scheme we base on ASME Y14.41 and ISO 16792
Move and re-attach PMI Working in the live session Existing PMI moved to a clear position, or one dimension attached to several faces
Hole and thread callouts Built, in testing Created from the hole feature, which NX associates the callout with
Datum features, feature control frames, surface finish, notes Built for 2D drawings, not yet on the 3D model Our drawing tools cover the 15 geometric characteristics with MMC and LMC modifiers; on the 3D model these types are the next step
Tolerances on 3D PMI Not in the tools Set by the engineer; the agent has no input in which to invent one

The reference test continues the session in which the AI modeled a block with two holes, a blend and a chamfer, described on our page about AI CAD modeling in Siemens NX. On the finished block the agent then placed linear and radial PMI: measured by NX, attached to the faces, listed in the MBD Navigator like PMI an engineer creates in the PMI tab.

How an AI PMI request runs in NX

Take a request on an open part: dimension the overall length and the blend radius, and show them in the main view. The engineer asks once; the work follows a fixed pattern.

// NX MBD with AI: one request, typed PMI tools, one live NX session

Engineer

Asks for the PMI

Which dimensions, on which features, in which view, in plain language.

Read tools

Finds faces and PMI

The faces the request refers to, plus the PMI already on the part.

NXOpen · live session

The PMI builder runs

Annotation plane set, NX measures, the PMI is attached to the geometry inside its own undo mark.

Result

Shown in model views

The PMI appears in the views your guideline defines, ready for the engineer’s review.

The AI never types a measured value. It names the geometry; NX computes the dimension. A wrong pick shows up as a wrong reference the engineer can see, not as a plausible number that is quietly false.

  1. Read what exists. The agent lists the PMI already on the part, so it can add what is missing instead of duplicating it.
  2. Pick the geometry. The agent reads the faces and edges of the part and picks the ones it needs.
  3. Set the plane. The tool sets the annotation plane explicitly instead of inheriting whatever view is current.
  4. Create. NX’s PMI builder creates the dimension, and the value comes from NX.
  5. Show. The PMI is assigned to the model views your guideline defines.
  6. Check. The engineer, PMI Advisor or our compliance checks review the result before anything is released.

Where AI-placed PMI goes wrong, and what the tools do about it

In our builds, creating a PMI object was rarely the problem. Placing it correctly was.

That does not show in a demo on a perfect part. It is why we test every PMI tool on parts that look like yours before it goes live, and why the engineer reviews what the agent placed.

Datums, tolerances and release stay with engineers

A datum scheme encodes how the part is located, measured and assembled; a tolerance encodes what the function needs. Neither can be read off the geometry, and neither should come from a language model guessing. With our current tools the agent cannot set tolerances on 3D PMI at all, and datum features and feature control frames on the model remain the engineer’s work.

Where your standard already says what goes where, rules can prepare part of it: rule-based PMI creation prepares datum schemes from a family strategy and standard callouts, and the engineer completes and releases them. Release itself is a person’s decision, recorded in your PLM. Checking before release is a job for PMI Advisor, Check-Mate or our MBD/PMI compliance automation, not for the agent that placed the PMI.

NX MBD with AI and its neighbors on this site

MBD touches several of our pages. They cover different steps of the same process, and most projects use more than one of them.

How NX MBD with AI relates to the MBD versus 2D decision, rule-based PMI creation, the MBD and PMI Authoring Assistant, MBD/PMI compliance automation and AI CAD modeling
Page Step What it answers
MBD or 2D drawing as the master Method decision Whether the model or the drawing is the reference, and what MBD changes in your process
Rule-Based PMI Creation Rules over part families How routine PMI is prepared from your norms and templates
MBD & PMI Authoring Assistant Conversational, one part at a time How a designer gets PMI authored on request
MBD/PMI compliance automation Checking How PMI is verified before release
AI CAD modeling in Siemens NX Modeling tools How the same session builds the geometry the PMI sits on
NX MBD with AI (this page) PMI tools in the live session What an agent can place on the model today, next to NX’s own MBD tools

What NX MBD with AI does not do

An agent takes repetitive placement off the designer’s desk. It does not take over the engineering decisions behind the PMI, and it does not remove the limits of NX or of its API.

Good fit for AI in NX MBD

  • Machined, prismatic parts with many dimensions of the same kind
  • A written MBD guideline, or a set of good reference models
  • Teams moving to MBD who want routine placement off the designers’ desks
  • Engineers who would rather review PMI than click it into place

Limits we state up front

  • No datum schemes, GD&T or tolerances decided by the AI
  • Datum features, feature control frames, surface finish and notes on the 3D model are not in the catalog yet
  • No 2D-to-3D PMI migration and no drawing generator
  • After a model change the agent reads the part again; session IDs do not carry over
  • Release stays with a person; the agent never signs off its own work

For checks that run while a designer works, rather than on request, the event-driven live model check assistant is the better pattern. Two projects where rule-based PMI automation already runs in production, without a language model, are our automotive PMI automation case study and the automated PMI placement and numbering case study.

Deployment, data and hosting options

The PMI tools run in the live NX session, and NX stays usable between calls. Our current PMI tools are built and tested on NX 2606.

Part files stay in your network. What reaches an AI model is what a tool returns: face types, positions, the list of existing PMI and, where you allow it, an image of the part in a standard view. Where the models run is decided per project: through Anthropic, through cloud providers with EU regions, in Germany or on premises. Which PMI tasks a smaller local model handles well is something we test on your parts before we promise it.

How we introduce AI into your NX MBD

The order matters more than the model. A guideline that exists only in the heads of two senior designers is the first thing to fix, with or without AI.

  1. Guideline first. Which PMI each feature type gets, which model views exist, which standard applies, ISO GPS or ASME Y14.5. If it is not written down yet, we derive it with you from good reference models.
  2. Split the work. What goes into Logic Editor rules, what into typed PMI tools for the agent, and what stays manual.
  3. Read tools first. The agent lists existing PMI and geometry before it places anything, so the team sees it understand their parts.
  4. PMI tools on your parts. Each tool is tested on parts that look like yours.
  5. Checks in place. PMI Advisor or Check-Mate rules are set up, so every result is checked the same way, whoever placed it.
  6. Pilot. One part family and a small group of designers, measured on PMI placed, corrected and deleted.
  7. Rollout. More PMI types and more families, with recurring requests turned into rules.

Start with one part family in your NX

We do not start with every PMI type at once. We start with one part family, the PMI your designers place by hand on every part of it, and your guideline as the reference.

Fixed scope · fully credited

NX Proof-of-Value Sprint

One PMI task on your parts and a running prototype at the end: the step between a demo and a pilot.

  • One part family and one set of PMI, chosen with your MBD lead
  • NX’s own rules where they fit, the agent where they do not
  • Every dimension measured by NX, one undo step per change
  • Credited in full against the follow-up project
See the fixed-price entry points

On-premises or EU cloud · your tools stay yours either way

For comparison

PMI placed by hand

Every dimension clicked into place through the dialogs, view by view.

  • The same picks on every similar part
  • Placement and views vary by designer
  • Review effort grows with every part

Frequently asked questions about NX MBD with AI

What is NX MBD?
NX MBD is model-based definition in Siemens NX: the 3D model carries the product and manufacturing information (PMI), such as dimensions, datum features, feature control frames and notes, as semantic objects attached to its geometry, organized in model views. The model, not a 2D drawing, becomes the reference for manufacturing and inspection.
Can AI create PMI in Siemens NX?
Yes, for part of the PMI. An AI agent with typed NXOpen tools can place associative linear and radial PMI dimensions in a live NX session, show them in model views and move or re-attach existing PMI. The agent names the geometry; NX measures the value. Hole and thread callouts are built and in testing.
Which PMI types can AI not create in NX yet?
On the 3D model, datum feature symbols, feature control frames, surface finish symbols and notes are not in our tool catalog yet; we have built them for 2D drawings. Tolerances on 3D PMI are set by the engineer, and the agent has no input in which to invent one.
Does the AI decide tolerances or the datum scheme in NX MBD?
No. Datums and tolerances encode how the part functions, is located and is measured, and that is engineering judgement. Where your standard defines them per feature type, rules can prepare part of the scheme for the engineer to complete; the AI never decides them, and release stays with a person.
Is AI-created PMI in NX associative?
Yes. The tools create PMI through NX’s own PMI builders and attach it to faces and edges of the model, and NX computes the value. The result is a semantic PMI object like one an engineer creates in the PMI tab, listed in the MBD Navigator and shown in model views.
How does AI for NX MBD relate to the MBD Logic Editor and PMI Advisor?
They complement each other. The MBD Logic Editor applies rules you write as logic, the same way every time; if a rule fits there, it belongs there. PMI Advisor and Check-Mate check PMI against your standard. The AI agent handles requests no rule covers yet, through NX’s own PMI builders.
Where do our data and the AI models run?
The tools and the logs run on your infrastructure, and part files stay in your network. Models see tool results such as face types, positions and existing PMI, and images only where you allow it. Models can run through Anthropic, cloud providers with EU regions, in Germany or on premises, decided per project.
How do we start with AI for NX MBD?
With a workshop or a fixed-scope proof-of-value sprint on one part family: your MBD guideline as the reference, NX’s own rules where they fit, typed PMI tools where they do not, and a running prototype on your parts. The entry points are on our NX AI assistants and chatbots page; the sprint is credited in full against the follow-up project.

Name the PMI your designers place by hand on every part.

We will tell you which share belongs in NX rules, which an agent can place today, and honestly what has to stay with your engineers.

Talk to an Engineer

Siemens NX, NXOpen and Teamcenter specialists · nearshore from Romania, one hour ahead of CET · on-premises option