Siemens and NVIDIA Link Digital Twins with PCB Inspection

Siemens and NVIDIA are developing a workflow that uses a digital representation of a printed circuit board assembly to prepare automated optical inspection before the first physical board is produced. The announcement combines Teamcenter and Process Preparation X with NVIDIA Cosmos for synthetic images of acceptable and defective conditions. The work is being demonstrated at Siemens’ Erlangen electronics factory. Its stated performance improvements are project targets, rather than measured results that every customer can assume it will obtain.
A virtual reference must represent the production item
Preparing inspection from design data can reduce dependence on waiting for a physical reference board. The benefit depends on whether the virtual assembly reflects the configuration that will actually be manufactured. Component identification, orientation and the applicable bill of materials need to remain aligned. A photorealistic image of an outdated board is not a useful reference for accepting the current production item.
The announced workflow includes validation and normalisation of the source information before the virtual board is assembled. That step matters because design datasets can contain different representations and levels of completeness. The engineering team should identify which attributes are needed for inspection preparation and which gaps require resolution before synthetic examples are generated. The quality of the source definition limits the quality of the simulated reference.
Synthetic defects need manufacturing relevance
Generating many images is useful only if the examples represent conditions the inspection system needs to distinguish. The defect definition should relate to the applicable acceptance criteria, not simply to visual variation. An unusual appearance may be acceptable, while a subtle difference may require attention. Quality specialists should therefore help define the examples and their expected classification.
The relationship between synthetic and physical images also needs assessment. Lighting, materials and the actual optical arrangement can influence what the production camera records. A virtual preparation method should be checked against representative boards once they become available. That does not remove the earlier preparation benefit; it establishes how confidently the synthetic material supports the installed inspection process.
False calls and missed defects are different risks
The project includes secondary AI-supported review of potential defects identified by AOI. Reducing false calls can save manual effort, but the evaluation must also check whether actual defects remain detectable. A system that rejects fewer boards is not necessarily better if it overlooks conditions that matter to function or reliability. The acceptance study should report both sides of that trade-off.
The same principle applies when the board design changes. A method trained or configured for one assembly should not quietly be assumed valid for another. The workflow needs a rule for determining when new preparation or verification is required. Change control connects the digital twin to the quality process rather than leaving the generated images as an isolated development dataset.
Keep inspection decisions traceable
The virtual reference, image-generation configuration and production inspection version should remain linked to the applicable board revision. When a result is challenged, the team should be able to identify what the system was expecting and which evidence supported its classification. A faster preparation route is more useful when it preserves that ability to investigate.
The collaboration is significant because it brings inspection preparation earlier into the product-development process. Its practical value will depend on the fidelity of the product definition and on validation against real production conditions. The engineering objective is earlier readiness with dependable acceptance, rather than assuming that synthetic realism alone establishes an effective quality-control method.
Source — dailycadcam.com. Source publication date 2026-10-22. Engineering analysis: The CAD Verse.



MEMBERS · ONE ENGINEERING COMMUNITY
Engineering videos and tutorials
Engineering marketplace and professional opportunities
Courses, tutorials and learning paths
Engineering files and CAD models
Professional engineering projects
Discussion
0 CommentsShare your thoughts and engineering insights with the community.
Comments are closed for this article.
No comments yet. Be the first to share your thoughts.