Siemens and NVIDIA bring AI-Powered PCB inspection into the virtual world

Siemens is collaborating with NVIDIA to develop adaptive AI-powered manufacturing capabilities that combine industrial software, Siemens’ Digital Twin technology and NVIDIA Cosmos to accelerate production planning, inspection preparation and optimization

Siemens today announced a new collaboration with NVIDIA to develop adaptive AI-powered manufacturing capabilities that combine industrial software, Siemens’ Digital Twin technology and NVIDIA Cosmos to accelerate production planning, inspection preparation and optimization.

The collaboration combines Siemens’ Teamcenter® software for product lifecycle management (PLM) and Process Preparation X with NVIDIA Cosmos to create a new AI-powered workflow for PCB inspection, helping electronics manufacturers prepare and optimize automated optical inspection (AOI) processes before physical production begins.

The solution is being demonstrated at Siemens Electronics Factory Erlangen, Siemens' Customer Zero for adaptive AI-powered production – with the goal of creating a highly adaptive factory that can rapidly introduce near daily changes (3 to 5 times a week) without disruption.

“By connecting the comprehensive Digital Twin to the shop floor and AI we are improving both line productivity and product quality. At the Erlangen factory, we're demonstrating how the deterministic nature of Siemens’ Digital Twin is the foundation for leveraging physics AI for preparing production,” said Mark Hindsbo, executive vice president operations software, Siemens. “By leveraging a ‘virtual’ golden board, electronics manufacturers can change production dramatically quicker without taking the line out of production. This enables electronics manufacturers to respond more quickly to change and helps reach production readiness at scale and speed – while helping to reduce manual engineering effort, accelerate AOI readiness and improve manufacturing quality.”

“By connecting the comprehensive Digital Twin to the shop floor and AI we are improving both line productivity and product quality. At the Erlangen factory, we're demonstrating how the deterministic nature of Siemens’ Digital Twin is the foundation for leveraging physics AI for preparing production,” said Mark Hindsbo, executive vice president operations software, Siemens. “By leveraging a ‘virtual’ golden board, electronics manufacturers can change production dramatically quicker without taking the line out of production. This enables electronics manufacturers to respond more quickly to change and helps reach production readiness at scale and speed – while helping to reduce manual engineering effort, accelerate AOI readiness and improve manufacturing quality.”

Factory operator capturing images of a production PCB and marking areas for potential defects prior to handoff for manual AOI programming

Digital twin of SMT AOI inspection area at Siemens Electronics Factory Erlangen created using the Digital Twin Composer

Traditionally, introducing a new product or product variant requires a physical prototype before optical inspection processes can be prepared and trained for quality control. The new approach moves more of that work into the virtual world. Teamcenter provides PCB design, bill of materials and component information while Process Preparation X validates, normalizes and assembles this information into a Digital Twin of the printed circuit board assembly (PCBA). This provides the comprehensive product and manufacturing context needed to prepare inspection before the first physical board reaches the production line.

 
This Digital Twin of the “perfect” reference board, often referred to as the “golden board,” is used with NVIDIA Cosmos to create a fully photorealistic virtual golden board along with synthetic photorealistic samples that represent both good and false conditions created during production.  Siemens uses NVIDIA Cosmos Data Factory to generate synthetic circuit-board defect images, enabling inspection-machine calibration before physical boards are produced. 

Image capture of a production PCB and marking of potential defects prior to handoff for manual AOI programming

Digital twin modeling of the Golden Board is automated using Process Preparation X software

The workflow then extends into production through an AI-supported secondary inspection capability. When an AOI system identifies a potential defect, the secondary inspection capability can help analyze the detected condition and determine whether it represents a real defect or a false call. Together, these technologies are intended to create an end-to-end digital and AI-supported workflow spanning virtual AOI preparation, calibration and inspection support once production begins.

3D component models are included in the Golden Board PCB BOM in Siemens’ Process Preparation X software

Images of the Golden Board and potential defects are automatically produced from the digital twin of the PCB using Process Preparation X software and NVIDIA Cosmos

Customer Zero at Siemens Electronics Factory, Erlangen
The collaboration is being demonstrated at the Siemens Electronics Factory in Erlangen, Germany, which serves as Customer Zero for broader Siemens and NVIDIA work on adaptive, autonomous and AI-powered production. The factory contributes the manufacturing use case, production environment and subject-matter expertise in a highly adaptable, high-mix environment with ever-changing production processes. The project is targeting up to 85 percent less manual engineering effort in AOI program generation – with 40 percent fewer false calls and more than 5 percent first-pass yield improvement.

For manufacturers, the intended value is earlier inspection preparation, reduced manual engineering effort, lower dependency on physical samples, better-calibrated inspection processes, reduced manual verification, faster production readiness and improved manufacturing quality. The collaboration also demonstrates how the Digital Twin and industrial AI can connect engineering intent with real-world inspection across the digital and real stages of production and how the comprehensive Digital Twin evolves from a virtual representation of the product into an active foundation for AI-powered manufacturing and quality processes.