Advantech and NVIDIA Unveil AI-Powered Factory Architecture for Next-Gen Manufacturing

Problem: Integrating AI into Real-Time Factory Operations Manufacturing facilities face increasing pressure to enhance production flexibility, optimize labor management, and improve energy efficiency. Traditional systems often struggle to provide the real-time insights and automation capabilities necessary for these objectives. The challenge lies in developing a comprehensive AI architecture that can seamlessly integrate with existing factory systems and deliver tangible improvements in operational efficiency and resource management.

Solution: AI Factory Brain and Physical AI Platforms Advantech, in collaboration with NVIDIA, has developed an AI-based factory architecture designed to address these challenges. The solution leverages the NVIDIA Factory Operations Blueprint (FOX) to create the AI Factory Brain, a multi-agent system led by a factory manager agent. This system utilizes NVIDIA's accelerated computing platforms, including the IGX Thor, Jetson Thor, and Jetson Orin, to process data from corporate systems like SAP, MES, and WMS, as well as edge sensors.

Key components of the architecture include: - **AI Factory Brain**: Monitors anomalies, identifies root causes, and coordinates agents and operators to address issues. It employs NVIDIA NemoClaw, Omniverse, Metropolis, and Isaac Sim to automate workflows, optimize energy use, and improve overall equipment effectiveness (OEE), yield, and mean time to repair (MTTR). - **Physical AI Platforms**: Advantech's AIR-427A and AFE-A702 platforms, powered by NVIDIA IGX Thor and Jetson Thor, support autonomous navigation, robotic manipulation, material handling, and shop floor coordination. These platforms enable AI agents to interpret workflow context, adjust production priorities, and assist operators in real-time. - **Edge Agentic AI**: Advantech's ICAM-540 camera and MIC-743-AT platforms, powered by NVIDIA Jetson Orin NX and Jetson Thor, support AI-driven automated optical inspection (AOI) and on-premises large language model (LLM) and visual language model (VLM) chatbot deployment. This enhances defect detection, increases production efficiency, and facilitates natural language interaction for factory operations.

Results: Proven Performance and Energy Savings The architecture has been validated through two AI agent pilots in Advantech's own manufacturing operations: - **iEnergy Agent**: This agent manages energy consumption by comparing production schedules with real-time Vision AI feeds and SCADA systems to control HVAC and lighting. The pilot projected a 10% reduction in total factory energy consumption upon full deployment. - **Production Line Efficiency Agent**: Using Vision AI, this agent captures real-time assembly line data, analyzes productivity metrics, detects anomalies and bottlenecks, and generates improvement recommendations and shift reports. Since its deployment six months ago, it has achieved a 12% improvement in assembly line productivity.

Enterprise-Grade Security and Safety The architecture incorporates NVIDIA NemoClaw, which includes the NVIDIA OpenShell secure runtime and policy frameworks, to provide security, access control, and operational governance. The MIC-735-IT and AIR-427A platform, powered by NVIDIA IGX Thor, also supports functional safety (FuSa)-ready AI computing for humanoid robotics and autonomous industrial systems, ensuring operational reliability and worker safety.

Future Outlook Building on its long-term collaboration with NVIDIA, Advantech plans to expand its edge AI and accelerated computing portfolio. This will support AI adoption across various domains, including smart manufacturing, AOI inspection, industrial robotics, warehouse automation, autonomous logistics, and digital twin applications.