AI Assistant for Smart Manufacturing with AWS IoT TwinMaker and Amazon Bedrock
AWS demonstrates a real-world implementation with its Cookie Factory sample solution targeting manufacturing customers. The challenge is to extract insights from unstructured manufacturing data and improve production monitoring and troubleshooting. The solution is an AI Assistant integrated with a digital twin built using AWS IoT TwinMaker, Amazon Bedrock for LLMs, AWS IoT SiteWise for equipment measurements, and Amazon OpenSearch Service for scalable document search. The AI Assistant uses generative AI large language models to interact naturally with operators and managers, diagnose production issues, suggest root causes, and navigate 3D views of the manufacturing site for troubleshooting. This is implemented as a Python application with chat UI and LLM agent workflows orchestrated with LangChain, applying a retrieval augmented generation (RAG) pattern grounded in a knowledge graph and user documentation. Impact includes simplified diagnosis and troubleshooting, enhanced user experience with interactive 3D visualization, and a foundation for customized advanced manufacturing AI solutions.
- Organization
- Cookie Factory
- Industry
- Manufacturing
- Location
- United States
- Published
- November 2023
Reported outcomes
Strategic outcomes
Primary read
Use case focus
Showing 3 of 3
- 1AI-assisted troubleshooting
- 2Digital twin interaction
- 3Generative AI chat assistant
- Manufacturing data is largely unstructured, complicating insight extraction and production issue diagnosis.
- Operators and managers lack intuitive, efficient ways to interact with complex manufacturing data and digital twins for troubleshooting.
- Developed an AI Assistant module integrated with the Cookie Factory digital twin for manufacturing monitoring.
- Used AWS IoT TwinMaker to create a knowledge graph and represent physical systems as digital twins.
- Powered AI Assistant LLM workflows with Amazon Bedrock, leveraging foundational models like Claude and Amazon Titan.
- Incorporated AWS IoT SiteWise to provide real-time equipment data and simulated alarm event sources.
- Implemented chat UI for natural language interaction, using LangChain to orchestrate multiple LLM chains for complex workflows.
- Used a retrieval augmented generation (RAG) pattern and grounded LLM responses with the Knowledge Graph and user documents.
- Leveraged Amazon OpenSearch Service for scalable document search backing the RAG workflow.
- Enabled operators/managers to use natural language to diagnose and troubleshoot production issues efficiently.
- Improved user experience with 3D visualization tied to AI diagnoses and SOP document navigation.
- Provided a modular, extensible AI assistant blueprint for smart manufacturing applications using AWS AI and digital twin technologies.
Architecture
The AI Assistant is a Python application module within the Cookie Factory sample solution. It provides a chat UI front end and an LLM agent backend implemented using LangChain. The LLM agent executes multiple chains including a router chain to delegate workflows, chains for querying the AWS IoT TwinMaker Knowledge Graph, chains implementing the retrieval augmented generation (RAG) pattern with user-provided documentation in S3, a fallback chain, and using foundational models from Amazon Bedrock. The system also uses AWS IoT SiteWise for equipment sensor data and simulated alarm event sources for context.
Sources & evidence1
AI-generated summary. Verify important details with the linked sources before relying on this case.
Explore related AI use cases
Was this useful?