We previously explored the significance of log files at Edge AI applications in enhancing root-cause analysis and risk assessment in large scale manufacturing. We discussed how the strategic use of log files combined with advanced technologies like Language Models (LLMs) and Retrieval-Augmented Generation (RAG) applied at the edge can provide a deeper understanding of operations, enabling proactive, precise, and efficient problem-solving.
Now we look deeper into the benefits of integrating Edge AI with log files. We show how such a system works together when introducing a new feature for further enhancement of accuracy and efficiency: Equipment Manual Reading with AI.
The Importance of Log Files in Manufacturing
As discussed before, log files are essential for capturing detailed, time-stamped information about machine performance, system errors, operational deviations, and maintenance records. These files provide a comprehensive record of the operational state of machinery and processes, enabling informed decision-making and proactive problem-solving.
The Role of Edge AI in Industrial Manufacturing
By bringing LLMs and RAG to the edge, we can process data in real-time, reducing latency and enhancing security. Edge AI enables localized processing, allowing for faster analysis and decision-making. This approach also supports scalability, flexibility, and adaptability to various operational scales.
Component Interaction of this application
The following interaction graph shows docker containers and database interaction of the whole application which combines the functionality of the Log File Analysis and the Equipment Manual Reader. This is also giving flexibility in deployment and scalability. To achieve transfer learning Edge Nodes with similar „skill set“ are supposed to learn and interact with each other.
Introducing Equipment Manual Reading with AI
Equipment manuals contain critical information about tool performance, specifications, and operational parameters. However, extracting this information manually can be time-consuming and prone to errors. AI-powered equipment manual reading offers a solution to this challenge.
By leveraging AI models, we can automatically extract relevant information from equipment manuals, including SECS/Gem manuals, which contain critical data variables and status variables. These variables, such as DVID (Dynamic variable identification) and SVID (Status variable identification), are essential for understanding tool performance and optimizing production processes.
Technical Aspects of Equipment Manual Reading with AI
To develop an effective AI-powered equipment manual reading system, we need to address the following technical challenges:
- Document structure and naming conventions: Equipment manuals often lack a clear data structure, and document names may not accurately reflect their content. To overcome this challenge, we can use natural language processing (NLP) techniques to analyse document titles and contents and identify relevant information.
- Version control: Manuals are versioned, and updates may not always include all initial documents. To address this challenge, we can use version control systems to track changes and updates to equipment manuals and ensure that the AI model is trained on the most recent and relevant information.
- Missing documentation: Older tools may not have available documentation, requiring alternative methods for data extraction. In such cases, we can use machine learning algorithms to predict missing information based on patterns and trends in available data.
- Format and quality variations: Documents may be in different formats (PDF, Word, Excel) and quality, requiring adaptable AI models. To address this challenge, we can use format-agnostic AI models that can handle different document formats and quality levels.
AI Models for Equipment Manual Reading
Several AI models can be used for equipment manual reading, including:
- OCR models: Text extraction using open-source OCR solutions like Tesseract, OCRopus, and GOCR. These models can be trained to recognize and extract relevant information from equipment manuals.
- Transformer models: LLMs like Aryn/deformable-detr-DocLayNet for document information extraction. These models can be fine-tuned for specific equipment manuals and formats.
- NLP models: Spark NLP for large language models (LLMs). These models can be used for natural language understanding and information extraction from equipment manuals.
- Python packages: pypdf, python-docx, pandas, and python-calamine for reading out pages from PDF, Word, and Excel documents. These packages can be used to pre-process documents and extract relevant information.
In the following graph we see how this would look like working with Spark NLP.
Conclusion: Integrating AI-Powered Equipment Manual Reading with Edge AI and Log Files
By integrating AI-powered equipment manual reading with Edge AI and log files, we can create a robust system for manufacturing that enables proactive, precise, and efficient problem-solving. This approach supports a culture of continuous improvement and innovation in manufacturing processes, ensuring enhanced operational reliability, product quality, and cost savings.
For deeper understanding please feel free to get in contact with us!
Content was presented at EEAI 2024 Conference . We are partner of EdgeAI Project. EdgeAI “Edge AI Technologies for Optimised Performance Embedded Processing” project has received funding from Chips Joint Undertaking (Chips JU) under grant agreement No 101097300. The Chips JU receives support from the European Union’s Horizon Europe research and innovation program and Austria, Belgium, France, Greece, Italy, Latvia, Netherlands, Norway.
