Puneet Kaur
Associate Professor, Department of Electrical & Electronics Engineering, University Institute of Engineering & Technology, (UIET), Panjab University, Chandigarh
Abstract
Industrialprocess controlsystems generate vast amounts ofreal-time sensor data, making itchallengingto extract meaningful insights, particularly in relation to system dynamics and historical performance. This paper presents a Artificial Intelligence framework PROCON-AI, that leverages Large Language Models (LLMs) combined with Retrieval-Augmented Generation (RAG) using Lang Chain to enhance query capabilities for these systems. The proposed framework integrates real-time process data acquired from sensors, with domain-specific knowledge, enabling users to query operational status, fault diagnostics, and historical trends using natural language. The RAG mechanism enhances LLM responses by incorporating relevant information from industry-specific documentation and best practices, facilitating contextually rich and accurate outputs. The system is built using OpenAI's GPТ-3.5 Turbo but is also compatible with open-source models like LLaMA2 and LLaMA3. Performance evaluations across multiple industrial processes show significant improvements in query accuracy, fault diagnosis speed, and response quality when compared to traditional SCADA systems and non-augmentedLLMs.
Keywords-