Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Knowledge Graph Modeling-Driven Large Language Model Operating System (LLM OS) for Task Automation in Process Engineering Problem-Solving

Sakhinana Sagar Srinivas, Vijay Sri Vaikunth, Venkataramana Runkana

TL;DR

The paper presents Process Engineering Operations Assistant (PEOA), a modular LLM OS for solving complex chemical and process engineering problems. It introduces a meta-agent–driven architecture that uses an action generator and domain-specific expert models to decompose tasks, select tools, and form tool-integrated solution trajectories, enhanced by Graph Retrieval-Augmented Code Generation (GRACG) and knowledge graphs. A teacher-student transfer learning approach using GPT-4 (Omni) creates synthetic tool-trajectory data to train smaller models, enabling domain adaptation and robust error handling through iterative reflection and revision. Comprehensive experiments on MathComp and ChemProc benchmarks show that POEA achieves performance close to leading proprietary LLMs, with ablation studies confirming the value of GRACG, instruction-tuning, and error-handling components, and additional Graph RAG comparisons highlighting the framework’s effectiveness in structured knowledge grounding. The work demonstrates significant potential for automating calculations, accelerating prototyping, and providing AI-augmented decision support for industrial processes, thereby enhancing efficiency, safety, and innovation in process engineering.

Abstract

We present the Process Engineering Operations Assistant (PEOA), an AI-driven framework designed to solve complex problems in the chemical and process industries. The framework employs a modular architecture orchestrated by a meta-agent, which serves as the central coordinator, managing an action generator and instruction-tuned small-scale language models (expert models). The action generator decomposes complex problems into sub-tasks and identifies suitable expert models to execute each, delivering precise solutions for multi-step problem-solving. Key techniques include advanced knowledge modeling using property graphs for improved information retrieval, facilitating more accurate and contextually relevant solutions. Additionally, the framework utilizes a teacher-student transfer-learning approach with GPT-4 (Omni) to fine-tune the action generator and expert models for domain adaptation, alongside an iterative problem-solving mechanism with sophisticated error handling. Custom datasets were developed to evaluate the framework against leading proprietary language models on various engineering tasks. The results demonstrate the framework effectiveness in automating calculations, accelerating prototyping, and providing AI-augmented decision support for industrial processes, marking a significant advancement in process engineering capabilities.

Knowledge Graph Modeling-Driven Large Language Model Operating System (LLM OS) for Task Automation in Process Engineering Problem-Solving

TL;DR

The paper presents Process Engineering Operations Assistant (PEOA), a modular LLM OS for solving complex chemical and process engineering problems. It introduces a meta-agent–driven architecture that uses an action generator and domain-specific expert models to decompose tasks, select tools, and form tool-integrated solution trajectories, enhanced by Graph Retrieval-Augmented Code Generation (GRACG) and knowledge graphs. A teacher-student transfer learning approach using GPT-4 (Omni) creates synthetic tool-trajectory data to train smaller models, enabling domain adaptation and robust error handling through iterative reflection and revision. Comprehensive experiments on MathComp and ChemProc benchmarks show that POEA achieves performance close to leading proprietary LLMs, with ablation studies confirming the value of GRACG, instruction-tuning, and error-handling components, and additional Graph RAG comparisons highlighting the framework’s effectiveness in structured knowledge grounding. The work demonstrates significant potential for automating calculations, accelerating prototyping, and providing AI-augmented decision support for industrial processes, thereby enhancing efficiency, safety, and innovation in process engineering.

Abstract

We present the Process Engineering Operations Assistant (PEOA), an AI-driven framework designed to solve complex problems in the chemical and process industries. The framework employs a modular architecture orchestrated by a meta-agent, which serves as the central coordinator, managing an action generator and instruction-tuned small-scale language models (expert models). The action generator decomposes complex problems into sub-tasks and identifies suitable expert models to execute each, delivering precise solutions for multi-step problem-solving. Key techniques include advanced knowledge modeling using property graphs for improved information retrieval, facilitating more accurate and contextually relevant solutions. Additionally, the framework utilizes a teacher-student transfer-learning approach with GPT-4 (Omni) to fine-tune the action generator and expert models for domain adaptation, alongside an iterative problem-solving mechanism with sophisticated error handling. Custom datasets were developed to evaluate the framework against leading proprietary language models on various engineering tasks. The results demonstrate the framework effectiveness in automating calculations, accelerating prototyping, and providing AI-augmented decision support for industrial processes, marking a significant advancement in process engineering capabilities.
Paper Structure (13 sections, 4 equations, 1 figure, 14 tables)

This paper contains 13 sections, 4 equations, 1 figure, 14 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Proposed Method
  4. Knowledge Modeling : Document Parsing/Indexing for Graph-Based Semantic Search and Retrieval
  5. Experiments
  6. Benchmark Datasets:
  7. Experimental Settings:
  8. Evaluating Tool Proficiency:
  9. User-Centric Evaluation:
  10. Experimental Results
  11. Ablation Studies:
  12. Additional Experiments:
  13. Conclusion

Figures (1)

  • Figure 1: The figure shows the architecture of the PEOA framework. It illustrates the key components and data flow of the framework, including the central meta-agent, action generator, and expert models. The framework processes user queries, selects specialized tools as expert models for sub-task solving, and employs error-handling mechanisms to generate solutions for complex chemical and process engineering problems. The interconnected nature of the components highlights the framework's ability to decompose tasks, select appropriate tools, and iteratively refine solutions through a sophisticated orchestration process.