Year: 2025
Type: conference-paper
Abstract: Modern intelligent manufacturing systems are inherently complex, driven by product variety, dynamic scheduling, and intricate interactions among humans, machines, and materials. This complexity presents significant challenges for production planning, control, and decision-making. To address these issues, we propose ABCM (Agent-Based Complexity Management), a novel method that integrates system dynamics simulation with a large language model (LLM)-based agent framework. ABCM utilizes PySD to simulate manufacturing processes and extract structured model outputs, which are analyzed through a suite of modular tools registered within a LangChain agent. These tools enable the computation of descriptive statistics, dynamic performance indicators, and complexity metrics—including entropy, rise time, overshoot, and integral errors—thus allowing for automated, data-driven complexity assessment. The agent autonomously interprets simulation results, identifies performance bottlenecks, and offers optimization recommendations. A case study demonstrates the practical application of ABCM in managing production variability and enhancing system responsiveness. The modular and extensible architecture supports scalable deployment in diverse intelligent manufacturing scenarios, contributing to improved adaptability, efficiency, and complexity control.
Authors: Xuan Kan, Yu Dai, Jiancheng Shi
Institution: Sichuan University
Language: English
Field: Decision Sciences
Domain: Social Sciences
Open Access status: hybrid
OpenAlex ID: W4414289159