Generative AI for Process Optimization: Exploring the Potential of Reinforcement Learning for Automated Process Improvement in Manufacturing

EasyChair Preprint 14305

Abstract

In the ever-evolving landscape of manufacturing, the drive for enhanced efficiency, reduced costs, and improved quality has spurred significant interest in leveraging advanced technologies for process optimization. Generative AI, particularly through the lens of reinforcement learning (RL), presents a transformative approach to achieving automated process improvement. This paper explores the potential of generative AI in optimizing manufacturing processes by developing RL models that dynamically learn and adapt to complex production environments. Reinforcement learning, with its capacity for continuous learning and decision-making under uncertainty, enables the identification and implementation of optimal strategies for process enhancement. Through a comprehensive review of current methodologies and applications, this study examines how RL can be integrated into manufacturing systems to autonomously fine-tune production parameters, anticipate maintenance needs, and minimize downtime. Case studies highlight successful implementations where RL has led to substantial gains in efficiency and productivity. The paper also addresses the challenges and ethical considerations in deploying AI-driven optimization in manufacturing, emphasizing the need for robust, transparent, and ethical AI practices. Ultimately, this exploration underscores the profound potential of generative AI and reinforcement learning in driving the next wave of innovation in manufacturing process optimization.

Keyphrases: Internet of Things (IoT), Reinforcement Learning (RL), Total Quality Management (TQM)

@booklet{EasyChair:14305,
  author    = {Abill Robert},
  title     = {Generative AI for Process Optimization: Exploring the Potential of Reinforcement Learning for Automated Process Improvement in Manufacturing},
  howpublished = {EasyChair Preprint 14305},
  year      = {EasyChair, 2024}}