GENERATIVE AI FOR MES OPTIMIZATION LLM-DRIVEN DIGITAL MANUFACTURING CONFIGURATION RECOMMENDATION

Abstract

Many factories use Manufacturing Execution Systems (MES) to track and manage work on the shop floor. Setting up a single MES system can take weeks. It often needs expert help. The setup process involves selecting settings, defining workflows, and connecting tools or machines. Mistakes or delays in setup can result in poor outcomes or wasted time.
This paper examines how large language models (LLMs) can facilitate a more streamlined and efficient setup. The goal is to support teams by offering smart setup tips based on past work, system rules, and best practices. These tips focus on digital manufacturing, the market available tools that many factories use to run their MES.
The testing methodology utilized both structured data from MES system logs, current configurations, and previous records, and unstructured data, including work instructions, operator notes, system flow diagrams, etc. The model learned to suggest settings based on what had worked well in the past. When given a few inputs, it can now suggest complete setup plans. These can include production steps, quality checks, and machine data links.
The model also verifies if the suggested setup adheres to the rules in Digital Manufacturing. It flags problems early, so users can fix them before they go live. This helps teams avoid long delays later in the process.
To test this, we ran the tool on actual machine setups. It provided clear information and reduced setup time by almost 30%. It also helped new users get started quickly without needing extensive training. In many cases, the suggestions matched what experts would choose. The method is not meant to replace people. It helps them move faster and avoid simple errors. Experts still review the final setup. However, with fewer manual steps, they can focus on the more challenging aspects.
The research gap lies in the fact that digital manufacturing systems still rely on manual rules. Current tools can’t process large text-based instructions or connect them to real-time MES configurations. This work addresses that gap by testing automated configuration suggestions at scale.
In short, this paper demonstrates how innovative tools can support MES setup by providing valuable insights and practical tips. We used past setups and system rules to train the tool. Early results indicate that it can save time and enable users to produce better work. This can help factories utilize the digital manufacturing solution more effectively.
Method: In this paper, we propose an Enhanced Association Rule Mining (EARM) framework for feature selection on text classification tasks. EARM's process for finding rules works at different levels, helps distinguish between specific and overall rules, enhances traditional association rule mining, uses a new method to balance support, confidence, and lift metrics, and combines different rule mining techniques to make it more reliable.