From Natural Language to Code: AI Automation in Cyber-Physical Manufacturing Systems

The complexity of modern manufacturing systems necessitates advanced control programs, yet traditional heuristic and formal methods are hindered by being error-prone, time-consuming, and complex. This paper proposes an AI-driven approach to automate control program development for Cyber-Physical Manufacturing Systems (CPMS). Using Natural Language Processing (NLP), our system converts user-defined natural language requirements into executable control code. We developed the AutoFactory dataset using Large Language Models (LLMs) to capture diverse manufacturing requirements. We fine-tuned the BERT model to accurately extract key components such as actuators, sensors, and pre-actuators from these specifications. Our approach aims to fully automate the transition from informal requirements to formal control programs, significantly reducing the need for manual intervention and minimizing human error. Future work includes developing a library of code for each component, enabling automatic code generation without requiring prior experience with PLC programming languages. This method leverages the speed and efficiency of computers to generate accurate control code swiftly, democratizing control program development and paving the way for more efficient, reliable, and scalable manufacturing systems.