Assessing the potential of LLMs for optimizing SCL-based PLC programs in industrial motion control

This article examines two artificial intelligence tools, GitHub Copilot and OpenAI ChatGPT5. The goal of the study is to determine whether general-purpose LLMs perform meaningful optimization of SCL code, specifically of an S7-1500 technological PLC. The optimization focuses on the structure of the code, its readability, and the reduction of its size. Additionally, an important parameter, the optimization of runtime, is also discussed. The code implements a motion control task, the kinematic control of a delta picker system. The results show that these tools are capable of interpreting the PLC code. However, ChatGPT5 was not able to write compilable PLC code. After optimization, the runtime of the PLC code became 31.79% faster. The results also indicate that the performance of LLMs is limited, engineering validation is necessary.