Standard Conformant Prompting (SCP): A Reproducible Framework for Compiler-Ready PLC Code Generation with Large Language Models : From Post-Hoc Verification to Standard-Driven Code Synthesis in Industrial Automation

Abstract

Generating PLC programs with Large Language Models (LLMs) that are directly deployable in industrial toolchains remains challenging because outputs must satisfy IEC 61131-3 semantics and vendor-specific compilation constraints. Verification-centric prompting such as Chain-of-Verification (CoVe) can reduce reasoning errors via post-hoc self-checking, but it does not enforce conformance during generation; consequently, outputs may remain non-compilable or structurally non-compliant. This paper introduces Standard Conformant Prompting (SCP), a compliance-driven prompting architecture that embeds IEC 61131-3 rules, vendor syntax profiles, declaration discipline, and safety-oriented structural templates directly into the prompt to constrain first-pass synthesis toward TIA Portal-targeted PLC code. SCP is evaluated on a fixed 25-industrial use cases under a reproducible protocol, with primary validation based on LLM-in-the-Loop (LITL) semantic assessment using five independent validator LLMs. Across validators, SCP achieves higher and more stable correctness ratings than CoVe while reducing safety-related failure modes and post-hoc correction needs; BLEU and expert Human-in-the-Loop (HITL) review are additionally reported as secondary comparisons. Future work will strengthen industrial applicability by adding compiler-and simulation-in-the-loop validation, extending modular vendor constraint libraries, and integrating SCP into real engineering workflows (e.g., TIA Portal and TwinCAT) to quantify commissioning effort, maintainability, and certification readiness.