The “Engineering Data Funnel”: Knowledge-Enhanced, Agentic-AI-Based Data Processing for Automation Engineering
Nicolai Schoch, Mohamed Elsheikh, Mario Hoernicke, Nika Strem, Katharina D.C. Stärk, Sebastián M. Palacio, Virendra Ashiwal · IFAC-PapersOnLine · 2025
Today, many software applications exist to support the engineer in doing the process and automation engineering, but a major challenge is the initial transforming of unstructured multimodal engineering design data (like P&I diagrams or control narratives) into structured formats for further processing. This transformation still requires manual effort, which is time-consuming and error-prone. In this work we present the “Engineering Data Funnel” (EDF), a system which converts unstructured data from formats like PDF or paper into structured formats for established engineering applications. The EDF uses a neuro-symbolic agentic AI workflow, where specialized expert models process multimodal data under consideration of domain expert knowledge and in mutual coordination. It thus improves robustness, correctness, and consistency of data processing. We present the setup and demonstrate its strengths by means of examples from different industry segments. With the EDF, through obtaining a structured representation of the initial engineering design data, we provide the foundation for the digital twin of the industrial process plant, which can be utilized and built upon throughout the entire plant life cycle.