Problem: Industrial facilities, particularly energy systems, are characterized by high complexity and an immense volume of technical documentation, such as Piping and Instrumentation Diagrams (P&IDs). These documents contain essential information about components, processes, and their interconnections, which are critically important for the identification of optimization potential (e.g., energy efficiency). However, the manual analysis and cross-referencing of this information across numerous P&IDs is time-consuming, error-prone, and scales poorly within complex innovation processes. This leads to delayed decision-making and the insufficient utilization of the knowledge hidden within the document archives, consequently hindering the introduction of process innovations for efficiency improvement.

Practical objective: To develop an AI-supported, methodical approach for the automated, semantic analysis of P&ID diagrams using Large Language Models (LLMs). By transforming large volumes of complex, heterogeneous technical documentation into structured, machine-interpretable knowledge, this approach is intended to enable faster and more reliable identification of optimization potentials in industrial energy systems. In particular, the solution aims to reduce the manual effort and error-proneness of P&ID analysis, improve the ability to cross-reference information across many diagrams, and thereby create a scalable basis for identifying and assessing energy-efficiency and process-improvement measures.

Academic objective: Academically accompany and analyze the development of this AI-supported approach through the lens of Innovation Management theory. Depending on the specific focus, the thesis may draw on theoretical perspectives such as absorptive capacity, process innovation, dynamic capabilities, organizational learning, or corporate innovation systems to explain for example, how LLM-based technical document analysis can influence the generation, integration and utilization of knowledge, the design of innovation processes or the capability of industrial firms to identify and implement optimization measures.

This thesis is supervised in cooperation with etalytics GmbH.
Interested in this thesis? Please send your application including your CV, transcript of records and short letter to William Schütte