While Artificial Intelligence (AI) methods are already present in many consumer applications, they are currently less used and accepted in the process industry and in the field of engineering. At the same time, the availability of digital engineering data is increasing due to the increasing digitization and standardization of data formats, for example DEXPI (Data Exchange in Process Industry). This work provides answers concerning important questions that need to be addressed when applying AI methods in engineering. How must plant data be structured and which standards can already be accessed? Which areas of engineering are promising for the application of AI methods and which models can be used? In this context, new and innovative approaches for data-driven modeling of engineering data will be introduced, applied and evaluated. A graph-based, machine-learning information model is used to provide a basis for describing information from plant topologies in the form of a piping and instrumentation diagram (P&ID) and process simulations based on unit operations. The use of machine learning models enables to learn the relationship between process data and separation units in conceptual engineering to support the generation of downstream processes. In detail engineering, AI algorithms are used that learn the topology of process units. In this way, Recurrent Neural Networks predict P&ID components, while Graph Neural Networks check the consistency to support the engineering and drawing of P&IDs. The combination of the graph-based information model with a deterministic algorithm also enables automated safety assessments in an early engineering and design phase. The different modeling approaches will be validated in initial feasibility studies and integrated into a smart engineering workflow and first prototype applications as part of this work.weiterlesen