"The most important and simplest thing is a single point of truth"

Dr Stefan Krämer on process optimization of plants and how engineering can provide support

Chemical engineer Stefan Krämer is an expert in improving process control and resource efficiency through digitalization. He heads the Process Performance Improvement group at Bayer, which is globally concerned with the optimal operation and energy efficiency of production processes by means of simulation, data analysis and process control. At NAMUR, he is responsible for the field of "Process and Operation Management Systems" and also teaches "Batch Process Operation" at TU Dortmund University. Stefan Krämer has been involved in EU projects on resource efficiency and is co-author of a reference book on the subject. He currently represents Bayer in the KEEN project - Artificial Intelligence in the Process Industry. If the father of two children still has free time, he likes to use it to go gliding.

Dr Krämer, in theory and practice you deal with numerous topics related to the optimization of process dynamics: What information from plant engineering do you need for this?

In order to optimize the control and system operation, various data is required. For very simple problems, process data is often sufficient. But even with a simple flow controller, you sometimes also need the valve design data from engineering to be sure you are working in the right valve range. And it quickly becomes more complex. At the latest, when it comes to new apparatus or complex issues for which a dynamic model is needed, we need access to P&I flow diagrams, data on dimensions and fittings, as well as the thermodynamics of the material flows in order to predict behaviour. And we need to know where the measurements are placed. So we need data from PCT engineering, plant construction and, in the case of existing plants, also the knowledge of the operating team.

In what quality do you usually receive this data and from which sources?

Most of the time, the quality is satisfactory. At the latest after discussion and demand, we get the necessary data in good quality, but typically analogue as drawings, data sheets and tables. We interpret it and then validate our models by asking further questions or comparing them with measured data. There can be many sources, such as P&I flow diagrams, equipment drawings, pump and valve data sheets, and measurement data from process control systems and plant information management systems. For many projects we also need live data from the process. They are also often spread far and wide and we need to bring them together.

Do you see potential for improvement in the information situation?

There are different levels. It becomes interesting when the required data is available digitally in the correct format, so that, for example, a simulation can be generated directly from the flow diagrams and equipment data of the engineering documentation. But a well thought-out and searchable data store alone would help in many places. But it is above all the elimination of data silos and the use of metainformation that will take us forward.

What needs to happen in the engineering tools used in your field to be able to optimize the operation of the plants more efficiently?

The most important and simplest thing is to have consistent data storage with a "single point of truth" so that you can rely on the information. We are working on an automatic connection of engineering data with simulation tools and the structuring of historical process data, which then enables an image of a plant to be created very quickly. On the other hand, it is not only important that the tools can do something, we humans have to keep control, for example with clear specifications.

If you could wish for something from engineering system manufacturers, what would that be?

An open exchange between different tools that doesn't take any effort.

Given scattered and inconsistent data sources, how do you think AI could help process engineers and automators?

Once an AI is able to bring together the hard facts and figures about the plant and its structure and answer simple questions directly, I could ask my AI engineering assistant: "Which pipe goes from tank A to B? And please tell me the diameter, what flowed through at 15:00 and how much." That would be something – and will probably make the work even easier than the combined data storage. If I can then fully rely on what the AI tells me, we're there!

Thank you very much for the interview, Dr Krämer!