Dr. Gregor TOLKSDORF from Evonik Technology & Infrastructure GmbH presented at the CAPE-OPEN 2019 Annual Meeting on what Evonik has in mind for the near future of process simulation and process engineering and how CAPE-OPEN could fit into that vision.
Dr. Gregor TOLKSDORF has participated to several CAPE-OPEN Annual Meetings (2015, 2016, 2017) while he was working on his Ph.D. at Technische Universität Berlin. His research work focused there on the development of MOSAICmodeling. He has co-authored several scientific papers describing how CAPE-OPEN fits with the approach of MOSAICmodelling (see for example paper in volume 40 of Computer Aided Process Engineering).
Gregor stated that CAPE-OPEN is not used at all at Evonik while the technology has been evaluated some years ago: its performance was found unsatisfactory. Evonik asked three questions to the CAPE-OPEN community:
- How about connections to other standardization communities?
As exemplified during the CAPE-OPEN 2019 Annual Meeting, CO-LaN is open to other standards. However most standards are data models while CAPE-OPEN is operating at execution time, when two applications need to talk one to another. Still data models could come handy for some aspects like giving semantic content to parameters of unit operations.
- Are “thermo performance” issues solved?
Infochem (now KBC), AmsterCHEM and Process Systems Enterprise jointly presented on performance with CAPE-OPEN thermodynamic interfaces at the CAPE-OPEN 2009 Annual Meeting so 10 years back from the CAPE-OPEN 2019 Annual Meeting: “Who Knows Where the Time Goes? Computational overheads of using the thermodynamics interfaces”. In particular, after having conducted many runs, they concluded that “Much of the overhead seems to be associated with the design and operation of the Material Object”. The training course organized within the CAPE-OPEN 2009 Annual Meeting brought helpful hints on what to do. They also pointed out that the Property Package design is also important for example stating that it is “essential to analyze the component list efficiently and only when SetMaterial is called”.
- How about integration of Machine Learning (e.g. regarding thermo)?
Machine learning is not apparently calling for any specific development by CAPE-OPEN. Within the APECS project, Reduced Order Models were developed as CAPE-OPEN Unit Operations by developing ways to automatically deduced these models from full scale CFD models.