Fraunhofer Institute of Optronics, System Technologies and Image Exploitation IOSB 
Motivation
With regard to Industry 4.0 technologies, the chemical industry and mechanical engineering have a great need to increase the efficiency and added value of processes and machines. In particular, the use of catalysts requires a predictive mode of operation of processes such as technical gas purification in order to minimize unexpected operational failures and the duration of maintenance-related downtimes due to aging and catalyst failure. Catalysts used to accelerate chemical reactions are indispensable components in both the chemical industry and technical gas purification. However, technical catalysts are subject to aging processes such as thermal deactivation (sintering), which leads to a loss of activity.
Short description of the project
As part of the "PrädiKat" project, an AI was developed that determines the aging state of a catalyst for methane oxidation (purification of exhaust gases from ship engines) based on current measured values. The aim was to be able to make a precise statement about how long a catalyst will remain active. This gives companies the opportunity to intervene in the catalyzed process with foresight. One focus of the project was on quantifying expert knowledge, since not all of the required variables could be recorded by sensors.
Project result
A partial result of the project is the experimentally determined data on the aging of the catalyst as well as the collected and quantified metadata, which together formed the data basis for the development of the AI solution. Various statistical techniques as well as ML methods were used in the project. It turned out that very good results can be achieved with classical neural networks. In order not to use these as a pure black box, the gradient of the network was additionally tapped during a prediction. The gradient allows conclusions to be drawn about which feature has the greatest relevance for the prediction step. This in turn provides the process expert with information on how to intervene in the process, for example, to extend the service life of the catalyst.