In accordance with the principles of Industry 5.0, the development of technological solutions is directed toward supporting people in the industrial production environment. Recent trends place greater emphasis on the development of support systems that assist production operators in managing their activities by monitoring their performance and suggesting which corrective measures should be taken at a given moment. In this work, various decision support systems were developed and tested in the IndPenSim simulation environment for penicillin production. For this purpose, decision support systems were designed to provide operators with corrective action recommendations at predefined time intervals (every 24 hours). Three families of decision support systems were introduced, based on predictive model optimization, GNN-based recommendation systems, and hybrid methods. The results demonstrate that the application of different decision support systems can increase the final yield of penicillin, with the hybrid approach standing out in particular, as it exploits the advantages of the other two approaches. A comparison with prior studies confirms that the proposed methods reach yields equivalent to those of the strongest IndPenSim-based optimization models reported so far. Moreover, interventions in the process were required significantly less frequently than in previous studies. These findings confirm the suitability of the developed approaches for application to other processes as well.
|
