Automated teller machines (ATM) belong to today's townscape as naturally as petrol stations belong to our road network. Despite large customer acceptance and wide dissemination, they still offer potential for optimization. In this thesis, the design methodology for mechatronic systems is adapted and applied to the analysis and optimization of the separation process inside an ATM. The integration of the often challenging model validation into the model-building process is emphasized. Along with this, practical steps of an approach to finding an adequate modeling depth are recommended. In the course of optimizing the banknote separation process, models in three different modeling depths are utilized: a 3-D FE model, a 2-D MBS model, and a highly idealized 1-D Simulink model. Identification and validation of the simulation models are achieved by means of measurement results fielded by specially developed automated roll-test bench. The outcome of the model-based analysis and optimization is a novel hierarchical control concept. At an underlying level, a draw-off roll control reduces the mechanical wear of the banknotes during the separation process significantly - without compromising functional safety. At a superordinate level, an iterative learning control concept ensures an automatic adjustment of the roller-pressing forces depending on the friction conditions between the banknotes. |