Error propagation analysis is an important part of a system development process. This paper addresses a model based analysis of spreading of data errors through mechatronic systems. Error propagation models for such kind of systems must use an abstraction level, which allows the proper mapping of the mutual interaction of heterogeneous system elements such as software, hardware and physical parts. A number of appropriate approaches have been introduced in recent years. The majority of them are based only on a data ow analysis. It is shown in this paper that for a complete picture the system control ow has to be considered as well. A new approach based on probabilistic control ow and data ow graphs is presented. The structures of the graphs can be derived systematically from an UML/SysML model of a system. The knowledge about an operational system profile allows the definition of additional system properties. Initially this model was developed for software errors localization. This paper shows its applicability to the error propagation analysis of an entire mechatronic system. The paper presents the modeling concept, the complete mapping process and application of the model for error localization. A reference robot control example demonstrates the main modeling steps.