Cyber Physical Systems (CPSs) are both software and hardware intense systems which are integrated into the digital world. An increasingly relevant application area for CPSs are software-driven industrial Production Automation Systems. A major driver of these Cyber Physical Production Systems is the need to integrate intelligent and smart functionalities into the production process at runtime that can run in the cyber world and do not radically change the underlying control flow of the physical plant. To gather the necessary knowledge for smart functionalities both cyber and physical parts must be captured, kept and analyzed together in a suitable CPS architecture – also despite frequent evolutionary changes during their whole (usually relative long) life span. This paper addresses this problem by presenting a service-component architecture for smart service-oriented CPSs which can manage implicit knowledge by acquiring, handling and analyzing domain specific service-components with respect to the CPS context.