Cognition in technical systems, as the ability to autonomously generate decisions, enables the systems to cope with heavily changing situations. But as its future behavior is not known a-priori, cognition adds new attributes to safety issues. The paper presents the application of an architecture for cognitive safety control to a manufacturing scenario. The architecture consists of an operational part, which provides certified safety, and a strategic functional part which computes safe and optimal strategies based on appropriate dynamic models, safety specifications, and learned knowledge about avoiding safety critical scenarios and attaining goals. Thus, the architecture combines cognitive functions and certified components to obtain safe behavior.