Building probabilistic models for industrial applications cannot be done effectively without making use of knowledge engineering methods that are geared to the industrial setting. In this paper, we build on well-known modelling methods from linear dynamic system theory as commonly used by the engineering community to facilitate the systematic creation of probabilistic graphical models. In particular, we explore a direction of research where the parameters of a linear dynamic system are assumed to be uncertain. As a case study, the heating process of paper in a printer is modelled. Different options for the representation, inference and learning of such a system are discussed, and experimental results obtained by this approach are presented. We conclude that the methods developed in this paper offer an attractive foundation for a methodology for building industrial, probabilistic graphical models. 1