Abstract With the advent of knowledge based techniques and tools, more and more automation engineers find themselves in the position of knowledge engineers. They have to formalize developmental knowledge and experience in order to make it accessible to the computer. A formal and yet easy-to-grasp language for the representation of knowledge from the automation field, PATHOS, was developed at IRP at the University of Stuttgart. In this paper, the fundamental aspects of knowledge representation in PATHOS are presented. The main part of the paper deals with the treatment of inexact knowledge and data, and proposes a scheme for the coupling of symbol-oriented knowledge bases with numerically oriented technical processes through linguistic variables and fuzzy logic.