Welding is one of the most important joining techniques used in the metalworking industries. In order to exploit efficiency potentials to a maximum possible extent, robot systems are adopted to automate the welding process whenever possible. This has become standard for high volume manufacturers like the automotive industry where thousands or millions of products are manufactured in the same or a very similar way. There is, however, a large obstacle for using automated welding robots in industries with highly individualized products: The programming effort for robot systems very often exceeds the manual welding times due to small lot sizes.
Therefore this paper proposes an approach to automatically derive welding programs. The building blocks of this technique are an automatic identification of feasible candidate weld joints from a CAD model, a knowledge base and a weld program compiler. After verification of these joints by a design engineer the machine instructions for the robot system have to be derived. Therefore the proposed approach defines the necessary structures of a Welding Expert Knowledge Base (WKB) where the relevant knowledge of a welding engineer designer is modeled in terms of rules of fuzzy rule sets. On top of the WKB we propose a weld program compiler which automatically identifies the sequence of weld joints and the necessary number of welding passes, generates optimized tool paths and finally derives appropriate parameter settings.