To enable the development of design-for-manufacturability tools for thin-walled mechanical components, converters are needed to convert design representations of components into representations with which manufacturability evaluations can be performed. In our work components are designed with features and represented as graphs. Conversely, manufacturing representations consist of manufacturing features, that in general are different than design features. Thus, the problem of converting feature-based representations arises. In this paper, manufacturing features are defined using graph grammars which provide a general, formal, structured description of how to convert features. Unfortunately, using graph grammar parsing as the basis for conversion is not computationally viable due to the nature of graph isomorphism determination. By taking advantage of the known structure of design representations and of manufacturing features, and by utilizing AI techniques for efficient pattern-matchers, it is shown that efficient converters can be constructed that do not require graph isomorphism determination. Conditions are given for the construction of polynomial time converters and a general conversion module is presented. An example converter is illustrated for tooling cost evaluation in injection molding and die casting.