Design problems with variable requirements force designers to explore multiple alternatives between problem space and solution space of a design task. Despite several methods have been already proposed, CAD models hardly meet these variable requirements. To close this research gap, this study develops a generative design approach, addressing the automation of designing alternatives by capturing design knowledge into a set of design elements. These elements use essentially knowledge-based parametrics in terms of design requirements. Such design elements are used to assist designers in the creation of required design variants. Thus, the exploration of design alternatives is transformed into configuration of desired design elements. As a result, the invested time in routine work is reduced and designers are able to concentrate on the exploration of required design alternatives. Moreover, the output configuration also offers potentials in computer-aided optimization, since the design elements not only capture the design knowledge, but also incorporate manufacturing restrictions. At last, this approach is illustrated by taking a bicycle peddle crank as an example.