During the early phase of modeling, our knowledge of the system under development is often partial. Partial modeling technologies enable the explicit representation of partial knowledge, allowing developers to consider multiple design alternatives and manage uncertainties and potential contradictions. However, partial modeling is challenging, as models must adhere to design rules and well-formedness constraints, which are complex logic predicates. Therefore, it is important to exclude design alternatives that violate such constraints as early as possible to reduce the number of invalid design decisions and make the automated instantiation of design alternatives tractable.

In this paper, we propose a technique for translating well-formedness constraints into propagation rules, which are logic reasoning steps that can automatically exclude invalid design alternatives. Thus, propagation rules can reduce the number of apparent design decisions, thereby improving the productivity of development. Moreover, by applying propagation rules during the automated generation of design alternatives, the proposed technique can significantly improve the performance of logic solvers. We integrated the proposed technique into the Refinery graph generation framework and illustrated the performance improvement with case studies.