Automated conceptual design often shows significant promise for generating conceptual solutions automatically through searching and synthesizing known principle solutions for a desired function. However, existing automated conceptual design systems lack an efficient heuristic search strategy and a knowledge-based planning mechanism, which lead to less chance of finding promising principle solutions within a limited time. This article proposes a novel semi-heuristic planning approach for automated conceptual design which comprises two successive steps, i.e. DESIGN PLAN and DESIGN COMBINATION. During the stage of DESIGN PLAN, the planning-graph technique is employed to extract a design plan from a planning graph, which is composed of various principle solution clusters. Under the guidance of the design plan, in the second stage, DESIGN COMBINATION combines compatible principle solutions from the principle solution clusters in the design plan until finding promising conceptual solutions. A prototype system and design case illustrate that the proposed semi-heuristic planning approach can successfully achieve an automated conceptual design synthesis.