Concurrent engineering and life-cycle design have made an outsized imprint on today’s engineering enterprise. By focusing primarily on detailed geometry, Design for Manufacture (DfM) tends to recommend incremental changes. While many of the methods developed have proven effective in isolation, reconciling the competing goals that the represent requires ever-greater design freedom. Increasingly this means moving DfM consideration upstream in the design process to conceptual design. This work presents a DfM methodology for decision-based conceptual design whose goal is to afford the consideration of all aspects of performance and functional requirements where design freedom is greatest – the earliest stages of the design process. This methodology draws on both formal reasoning and case experience for generating design alternatives by applying functional modeling. These functional abstractions are merged with real part geometry associated with manufacturing knowledge from reverse engineering toward generating manufacturable design skeletons. Such direct connection from function to manufacturable form afforded by this methodology allows the designer to make better-informed design decisions at the earliest design stage.

Design is often cast as an iterative process of solution generation and selection. This work also shows initial efforts of integrating multiple abstractions of computational synthesis and evaluating this synthesis to the accomplishment of both function and performance goals. Design abstractions described in this work help manage the problem size:​ high-level design concepts are generated and the best selected for further development. The selection process is at all times based on uncertain models both of design performance and design value. This research combines the computational synthesis across design abstractions for generating design candidates and decision-based conceptual design both selecting among them and for refining the evaluation model driving this selection. Normative methods from decision theory are used for controlling design search and evaluation.

A liquid dispensing system design case is described to explore the computational framework we proposed and to illustrate our DfM methodology in the domain of mechatronic design. It demonstrates both the promise of the methodology and the challenges presented by computer-aided conceptual design.