The following is a special contribution to this blog by CCC Executive Council Member Mark Hill and workshop organizer Wade Shen of MIT Lincoln Laboratory.
Recently DARPA ISAT organized a workshop on the future of Computer Aided Design tools called “Rethinking CAD”. The purpose of this workshop was to bring together industrial and academic experts from manufacturing and mechanical design fields to understand the limitations of current tools for exploration of design trade spaces for geometry, materials and microstructure in light of recent developments in 3D printing and additive manufacturing. Here, we summarize the state of today’s technologies and the results of the DARPA ISAT workshop (full slides are available here).
During our workshop, panel members addressed the current state of CAD tools and their limitations. 3D printing and additive manufacturing methods have made it possible to build hardware quickly and easily. As a result mechanical prototyping is or soon will be limited by design.
Unlike software and electronic design, where specification of behavior (code) and testing are tightly integrated into IDEs, most CAD tools provide limited or no integrated models of physical and material properties to designers. Simulation is mostly done after the fact to validate designs. This means that designs must propose geometries and test material properties in a non-interactive fashion to understand if geometries that they would like to build can satisfy requirements/constraints.
Current tools are also very limited in terms of their ability to capture and make use of design requirements in such a way that they become testable during the design process. Suppose that you’d like to design a truss that can withstand a shearing force of 10Nm along its X-axis. CAD tools allow designers to specify geometry and material specifications, but behavior, requirements and function are not part of today’s tooling nor are there ways to make use of these specifications for the purpose of testing and validation during the design process.
Finally, it was clear to all participants that the exploration of design spaces requires automated or semi-automated synthesis. By this we mean that with functional/behavioral specification the space of possible designs is potentially very large and automating the process of generating, testing, and optimizing these designs allows for much faster exploration of this space. We are currently limited by the implicit knowledge (and design rules) of human designers and because of the long design cycle between geometry, simulation, prototype builds. Even with the advent of fast additive manufacturing methods, exploration of these spaces is extremely limited. Many electronic design environments already do optimization and simulation automatically for selected problems (e.g. component placement, interconnect routing, etc.), but the mechanical design tools are still lacking.
How can we make CAD better? The convened experts agreed on the need for the ability of design languages that communicate function, requirements and constraints in a computer-interpretable manner. Such specifications would enable automated testing/evaluation of design requirements and specifications and automated design generation and optimization, i.e. tools could suggest design optimal geometries, materials and structures. Automated design and testing requires more information than we currently supply our CAD tools. All attendees agreed that better, more natural interfaces are needed to make it possible to express this information easily and efficiently. The ability to “infer” or “fill-in” aspects of CAD designs based on knowledge of the user’s intent or based on prior designs could ease the user-interface burden.
