CHS: Medium: Collaborative Research: Discovery and Exploration of Design Trade-Offs

Massachusetts Institute Of Technology, Cambridge MA

Investigators

Abstract

To build functional, complex, and integrated products, we need tools that allow engineers to efficiently explore a design space that encompasses geometry, materials, electronics, software, and fabrication processes, and these computational tools must be able to simultaneously optimize conflicting objectives such as mechanical performance, weight, and cost. Currently, experts and novice users alike go through a long process of trial-and-error to identify top-performing designs, and a critical bottleneck then occurs when engineers convert concepts created by designers into high-performance products because design variations that improve one performance metric may worsen another, so that there is no single best possible design but rather a set of "Pareto optimal" designs that reflect different trade-offs. This research will create algorithms that can discover the Pareto front of trade-offs efficiently and can then present it to engineers and designers seeking to optimize the performance of a design along more than one axis. Project outcomes will make foundational contributions to computer-aided manufacturing, with demonstrations on a variety of pipelines for creating different types of objects. The resulting methods will both allow engineers to create high-performing products in less time while also dramatically improving the quality of the products themselves. The work will have additional broad impact through involvement of both undergraduate and high school students in the research activities. To achieve the project's goals, novel methods will be developed for capturing the full space of optimal design trade-offs whose shape is typically nonlinear and even disconnected. These methods will be driven by probabilistic analysis in the context of multi-objective stochastic optimization. Efficient representations and algorithms for exploring trade-offs with real-time feedback, showing which designs achieve desired performance characteristics, will also be defined. These techniques will then be used to implement a system that conveys an intuitive understanding of design trade-offs, allowing engineers to quickly discover desired solutions or find efficient ways to iterate on the design specifications. In addition, for complex designs requiring assembly-based approaches, tools will be created for iteratively constraining and refining the design space until the engineer feels confident with the assembled artifact. The workflow will allow engineers to find solutions that are better than what a human or a computer alone would find. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.