Center for Computational Engineering Seminar (CCE) HOST: Professor Alan Edelman DATE: Thursday, December 13, 2018 TIME: 2:00 PM - 3:00 PM LOCATION: Building 32, Room 124 (STATA) 32 Vassar Street, Cambridge TITLE: Machine-Learning Enhanced Energy-Product Design SPEAKER: David Tew (ARPA-E) ABSTRACT: Engineering design processes are generally characterized by iterative attempts at the solution of a well-defined market problem. Each iteration is frequently characterized by a hypothesis generation/conceptual design phase where low-fidelity/reduced order models are used to refine a high-level solution concept. In the next phase, the winning solution concept is then subjected to a high-fidelity (i.e. expensive) detailed design and evaluation process that ideally culminates in the demonstration of a successful solution to the problem. If not, the initial hypothesis is updated using lessons learned in the evaluation phase, and iteration continues until the solution is obtained or the effort is abandoned. Recognizing that high-risk and high-cost design processes are frequently significant barriers to entry for energy-efficient products and that emerging machine learning/artificial intelligence techniques have the potential to lower the cost and risk of certain aspects of the above-described energy-product design process, ARPA-E is seeking to accelerate the application of these techniques in the engineering design process to help engineers-- 1) to develop better & more novel product concepts, 2) to more efficiently execute the high-fidelity optimization of these concepts, and 3) to execute "inverse design" (i.e. no iteration) process for "simple" energy products components. ========================================================= Massachusetts Institute of Technology Cambridge, MA _______________________________________________ CRiB-list mailing list CRiB-list@mit.edu http://mailman.mit.edu/mailman/listinfo/crib-list