TITLE: MECHANICAL DESIGN WITH A BAYESIAN EXPERIMENTAL AUTONOMOUS RESEARCHER (BEAR)
ABSTRACT: Additive manufacturing (AM) has increased the complexity with which structures can be designed and fabricated. Computational tools, empowered by the control afforded by AM, have enabled the discovery and realization of structures with enhanced or tailored mechanical performance. However, this approach is limited to mechanical properties that can be reliably predicted using simulation. For other properties dependent upon geometric and materials nonlinearities, design typically occurs manually through iterative manufacturing and physical testing. This manual approach is impractical given the expansive design space now accessible with AM. Additionally, AM has introduced a host of new defects for which researchers and practitioners do not have the benefit of empirical engineering guidelines built upon decades of intense study. As a result, how best to design and optimize structures for properties that are challenging to simulate remains an open question.
In this work, we address the challenge of designing and optimizing structures for properties that are challenging to simulate by realizing a Bayesian experimental autonomous researcher (BEAR) that can plan and conduct experiments without human intervention. First, we describe how the high-throughput nature of this system, relative to manual testing, allows for the comprehensive exploration of a large family of structures over hundreds of experiments, a previously impractical concept. Then, we present the results of experimental campaigns conducted by the BEAR resulting in the identification of high performing structures in 60 times fewer experiments than a grid-based experimental campaign. Lastly, two additional projects are proposed. The first details an exploration of the response and uncertainty of a mechanical system that is highly sensitive to imperfections. The second seeks to leverage and modify the BEAR to optimize mechanical metamaterials for impact performance.
Collectively, this work shows the potential for BEARs to impact fields where computational tools are imperfect and experiments are slow and complex. The use of autonomous research systems for the design of structures for properties that cannot be effectively simulated represents a shift in the conventional design process and could have an impact in materials development and mechanical design in a manner that facilities the convergence of machine learning, physical experimentation, and design.
COMMITTEE: ADVISOR Professor Keith Brown, ME/MSE/Physics; Professor Elise Morgan, ME/MSE/BME; Professor Emily Whiting, CS; Professor Kristofer Reyes, MDI, University of Buffalo