Global optimization using LLIMAS for the design of vibration isolation systems of high-uncertainty payloads

Abstract

Modern space system prototyping calls for bold, non-sequential approaches to engineering design. Embracing such an approach poses a new challenge on the design of passive vibration isolation systems: Accommodating large uncertainties in the payloads they support. Isolators are typically tuned and configured to the exact mass properties of the payload and do not perform well outside those assumptions. The number of candidate isolator configurations across which random vibration performance must be assessed also presents a significant challenge. This effort is observed to scale with 10^N, where N is the number of design variables studied. Here, 10^24 practical, unique designs were available. Our work describes the application of robust optimization techniques, global search algorithms, and massively parallelized job execution inside the LLIMAS software environment to overcome such computational challenges and identify isolator configurations that provide acceptable attenuation over a wide range of payload assumptions. Final geometry for the selected point design is presented, and performance comparisons of gradient-based, local, robust, non-robust and genetic algorithms are discussed.