A network‐based approach to identifying key components in structural vibrations

Modeling the dynamics of systems with many interacting components, such as robots, wind turbines, and trusses, remains challenging today. These systems often display complex oscillatory responses to external inputs, and harmful vibrations might be excited along with the desired motion. Understanding the relative importance of individual components or systems aspects to the overall system dynamics could be a vital step towards focused design and maintenance efforts. This work proposes a network‐based approach to studying the dynamics of a mechanical system by representing the system as a network of coupled oscillators, where each node corresponds to a machine component and each link denotes a physical connection, such as a weld or bolt. Inspired by studies of dynamics in biological and social networks, we show how network measures can be used to predict the importance of a single oscillator, or component, for shaping the overall dynamics. We further demonstrate under which conditions these conclusions are possible, and where the metrics fail. This study hopes to contribute to the broader field of network‐based methods in engineering and yield insights that help focus design and maintenance efforts in the future.