An Information-Theoretic Exploration of Multi-Body Fluid-Structure Interactions

Abstract

Fluid-structure interactions are highly complex and difficult to resolve using classical physics tools. Recently, information theory has been proposed as an alternative for exploring these often highly non-linear and complex multi-physics problems. By viewing the fluid-structure system as a flow of information, valuable knowledge can be gained without needing to resolve the details of the interaction. Here, a combination of simulations and experiments are used to assess how the behaviour of the fluid-structure system can be recast into an information-theoretic framework. The proposed information-theoretic tool is transfer entropy, which takes time series of some aspect of the system behaviour, for example structural displacements, and infers directed casual relationships between the components. Furthermore, due to dependence on the lag parameter, transfer entropy can infer information about the critical time scales of the system. The first experiments consist of two structures, constrained to allow only unidirectional communication and designed to have enough known properties to validate the information-theoretic analysis. Additional experiments are performed that add a third structure, thereby increasing the number of communication pathways. This work shows that, while transfer entropy is model free, the order of operations and interpretation of the results are significantly impacted by the characteristics of the input data. As a result, it is imperative to have a clear research goal and an understanding of the critical behaviours of the system. With that, it is relatively inexpensive to gain valuable information from even the most complex fluid-structure systems. Overall, transfer entropy is proven to be a useful tool for the analysis of fluid-structure interactions, provided it is not naively applied and the data exhibits some degree of randomness.