Motivated by recent success in the dynamical systems approach to transitional flow, we study the efficiency and effectiveness of extracting simple invariant sets (recurrent flows) directly from chaotic/turbulent flows and the potential of these sets for providing predictions of certain statistics of the flow. Two-dimensional Kolmogorov flow (the 2D Navier-Stokes equations with a sinusoidal body force) is studied both over a square [0, 2π]2 torus and a rectangular torus extended in the forcing direction. In the former case, an order of magnitude more recurrent flows are found than previously (Chandler & Kerswell 2013) and shown to give improved predictions for the dissipation and energy pdfs of the chaos via periodic orbit theory. Over the extended torus at low forcing amplitudes, some extracted states mimick the statistics of the spatially-localised chaos present surprisingly well recalling the striking finding of Kawahara & Kida (2001) in low-Reynolds-number plane Couette flow. At higher forcing amplitudes, however, success is limited highlighting the increased dimensionality of the chaos and the need for larger data sets. Algorithmic developments to improve the extraction procedure are discussed.
Adaptive Diffusion Posterior Sampling for Data and Model Fusion of Complex Nonlinear Dynamical Systems
College of Information Sciences and Technology, Pennsylvania State University(宾夕法尼亚州立大学信息科学与技术学院);School of Mechanical Engineering, Purdue University(普渡大学机械工程学院)
Leveraging Scale Separation and Stochastic Closure for Data-Driven Prediction of Chaotic Dynamics
Sorbonne Université(索邦大学);Institut Jean le Rond d’Alembert(让·勒朗·达朗贝尔研究所);ISIR(智能机器人研究所);SCAI(认知科学与人工智能中心);Institut Universitaire de France(法国高等研究院);M2N, Conservatoire National des Arts et Métiers(巴黎国立工艺学院)