Extracting Flow Characteristics from Single and Multi-Point Time Series Through Correlation Analysis

Strongly driven fluid and combustion systems typically contain a few, nonlinearly coupled, major flow constituents. It is necessary to identify the flow constituents in order to establish the underlying dynamics and to control these complex flows. Due to non-trivial boundary condition in realistic systems and long-range coupling, it is often difficult to construct accurate models of large-scale reacting systems. The question then arises if these flow constituents can be identified and controlled through analysis of experimental data. The difficulties in such analyses originate in the presence of high levels of noise and irregularities in the flow. A typical time series contains high-frequency noise as well as low-frequency features originating from the near translational invariance of the underlying fluid systems. We propose a pair of approaches to study such data. The first is the use of auto and cross correlation functions. Auto-correlation functions of the time series from a single transducer can be used effectively to demonstrate the low dimensionality of the flow. Second, we show that multi-point time series from appropriately placed transducers can be used to establish spatial characteristics of these flow constituents. The novelty of the approaches lies in the establishment of geometric and dynamic features of the primary flow constituents based on sensor data only, without the need of expensive imaging tools. These methods can potentially identify changes in flow behavior within complex propulsion systems, such as aircraft engines, by utilizing data collected from embedded transducers.