Pressure-Informed Velocity Estimation in a Subsonic Jet

TL;DR

This study develops and compares pressure-informed velocity estimation methods in a subsonic jet, revealing pressure fluctuations' connection to flow structures and highlighting the bidirectional LSTM's effectiveness in capturing flow dynamics.

Contribution

Introduces hybrid spectral and POD-based linear models and neural network approaches, including bidirectional LSTM, for pressure-velocity estimation in subsonic jets, advancing flow and noise analysis.

Findings

Pressure fluctuations relate to large-scale flow structures.

Bidirectional LSTM outperforms other estimation methods.

The models capture space-time dynamics of acoustic sources.

Abstract

This work aims to estimate time-resolved velocity field that is directly associated with pressure fluctuations in a subsonic round jet. To achieve this goal, synchronous measurements of the velocity field and in-flow pressure fluctuations were performed at Mach number 0.3. Two different experiment campaigns were conducted, the first experimental campaign aims to explore the time-resolved dynamics of the axisymmetric velocity components, and second experiment focuses on the time-resolved, 2D velocity estimates on a streamwise plane. Two different methods were utilized to estimate the input-output relation between velocity and in-flow pressure measurements. A hybrid approach based on the spectral linear stochastic estimation and the proper orthogonal decomposition was applied to setup the model in a linear manner, and a wavelet-based filter was implemented to attenuate the noise level in the cross-correlation functions. In addition, the pressure-velocity relationship was also described by neural network architectures based on the multi-layer perceptron (MLP) and bidirectional long-short-term-memory (LSTM). In both experimental sets, pressure fluctuations inside the flow are found to be connected to the streamwise convection of large-scale coherent structures in the flow. A unique advantage of the bidirectional LSTM method was found among all estimation schemes is also reported in this work. The estimation result represents the space-time dynamics of the acoustic sources in the jet flow field, and it is of great importance to understand the noise generation mechanism.