Bayesian inference for modal identification in ducts with a shear flow.

Abstract

An in-duct modal decomposition technique is described. The basis for the technique is to consider the decomposition as an inference problem. Using transfer function measurements at the duct walls, a Bayesian inference is conducted to evaluate the acoustic modal coefficients in the presence of uncertainties. These uncertainties encompass model errors, microphone measurements error, and uncertainty on the flow profile. The formalism of the direct problem of modal decomposition in a ducted shear flow is first developed. The case of a circular cross-section duct is then treated without and with a flow, using synthetic noisy signals for the inference problem.