The main objective of this work is the use of inverse techniques for source identification in aero-acoustic problems, and the reconstruction of noise source distributions from pressure or velocity field measurements. The investigation is performed on a laboratory-based flow duct acoustic facility installed in the anechoic chamber of the UTC. A theoretical analysis of inverse methods for reconstructing ducted noise sources in the configuration of the testing platform has been developed. A set of experimental measurements has been used for a practical estimation of the capability of the different source reconstruction strategies in laboratory conditions. The inverse method for deducing the optimal source strengths lead to the inversion of the matrix of the frequency response functions between the set of assumed sources and the measurement array. Unfortunately, in practice the problem becomes very often bad –conditioned due to the effects of measurement noise or modelling errors. But even when the response matrix to be inverted becomes illconditioned, useful results can still be found by using regularisation methods. Traditionally, two regularisation methods, Tikhonov regularisation and singular value discarding (SVD) have been employed to find a good estimate. The choice of the regularisation parameters involved within each method is essential for the successful source strength reconstruction. The L-curve method and the ordinary cross-validation (OCV) and generalised cross validation (GCV) constitute useful analysis tool for the proper parameter selection. A comparison between the different regularisation techniques has been performed as a function of the three main parameters that influence the reconstruction results: the axial distance between the sensor array and the reconstruction sources, the frequency range of interest and the presence of noise.
|
![]() |
Diagram of the set-up considered and preliminary comparison between the measured pressure and velocity inside the duct with the corresponding reconstructed fields by inverse techniques |