Speaker: Dr. Iman Khatami,
PhD, Postdoctoral fellow,
Universitte de Sherbrooke
Title: FREE-FIELD INLET / OUTLET NOISE IDENTIFICATION ON AIRCRAFT ENGINES USING MICROPHONES ARRAY
Local Organizer: Dr. Akbarzade Tootoonchi
Time: Saturday, January 10, 2015
Location: Amphitheatre Hall, Department of Mechanical Engineering, Ferdowsi University of Mashhad (FUM)
Abstract:
This research considers the discrimination of inlet / exhaust noise of aero-engines in free-field static tests using far-field microphone arrays. Various techniques are compared for this problem, including classical beamforming (CB), regularized inverse method (Tikhonov regularization), L1- generalized inverse beamforming (L1-GIB), clean-PSF, clean-SC and two novel methods which are called hybrid method and clean-hybrid. The classical beamforming method is disadvantaged due to its need for a high number of measurement microphones in accordance with the requirements. Similarly, the inverse method is disadvantaged due to their need of having a priori source information. The classical Tikhonov regularization provides improvements in solution stability, however continues to be disadvantaged due to its requirement of imposing a stronger penalty for undetected source positions. Coherent and incoherent sources are resolved by L1- generalized inverse beamforming (L1-GIB). This algorithm can distinguish the multipole sources as well as the monopoles sources. However, source identification by L1- generalized inverse beamforming takes much time and requires a PC with high memory. The hybrid method is a new regularization method which involves the use of an a priori beamforming measurement to define a data-dependent discrete smoothing norm for the regularization of the inverse problem. Compared to the classical beamforming and the inverse modeling, the hybrid (beamforming regularization) approach provides improved source strength maps without substantial added complexity. Although the hybrid method rather solves the disadvantage of the former methods, the application of this method for identification of weaker sources in the presence of the strong sources isn’t satisfactory. This can be explained by the large penalization being applied to the weaker source in the hybrid method, which results in underestimation of source strength for this source. To overcome this defect, the clean-SC method and the proposed clean-hybrid method, which is a combination of the hybrid method and the clean-SC, are applied. These methods remove the effect of the strong sources in source power maps to identify the weaker sources. The proposed methods which represent the main contribution of this thesis show promising results and opens new research avenues. Theoretical study of all approaches is performed for various sources and configurations of array. In order to validate the theoretical study, several laboratory experiments are conducted at Université de Sherbrooke. The proposed methods have further been applied to the measured noise data from a Pratt & Whitney Canada turbo-fan engine and have been observed to provide better spatial resolution and solution robustness with a limited number of measurement microphones compared to the existing methods.