Trailing-edge serrations: improving theoretical noise reduction models

This paper discusses the elements that make up a theoretical model for predicting the noise generated by a serrated trailing edge, and in particular, the required input to such a model; the wavenumber frequency spectra of the turbulence at the edge. It is proposed that this input, which is often modeled empirically, varies between a straight edge and a serrated edge and this fundamental difference in turbulence structure at the trailing edge leads to inaccurate theoretical predictions of noise reduction. Experimental measurements are therefore taken for straight and serrated trailing edges, with a particular focus on measuring the quantities which arise in typical wavenumber-frequency spectra models such as the TNO model. Whilst elements like the boundary layer thickness, shear profile, and skin friction velocity are found to be similar across the straight and serrated edges, the normal velocity wavenumber spectra is observed to vary substantially between the straight edge, and different locations along the serrated edge. In particular, the high-frequency decay rate of this spectra is reduced for a serrated edge versus a straight edge. This observed difference is then incorporated into a simple Chase-style empirical wavenumber-frequency spectrum; theoretical predictions which take account of a weaker high-frequency decay rate for a serrated edge agree much better with the experimental far-field noise predictions than theoretical predictions which assume an identical turbulent structure for both straight and serrated edges.