A new analytical theory for multiple scattering of cylindrical acoustic waves by an array of finite impedance semi-cylinders embedded in a smooth acoustically hard surface is derived by extending previous results for plane waves [Linton and Martin, J. Acoust. Soc. Am. 117 (6) 3413 -3423 (2005)]. Although the computational demands of the new theory increase as the number of the semi-cylinders in the arrays and/or the frequency increases, the theory offers an improvement on analytical boss theories since the latter (i) are restricted to non-deterministic (infinite) random distributions of semi-cylinders with spacing/radii small compared to the incident wavelength and (ii) are derived only for plane waves. The influence on prediction accuracy of truncation of the infinite system of equations introduced by the new theory is explored empirically. Laboratory measurements have been made over deterministic random arrays of identical varnished wooden semi-cylinders on a glass plate. The agreement between predictions and measured relative Sound Pressure Level spectra is very good both for single deterministic random distributions and for averages representing non-deterministic random distributions. The analytical theory is found to give identical results to a Boundary Element calculation but is much faster to compute.PACS numbers: 43.50.Vt, 43.28.En
IntroductionSurface roughness is known to have significant influence on near-grazing sound. One approach to modeling long wavelength sound reflection from randomly rough surfaces considers scattering from idealized roughness elements or 'bosses'. Several measurements have been made of relative sound pressure level (SPL) spectra above rough surfaces, where the roughness height and spacing are small compared to the wavelengths of interest [1][2][3][4][5]. These data have been compared with predictions of models derived by Attenborough and Taherzadeh [1] from a boss theory by Tolstoy [6], [7]. It has been found necessary to adjust the impedance of the scatterers and imbedding plane to obtain good agreement between predictions based on Tolstoy's boss theory and the data. Tolstoy's effective admittance models [6], [7] predict that a surface wave is generated at grazingincidence above a hard rough boundary and that the effective admittance above a hard rough boundary is purely imaginary. However, comparison with data [2] indicates that Tolstoy-based predictions overestimate the surface wave component, especially at grazing incidence, and that it is necessary to include attenuation due to non-specular scatter to obtain a good fit with these data [4]. In other comparisons of predictions and data [5], the assumed location of the effective admittance plane has been adjusted to improve agreement with data at higher frequencies. Poor agreement between laboratory measurements of propagation over rough convex surfaces and