885095977

Contributed Papers

11:15

2UW6. Converting bottom lass measured from a rough layered sediment to the equivalent ⁴*flat bottom** loss. Diana F. McCammon (Appl. Res. Lab., Penn State Univ., State College. PA 16804)

In the BLUG parameter-estimation technique, measured bottom loss is inverted to obtain a “besl-fit” set of ten geoacoustic parameters that characterize the sediment. This inversion process has been im-proved and automated with Monte Carlo methods; however, in spite of these advances, the inversion process can słill give poor results, notably in rough thin sediment regions, becausc the inversion model assumes flat interfaces between water, sediment, and basement. The purpose of this paper is to describe a correction factor that can be applied to the measured data to convert it from rough surface loss to the equivalent loss if the interfaces had been smooth. With this correction, the data can be madę to conform to the assumptions of the model, which should lead the inversion process to a better filting set of parameters. Four examples of this application are shown; in two thin sediment cases, the correction gave improved modcl/data correlations and lower squared errors; in the two smoother thicker sediment cases, the correction did not signifi-cantly affect the rcsult. [Work supported by NAVOCEANO.]

11:30

2UW7. Time domain invcrsion of two-dimensional velocity ftelds using simulated annealing. P. Gcrstoft, J. M. Pedersen, and P. D. Vestergaard³⁾ (Odegaard & Danneskiold—Samsde ApS, Kroghsgade I, Copcnhagen 2100, Denmark)

Inversion of scattercd sound fields causcd by horizontal and lateral velocity variations can be done using an eflkieni Monte Carlo method called simulated annealing. The subsurfacc contains large velocity vari-ations in both depth and rangę and thus, for inversion of iransicnt signals, at leasł a twodimensional (2-D) representation of the velocity field is required. This 2-D description requires that the nonlinear inver-sion is carried out in a huge parameter spacc. Standard local optimiza-tion methods will be trapped in local minima and a search throughout is computationally prohibitive. Thus the simulated annealing method is used. The present implementation is here based on an ensemble ap-proach whereby several copies are annealed simultaneously. By using several copies it is possible to obtain statistical information about the optimal cooling ratę. In order to make the method converge in accept-able time, both optimization and the forward modeling method of the inversion have to be fast. Thus gcometrically flexible but computationally cxhaustive methods such as finitc difference are not yet used. At present, the forward modeling is done by either the one-dimensional convo!ution model or ray tracing. The 2-D effect of the one-dimensional convolution is obtained by geometrically requiring the structure to be stratified with a weak variation in rangę. For the ray tracing, the sub-structure shall also be stratified but here the wave propagation is in a real 2-D environment. ^a>Pre$ently at BIRPS, Cambridge, U.K.

11:45

2UW8. Linearization of the matched-field processing approach to acoustic tomography. A. Tolstoy (Acoust. Div., Naval Res. Lab., Washington, DC 20375)

This paper conlinues the approach presented in Tolstoy et aL [J. Acoust. Soc. Am. 89, 1119-1127 (1991)] but offers a much improved inversion technique, i.e., a linearization of the problem, which reformats the computations in terms of simple nonsquarc matrix inversion for an overdetermined system. This linearization results in sound-speed accu-racies that arc an order of magnitude better than the earlier technique. In addition, calculations confirm that for simulations with white, Gaus-sian, uncorrelated noise, the linear/Bartlett processor results are iden-tical to those of the minimum-variance/Capon processor. Fmally, optimal source-receiver configurations have been determincd by exhaustively computing the condition numbers for the associated ma-trices in the new linear formulation. Simulation results now show that three arrays located at optimal coordinates in a 250- by 250-km ocean region with shot sources distributed around the perimeter can result in 3-D sound-speed profiles determined to accuracies better than 0l07 m/s and better than 0.03 m/s for four arrays located at optimal coordinates. Such results presently assume perfect knowledge of sound-speed profiles at the arrays and around the region perimeter.

12:00

2UW9. Matched-field processing (MFP) tomography for inyerting the El Nino profile. E. C. Shang and Y. Y. Wang (CIRES, Univ. of Colorado/NOAA/Wave Propagation Lab., Boulder, CO 80303)

The matched-field processing (MFP) method has been substantially applied for source localization studies in reccnt years. It has been dem-onstrated that the MFP is very scnsitive to the mismatch of sound-speed profile (SSP) with a high-resolution MFP estimator. It tums out that the high-resolution MFP is also a potentia! powerful tool for SSP inyerting with a known source-receiver system (A. Tolstoy and O. Dia-chok, J. Acoust. Soc. Am. Suppl. 1 88, SI 17 (1990)]. In this paper, the high-resolution modę matching (HRMM) estimator (E. C. Shang, J. Acoust. Soc. Am. 86, 1960 (1989)] has been used for El Nino profile inverting. By matching a proper set of modal travc! time perturbation, the El Nino profile can be efficiently inverted in a 2-D parameter space based on a simple acoustic model of the 1982-1983 El Nino event. (Work supported by ONR and NOAA.]

12:15

2UW10. Tomographic reconstructłon of stratified fluid flow. Daniel Rouseff (The Johns Hopkins Univ. Appl. Physics Lab., Johns Hopkins Rd., Laurel. MD 20723), Kraig B. Winters, and Peter Kaczkowski (Univ. of Washington, Seattlc, WA 98105)

A new method for imaging a moving fluid using acoustic tomography is evaluated by numerical simulation. A cross section of the medium is probed by high-frequency acoustic waves from several different directions. It can be shown that the resulting measured travel time data contain sufficient information to reconstruct both the spatially varying sound speed and the transverse component of the fluid vorticity (K. B. Winters and D. Rouseff, Inverse Problems 6, L33 (1990)]. The results are exact to within the validity of the straight-ray geometrie acoustics approximation. To evaluate a discrele version of the reconstruction al-gorithm, a three-dimensional stably stratified mixing layer is simulated. The flow exhibits characteristic features in both density (sound speed) and vorticity. The dynamics of the fluid flow can be described as the instability of a vortex sheet. The acoustic travel time is calculated by integrating through the simulated flow fields. The synthetic data are then inverted to yield reconstructions of both the density and the vor-ticity of the evolving flow. (Work supported in part by the U.S. Navy under Conlract No. N00039-89-C-0001.]

1875

J. Acoust. Soc. Am., Vo1. 89, No. 4. Pt. 2, April 1991

121st Meeting: Acoustical Society of America

1875