정합장처리에서 오정합에 의한 바이어스와 민감도

Abstract

Matched field processing(MFP) is a parameter estimation technique for localizing the range, depth, and bearing of a point source from the signal field propagating in an acoustic waveguide. MFP involves the correlation of the actual acoustic pressure field measured at a receiver array with a predicted field based on a postulated source position and an assumed ocean model.

A high degree of correlation between the measured field and the predicted field indicates a likely source location. Thus an increased complexity of the ocean's structure provides a greater variability of the acoustic fields, which aids the estimation procedure. When the environmental data are inaccurate or incomplete, a "mismatch" occurs between the measured data and the predicted pressure field, that causes a degradation in MFP correlation and an appreciable bias.

In this thesis, I was concerned with quantitative evaluation of the effects of mismatches arising from inaccuracies in a number of important system and ocean environmental parameters in a shallow water. The motivation for this study is to examine the biases in the source localization and the sensitivities of the matching results from various mismatches.

Using a conventional estimator, I have investigated the bias of range and depth estimates caused by perturbations in array position, as well as ocean environmental parameters through the simulation. Replica fields are calculated using the normal mode methods with the exception of bathymetry case. Also this study examined the sensitivity of MFP to geometric, geoacoustic, and ocean sound speed parameters using the genetic algorithm. And this method is applied to measured data to overcome mismatch and accurately estimate source location with limited a priori environmental information by expanding the parameter search space of MFP to include environmental parameters.

As a result, significant biases can be introduced into the depth and range localization predictions of a MFP through erroneous estimates of environmental parameters. It can also be concluded that the impact of mismatch, both summer sound speed and sensor position in water layer, is more serious than the geoacoustic parameters. This implies that simulations of mismatch which consider only a few errors will provide very misleading results on source position. Water depth and bottom bathymetry errors can be offset significantly

it shifted progressively farther away and deeper from the actual source location as the true water depth became shallower. Errors in estimates of the sediment attenuation and density, and basement parameters appear to be of relatively minor importance.

From an experimental implementation viewpoint, these result should enable resources to be concentrated on obtaining reliable values for those parameters which are important to know accurately, avoiding unnecessary effort to overdetermine relatively unimportant ones. It is also necessary to understand the types of mismatches in MFP that may be introduced by inaccuracies in the various forward modeling parameters, so that specific types of information deficiencies may be identified and attempts can be made to compensate for them.