Optimizations for Fourier Synthesized Time Domain Pulse Propagation Calculations

摘要

Fourier transform methods are the standard way for determining time- domain pulse structure and arrival time from a set of continuous wave (discrete frequency) underwater acoustic model calculations. This technique requires a large number of computer model runs at closely spaced frequencies, often making it computationally expensive. It has the advantages of including the correct attenuation at each frequency component, and of correctly treating continuity requirements at the water/sediment interface. Direct time-domain computer models are not as accurate for ocean bottoms with strong attenuation over a large bandwidth of frequencies. In this work the frequency-domain/Fourier approach is optimized for maximum efficiency at a given level of acceptable imprecision. Techniques are presented to improve the efficiency of the individual frequency component calculations, and to avoid running many of the frequencies. Efficiencies at individual frequencies are gained through intelligent selection of grid parameters in the ocean acoustic model (a parabolic equation model). Further improvements are achieved through intelligent zero padding schemes, and by interpolating envelope functions at the receiver location in order to estimate (and hence avoid running) up to 90% of the calculations required by the Nyquist sampling theorem. The effects of the various approximations are shown in the examples.