The measurement of underwater acoustic noise radiated by a vessel using the vessel's own towed array

Abstract

The work described in this thesis tested the feasibility of using a towed array of hydrophones to: 1. localise sources of underwater acoustic noise radiated by the towvessel, 2. determine the absolute amplitudes of these sources, and 3. determine the resulting far-field acoustic signature of the tow-vessel. The concept was for the towvessel to carry out a U-turn manoeuvre so as to bring the acoustic section of the array into a location suitable for beamforming along the length of the tow-vessel. All three of the above were shown to be feasible using both simulated and field data, although no independent field measurements were available to fully evaluate the accuracy of the far-field acoustic signature determinations. A computer program was written to simulate the acoustic signals received by moving hydrophones. This program had the ability to model a variety of acoustic sources and to deal with realistic acoustic propagation conditions, including shallow water propagation with significant bottom interactions. The latter was accomplished using both ray and wave methods and it was found that, for simple fluid half-space seabeds, a modified ray method gave results that were virtually identical to those obtained with a full wave method, even at very low frequencies, and with a substantial saving in execution time. A field experiment was carried out during which a tug towing a 60-hydrophone array carried out a series of U-turn manoeuvres. The signals received by the array included noise radiated by the tow-vessel, signals from acoustic tracking beacons mounted on the tow-vessel, and transient signals generated by imploding sources deployed from a second vessel.Algorithms were developed to obtain snapshots of the vertical plane and horizontal plane shapes of the array from the transient data and to use range data derived from the tracking beacon signals to track the hydrophones in the horizontal plane. The latter was complicated by a high proportion of dropouts and outliers in the range data caused by the directionality of the hydrophones at the high frequencies emitted by the beacons. Despite this, excellent tracking performance was obtained. Matched field inversion was used to determine the vertical plane array shapes at times when no transient signals were available, and to provide information about the geoacoustic properties of the seabed. There was very good agreement between the inversion results and array shapes determined using transient signals. During trial manoeuvres the array was moving rapidly relative to the vessel and changing shape. A number of different array-processing algorithms were developed to provide source localisation and amplitude estimates in this situation: a timedomain beamformer; two frequency-domain, data independent beamformers; an adaptive frequency-domain beamformer; and an array processor based on a regularised least-squares inversion. The relative performance of each of these algorithms was assessed using simulated and field data. Data from three different manoeuvres were processed and in each case a calibrated source was localised to within 1 m of its known position at the source's fundamental frequency of 112 Hz.Localisation was also successful in most instances at 336 Hz, 560 Hz and 784 Hz, although with somewhat reduced accuracy due to lower signal to noise ratios. Localisation results for vessel noise sources were also consistent with the positions of the corresponding items of machinery. The estimated levels of the calibrated source obtained during the three manoeuvres were all within 4.1 dB of the calibrated value, and varied by only 1.3 dB between manoeuvres. Results at the higher frequencies had larger errors, with a maximum variation of 3.8 dB between serials, and a maximum deviation from the calibrated value of 6.8 dB. An algorithm was also developed to predict the far-field signature of the tow-vessel from the measured data and results were produced. This algorithm performed well with simulated data but no independent measurements were available to compare with the field results.