Underwater detection, classification and localisation

Improving the capabilities of towed sonar arrays

More Info


The end of the Cold War and the collapse of the Warsaw pact have resulted in a change of operational theatre for the naval forces of the North Atlantic Treaty Organisation (NATO). In particular, the focus of Anti Submarine Warfare forces has shifted from tracking Soviet nuclear ballistic missile submarine in the deep waters of the Atlantic ocean to hunting smaller and quieter Diesel electric submarines in coastal water. In most scenarios, towed array sonars are the best sensor to detect, classify and localise submarines. The long passive towed array sonars used during the Cold war are more difficult to use in coastal waters and are being replaced by most Navies by Low Frequency Active Sonars (LFAS) using a towed source and shorter towed receiving array. These shorter towed arrays can be used in both active and passive modes. In passive mode, their reduced size offer limited performance compared to their longer equivalent. In active mode, they can detect submarines at long ranges in shallow waters but are plagued by false alarms caused by echoes from features of the seafloor. This thesis deals with algorithms improving Detection, Classification and Localisation for towed sonar arrays, with a specific focus on LFAS sonars. In Chapter 2, we derive, analyse and apply to measured data a method for improving detection performance with short passive towed arrays. An important issue in detection of quiet acoustic source with short towed arrays is the improvement in signal-to-noise ratio (SNR) and bearing resolution for targets emitting low frequency signals. One of the techniques believed to improve these characteristics is Synthetic Aperture Sonar (SAS). The method is based on the artificial enlargement of a sonar array by coherently integrating acoustic snapshots at different antenna positions. We first derive theoretical measures of performance of passive SAS and report on its application in combination with other signal-processing algorithms. Its theoretical performance is compared with that of the frequently used incoherent integration. The used passive SAS algorithm is the method known as Extended Towed Array Measurement (ETAM) or the overlap correlator. It is based on the correlation of data snapshots on overlapping hydrophones. Correlation is a key issue in this method and since it is affected by noisy targets, some gain can be expected from noise cancellation. The influence on the performance of ETAM of a method of tow ship noise cancelling at hydrophone level (Inverse Beam Forming, IBF) is analysed. This approach increases ETAM performance by removing a loud and highly correlated noise source, the tow ship, and thus enhancing the other targets in the beam pattern. The results of the algorithms applied to two experimental datasets show that they bring an improvement close to theoretical expectations. Port starboard discrimination and the successful combination of IBF with ETAM make this approach innovative. In Chapter 3, methods for improving the localisation of a source with a short towed array are analysed and applied to data, both simulated and measured at sea. Localisation performance with sonar towed array is related to the array length. The knowledge of the position of a given acoustic source gives a critical tactical advantage to a ship. There are a limited number of ways to estimate the range of a source with a towed passive sonar, most requiring the towing platform to execute a manoeuvre. These manoeuvres are undesirable as they take a lot of time, cause bending of the towed array and can even put the towing platform in harm’s way. We present a number of source position estimation methods for both broadband and narrowband sources suitable for short towed arrays. Recursive methods based on the extended Kalman filter are first examined. A new method based on the integration of time delay of arrival measurements along the sonar path is described. We derive theoretical performance indicators and show that this method gives the possibility to estimate the position and speed of a source without a manoeuvre. In Chapter 4, the Classification performance of a broadband waveform is analysed and measured on data collected at sea. Like any long-range active sonar system, LFAS produces a large amount of unwanted sea bottom echoes or clutter. These echoes give rise to false alarms that increase the computational load of target trackers and jeopardise the correct classification of each echo. The number of false alarms due to clutter can be reduced either through echo classification techniques or through Doppler filtering provided the targets of interest are in motion. Much research has been carried out on waveform investigation for the efficient use of bandwidth capabilities of modern sonar transmitters. Among the quantity of waveforms, Binary Phase Shift Keyed (BPSK) pulses have emerged as exhibiting cross-correlation properties relevant to Doppler filtering while maintaining a range resolution comparable to Frequency Modulated (FM) pulses. We have successfully applied a false alarm reduction technique using contacts obtained with an FM pulse subsequently processed by Doppler filtering with a BPSK pulse. The Doppler classification performance for this pulse is evaluated on an experimental dataset and a few limitations of BPSK are identified.