Research On High-Resolution Localization Method For Underwater Near-field Noise Sources

The self-noise will make a great impact on stealth for naval vessels.Self-noise not only provides detection and localization information for the enemy's sonar,but also cause interference on underwater acoustic equipments used in the own vessel.In order to improve combat capability of vessels at sea,we must take effective measures to reduce vibration and noise.Vessels are large in size,and radiated noise sources are in different parts.So the research on underwater noise source localization technology is developed to obtain the spatial distribution of noise sources on vessel,which can be targeted to reduce vibration and noise.Compared to far-field source localization,near-field localization can obtain more accurate positions of noise sources.However,the traditional near-field source localization algorithms have low resolving power,the space aliasing occurs easily,which have an impact on the positioning accuracy.Based on linear array,the dissertation focuses on underwater near-field noise source high-resolution localization methods.The main research contents include:1.The line spectrum is a component part of ship radiated noise.Considering the line spectrum localization,an algorithm based on sparse signal reconstruction is presented.The algorithm constructs the optimization problem by minimizing the L1-norm,which solves the problem of source localizaion.Moreover,the algorithm can automatically select a tradeoff parameter by estimating the noise power,without the prior knowledge on environment noise.And the tradeoff satisfies the balance of sparsity and fidelity.For the near-field localization,the delay inequality between the source and each array element is a function of azimuth and range.If the two-dimensional space domain is divided,the high computational cemplexity is almost unacceptable.To solve the problem above,we utilize the combination of symmetric element data to construct a new covariance matrix which contains only the azimuth parameter.Thereby we transform the two-dimensional positioning problem into two one-dimensional parameter estimations,then the computational complexity of near-field source localization is reduced.The results of computer simulation and pool experiment show the effectiveness of the proposed algorithm.2.Compared to the far-field source localization,many near-field source localization algorithms transform the two-dimensional search into multiple one-dimensional searchs to reduce the quantity of calculation.However,two conditions must be guaranteed to avoid spatial aliasing.One condition is that the element spacing is less than quarter-wavelength,the other one is that the number of estimated sources is not more than half of element number.When the element number is constant,there is a problem of aperture loss.In fact,the resolving power and positioning accuracy depends on the array aperture size.Considering the aperture loss problem appearing on near-field localization,the co-prime symmetric array model is presented.The element spacing is increasing and need not limit to a quarter wavelength by adopting co-prime symmetric array,which makes array aperture extend effectively.Under the co-prime symmetric array model,we reconstruct a special four-order cumulant matrix to estimate the azimuths of near-field sources,and then the ranges of different sources can be estimated by searching MUSIC spectral peak in these directions of each estimated azimuth.The method need not additional paremeter pairing algorithm.In addition,utilizing the method can avoid the spatial aliasing.Theoretical analysis and computer simulation verify the proposed method can improve resolving power and positioning accuracy effectively.3.Considering the broadband continuous spectrum component part localization,we research near-field boradband source localization methods sufficiently,and present a positioning algorithm based the on subspace orthogonality principle.This algorithm estimates of the positions of different near-field noise sources by measuring the orthogonal relation between the signal subspaces of each frequency component and the noise subspace of the reference frequency component,and decreases the impact on noise subspace estimation error by the orthogonal projection of the array steering vector at the reference.The algorithm improves the resolving power and positioning accuracy in low SNR condition.The algorithm is an extension of test of orthogonality(TOPS)in near-field region.Moreover,it takes full advantage of the signal subspace at each frequency to solve the problem that TOPS depend too much on test the signal subspace estimation accuracy at reference frequency.Theoretical analysis and computer simulation show the effectiveness of the proposed algorithm.