基于反向散射强度的海底参数反演方法研究

发布时间：2018-05-05 19:45

本文选题：反向散射强度 + 地声参数反演　；参考：《哈尔滨工程大学》2014年博士论文

【摘要】：海底作为海洋波导声传播的重要边界,可靠的声传播模型和匹配场声源定位都需要已知有关海底的声学特性。然而,大范围的海底参数通常难以直接测量,并且需要付出相当高的代价。目前,利用声场测量数据来估计海底参数的声学反演方法已成为海底遥测的有效手段,并具有广阔的应用前景。海底声散射是声波与海底相互作用的结果,势必携带有关海底的信息,因此本文在构建反向散射强度正演模型的基础上,利用实验室环境条件下测得的水池底部沙质沉积物反向散射强度数据,开展了基于反向散射强度的海底参数反演方法的研究与验证。与传统的基于声传播模型的匹配场反演方法相比,该方法的优势在于算法简单,易于实现,能够获得更为全面的海底信息。在正演模型中,将海底沉积物视为由固体框架和孔隙水组成的双相系统,由基于Biot理论简化得到的等效密度流体模型(EDFM,有时也称为等效密度流体近似)来描述,适用于泥(包括粘土和粉砂)或细沙等软质沉积物,其中等效密度为一个与频率有关的复数;考虑海底表面粗糙散射和沉积物内部不均匀性引起的体积散射,分别采用复合粗糙散射模型和微扰流体近似体积散射模型;考虑沉积物的频散特性,声速频散和衰减的频率依赖关系由反演出的沉积物物理参数结合EDFM计算得到。因此,正演模型由10个模型参数组成,声速和衰减(以沉积物/海水声速比和损失参数形式给出)作为间接反演参数。在高斯数据误差的假设下,本文基于似然函数建立目标函数,用以衡量反向散射强度测量值与模型预报值的失配。一方面,为了求解目标函数的最优解,采用一种基于差分进化算法和粒子群算法的两级混合优化算法,以克服单一算法的缺点,有效提高反演结果的精度。另一方面,为了分析反演参数的不确定性和参数间的相关性,采用Bayesian反演方法,通过后验概率密度给出反演参数的边缘概率分布和协方差矩阵。仿真研究表明,这种间接反演地声参数(包括声速和衰减,以沉积物/海水声速比和损失参数的形式给出)的方法,对地声参数的反演表现出较强的稳健性,特别是损失参数也能够得到令人满意的估计结果。最后,根据实验测得的反向散射强度数据对池底沙质沉积物参数进行了反演,通过与取样测得的沙样品参数的比较来验证反演结果的可靠性。测量的沙样品参数包括:孔隙度、平均颗粒粒度、颗粒质量密度、高频段声速和衰减系数。在对沙质沉积物中声速和衰减系数进行取样测量时(频率范围90kHz-170kHz),提出一种基于脉冲压缩技术的宽带测量方法,通过一次测量即可获得测量频带内沉积物中声速频散和衰减的频率依赖关系。与窄带测量结果的比较可以发现,声速和衰减系数的宽带测量结果存在一定的起伏,特别是衰减系数在测量频带的后半部分与窄带测量结果的偏差较大,这一现象可能是实验中所采用的宽带信号时延分辨率偏低,直达波对应的相关峰与邻近多途有所叠加的结果。相比于仿真结果,根据实验数据得到的反演参数不确定性有所增大,部分参数间的相关性进一步减弱,这是数据误差总体上大于仿真情况的结果。相比于取样测量结果,反演得到的孔隙度偏小,而高频段声速偏大,除数据误差的影响外,在取样、搬运和测量过程中,沉积物的堆积状态难免发生改变,致使实测的沉积物物理参数和地声参数与其实际参数值也会存在一定的差异。此外,还对池底沙质沉积物低频段(0.5kHz-3kHz)声速进行了原位测量,实验结果表明大量气泡的存在会导致声速明显降低。
[Abstract]:As the important boundary of acoustic propagation in marine waveguides, a reliable acoustic propagation model and a matched field sound source location need to be known for the acoustic characteristics of the seabed. However, a large range of seabed parameters is usually difficult to be measured directly, and requires considerable cost. At present, acoustic field measurements are used to estimate the acoustic inverse of the seabed parameters. The method has become an effective means for seafloor telemetry and has a broad application prospect. Underwater acoustic scattering is the result of the interaction of sound waves with the seabed, and it is bound to carry information about the seabed. Therefore, on the basis of building the forward model of backscatter intensity, this paper uses the sand sediment at the bottom of the pool measured in the laboratory environment. The inversion method based on backscatter intensity based on backscatter intensity is studied and verified. Compared with the traditional matching method based on the sound propagation model, the advantage of this method is that the algorithm is simple and easy to be realized and can obtain more comprehensive sea bottom information. In the forward model, the seabed sediments are considered as A dual phase system consisting of a solid frame and pore water, described by an equivalent density fluid model (EDFM, sometimes also called equivalent density fluid) based on Biot theory, suitable for soft sediment, such as clay and silt, or fine sand, in which the equivalent density is a complex number related to the frequency; the surface roughness of the sea floor is considered. The volume scattering caused by the roughness scattering and the inhomogeneity in the sediment is composed of the composite rough scattering model and the perturbation fluid approximate volume scattering model. Considering the dispersion characteristics of the sediments, the frequency dependence of sound velocity dispersion and attenuation is calculated by the EDFM calculation of the physical parameters of the sediment. Therefore, the forward model is composed of 10 modules. Type parameter composition, sound velocity and attenuation (given in the form of sediment / sea sound speed ratio and loss parameter) as indirect inversion parameters. Under the assumption of Gauss data error, the objective function is built on the basis of likelihood function to measure the mismatch between the measured value of backscatter intensity and the model prediction value. On the one hand, to solve the objective function most. A two level hybrid optimization algorithm based on differential evolution and particle swarm optimization is adopted to overcome the shortcomings of the single algorithm and improve the accuracy of the inversion results. On the other hand, in order to analyze the uncertainty of the parameters and the correlation between the parameters, the Bayesian inversion method is used to give the inversion parameter through the posterior probability density. The marginal probability distribution and covariance matrix of the number show that this indirect inversion of ground acoustic parameters (including sound velocity and attenuation, sediment / sea sound speed ratio and loss parameters) shows a strong robustness to the inversion of the acoustic parameters, especially the loss parameters can also be satisfactorily estimated. Finally, the parameters of sediment sediment at the bottom of the pool are retrieved according to the measured backscatter intensity data, and the reliability of the inversion results is verified by comparison with the sample parameters measured by the sample. The parameters of the measured sand sample include porosity, average particle size, particle mass density, sound velocity and attenuation coefficient at high frequency section. The sound velocity and attenuation coefficient in sand sediment are measured (frequency range 90kHz-170kHz). A broadband measurement method based on pulse compression technique is proposed. The frequency dependence of sound velocity dispersion and attenuation in the measured frequency band can be obtained by one time measurement. There is a certain fluctuation in the wideband measurement results of the attenuation coefficient, especially the attenuation coefficient has a larger deviation from the narrowband measurement results in the second half of the measured frequency band. This phenomenon may be the result of the low resolution of the time-delay of the wideband signal used in the experiment and the result that the corresponding peak of the direct wave corresponds to the adjacent multipath. As a result, the uncertainty of the inversion parameters obtained by the experimental data is increased and the correlation between some parameters is further weakened. This is the result that the data error is larger than the simulation. Compared with the results of the sampling measurement, the porosity is smaller and the sound velocity of the high frequency section is large, in addition to the influence of the data error, the sampling and handling are carried out. In the course of the measurement, the accumulation of sediment is unavoidable, and the physical parameters of the sediment and the actual parameters of the sediment are also different from the actual parameters. In addition, the acoustic velocity of the low frequency section (0.5kHz-3kHz) of the sediment in the bottom of the pool is measured in situ. The experimental results show that the existence of a large number of bubbles will lead to sound. The speed is obviously reduced.

【学位授予单位】：哈尔滨工程大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TB56

【相似文献】