【作者】 邹士新；

【导师】 马远良；

【作者基本信息】 西北工业大学 ， 武器系统与运用工程， 2006， 博士

【摘要】 本文以浅海底质参数的匹配场反演和失配环境下水面噪声源的匹配场干扰抑制技术为主要的应用背景，从浅海声场的抛物方程建模分析、匹配场反演优化算法的构造、匹配场反演结果的不确定性评价算法的构造、失配环境下的匹配场干扰抑制算法的设计等方面进行了深入的研究。主要研究内容和取得成果有： (1) 依据广义邻域优化算法构造理论设计了匹配场反演所必须的全局优化算法。对现存优化算法的优缺点进行了对比分析，利用匹配场反演标准测试问题选择了构成混合优化算法的基本算法，并根据匹配场反演问题的特点确立了混合优化算法的结构和混合原则，提出了衡量算法性能的指标。最后利用标准测试问题对新、旧算法的性能进行了对比分析，表明混合寻优算法是高效的。 (2) 应用贝叶斯分析与推断理论，研究了匹配场反演结果的不确定性问题。建立了完全贝叶斯分析与经验贝叶斯分析的数学模型，并讨论了与不同模型相关的计算复杂性问题。用标准测试问题，分别就小维度参数反演问题和高维度参数反演问题进行了详细的定性与定量的分析。对于参数数量少的小维度参数反演，重点分析了数值计算模型的精度。利用所导出的解析解，从反演频率的高低、使用频率的数量多少、信噪比的变化等方面检验了数值模型的精确性，结果表明数值模型能完全满足匹配场反演结果不确定性分析的要求。在高维参数反演的场合，已有的分析表明，计算复杂度与反演参数数量呈指数关系。应用上述数值分析模型和高等统计模拟计算方法，导出了一种基于全局优化算法的不确定性量化算法，在计算精度与计算时间上达到了较好的平衡。 (3) 进行了匹配场反演方法的试验数据验证。运用1993年SACLANT中心ELBA岛浅水试验数据，对本文提出的算法进行了全面的检验与验证。利用所获得的数据，对不同频率范围、不同时段的数据分别进行了反演。对各种反演情况下的反演结果进行了对比分析，并对反演结果进行了解释。结果表明本文提出的算法能够应用于实际。 (4) 进行了稳健匹配场噪声抑制算法研究。对常规空域滤波器对环境参数失配敏感的原因进行了分析，提出使用扰动环境参数，结合具体的搜索位置，利用离散Karhunen-Loeve展开对获得的拷贝场向量进行处理，以获得新的对环境参数失配稳健的特征向量。同时基于稳健空域滤波器设计过程中优化变量多，直接优化所得的空域滤波器性能较差的事实，提出使用离散余弦变换来降低空域滤波器的维数，减少优化变量数，提高空域滤波器性能。更多还原

【Abstract】 This dissertation studies mainly the inversion of ocean environmental parameters in shallow water and the suppression of sea-surface noise interferences in mismatch environment by Matched Field Processing (MFP). Acoustic field modeling using Parabolic Equation method is analyzed, a global optimization algorithm is developed, the uncertainty analysis of matched field inversion is conducted, and the surface interference suppression algorithm in mismatch environment is obtained. The main contributions are as follows:(1) The global optimization algorithm applied matched field inversion is developed based on generalized neighboring region theory. The optimization algorithms presently applied to matched field inversion are compared. In consideration of a Benchmark problem, two basal algorithms, i.e. Differential Evolution algorithm and Down Hill Simplex algorithm, are selected and a hybrid optimization algorithm based on the basal algorithms is proposed. The structure and performance parameters of the hybrid optimization algorithm are deduced, and a flow chart of the algorithm based on generalized neighboring region theory is shown. Using Matched field inversion Benchmark problem a comparison has been made between the hybrid optimization algorithm, it is called Down Hill Simplex Differential Evolution algorithm, and the basal algorithms. The results indicate that hybrid optimization algorithm is more effective.(2) The uncertainty problem of the matched inversion is investigated using Bayesian statistics and decision theory. The Full Bayesian approach and Empirical Bayesian approach are introduced and the computational complexity is discussed. Qualitative and quantitative analysis for the inversion is performed in low and high dimensional parameter space using the benchmark problem. For the low dimensional case the Full Bayesian approach is adopted for validating the computational model using different signal to noise rate and inversion frequency. For the high dimensional case the computational complexity problem is considered and an algorithm which calculates the inversion uncertainty is developed based on the optimization algorithm.(3) Using experimental data in the Mediterranean Sea the developed algorithms are validated. According to the ocean environment, range dependent inversion is conducted and the results are analyzed.更多还原