返回
类型 基础研究 预答辩日期 2018-01-22
开始(开题)日期 2016-05-26 论文结束日期 2017-11-27
地点 李文正楼北614 论文选题来源 973、863项目     论文字数 7 (万字)
题目 近场高频渐进近似方法
主题词 近场物理光学,线积分,驻相法,近场弹跳射线法,近场绕射
摘要 为了应对未来信息通讯对超大带宽、超高容量的需求,通信设备和民用探测雷达逐渐向大带宽 的微波、毫米波波段拓展,电磁散射和辐射逐渐成为电大、超电大尺寸目标的近场计算问题。传统 的全波方法己不能满足实际工程电磁计算的需要,目前高频渐进方法对解决这些问题具有更大优 势。本文主要研究高频渐进近似方法中的近场物理光学的线积分技术、近场弹跳射线法、近场截断 劈等效边缘电磁流法和基于近场电磁数据的深度学习方法,本文的主要工作和创新点可概括为: ? 基于物理光学,分别推导出频域近区电磁场的双站散射和单站散射线积分表达式。通过构造并矢 势函数并结合变形形式的 Poincaré 引理,将偶极子源照射理想导电平板的双站散射物理光学面积 分表达式退化为无奇异性的围线积分形式;使用全局坐标下的矢量代数定理推导出矢量形式的近 场单站线积分表达式,避免了现有标量线积分表达式中 Dirichlet 函数在数值实现中的奇异性问题。 与传统物理光学面积分技术相比,这些线积分表达式有更高的计算效率。 ? 对近场物理光学面积分积分核进行二次相位和线性幅度近似,使用驻相法推导出具有一致性绕射 理论形式的解。相对传统驻相法,无需驻相点搜索并避免了驻相点相互靠近时的数值精度问题。 基于三角面元网格的自适应细分技术,实现了按场值误差自适应收敛的任意目标近场散射计算。 该方法的计算效率与频率无关,计算超电大尺寸目标(几百波长或上千波长)时相对于传统面积 分方法有很大的优势。 ? 通过构造闭合参量区间积分形式的时域并矢势函数,结合矢量定理将本文导出的频域物理光学近 场线积分表达式拓展到时域情形。特殊地,当激励源位于从观察点/镜像点到板边缘的连线上时, 引入了开放参量区间积分形式的时域并矢势函数,避免了这一特殊情形时的奇异性。基于时域线 积分表达式的直接瞬态场计算相对于频域线积分技术结合 IFFT 在计算效率上有很大的优势。 ? 把本文导出的频域/时域线积分表达式退化至平面波激励情形,应用于频域/时域弹跳射线法,实 现了平面波激励的高效近场射线管积分计算。通过天线方向图的矢量点源分段近似和偶极子对分 段近似,把射线管积分计算转化为矢量点源/偶极子源激励的近场物理光学积分,使用本文导出的 矢量点源/偶极子源激励的频域/时域线积分表达式,实现了任意方向图激励的快速近场计算。 ? 基于局部远场近似条件对劈边缘进行细分,将远场截断劈等效边缘电磁流方法拓展至近场绕射计 算。照亮劈的近场绕射贡献表示为沿劈边缘的等效电磁流的线积分形式;照亮面的物理光学贡献 使用本文实现的线积分技术表示为沿面片边缘的围线积分形式。因此总的散射场可表示为一系列 线积分的叠加形式,得到高频渐进近似方法的完整线积分算法框架。 ? 使用深度学习方法直接对近场双站雷达回波信号进行识别,相应的训练与测试数据库通过本文实 现的近场高频渐进近似方法得到。该方法避免了复杂的回波信号预处理和特征提取,极大的简化 了识别过程,提供了一种全新的近场目标识别方法。
英文题目 NEAR-FIELD HIGH-FREQUENCY ASYMPTOTIC METHOD
英文主题词 Near-field Physical Optics, Line-integral, Stationary Phase Method, Shooting and Bouncing Ray Method, Near-field Diffraction
英文摘要 With the increase of working frequency of modern communication electronic devices, the requirement of analysing nearfield EM characteristics of target with super large size become urgent. Comparing with traditional full wave methods, the high frequency asymptotic method saves the storage and possesses a higher efficiency. High frequency asymptotic method therefore become one of the most important tools in analysing EM characteristics of targets with large electric size. Nearfield high frequency asymptotic method has been investigated in this thesis and the author’s major work and contribution are as follows: ? A novel line-integral representation of the physical-optics radiation integral from a perfectly conducting surface illuminated by a finite number of electric/magnetic Hertzian dipoles is presented. This novel representation guarantees the integrand free from singularities along the computational path. As such, it can be easily integrated for arbitrary positions of the source and observation points. The time reduction performance would be improved by this novel formula. The line-integral representation offers an alternative method to rapidly solve the scattering problem as it is usually more efficient than producing the same result in conventional surface-integral. ? Contour-integral representations have been presented to evaluate the physical-optics backscattered electric and magnetic fields from a perfectly conducting object illuminated by a dipole source. The advantage of the proposed representations is that the integrands in the contour integrals along the rim of the scatterer are free from singularities for all the source/observer positions. In addition, the proposed algorithm is easily extensible to analyse the scattering from complex objects. Such a liberty is achieved by deriving the novel representations using vector-algebra theorems in global coordinates. ? The field scattered by a scattering target can be described in terms of a double integral. Such integral on a triangular patch, with linear amplitude and quadratic phase variation approximation, can be exactly expressed in the format of uniform geometrical theory of diffraction which is convenient for numerical computations. This method can be applied to near-field scattering calculation scenario of arbitrary complex targets. The algorithm has some excellences like robustness and high efficiency. ? A time domain line integral representation of the physical-optics radiation integral is developed for the scattering from a perfectly conducting surface illuminated by an electric Hertzian dipole. The proposed line integral representation, which is valid and singularity-free for all near-field observers, is derived directly in the TD domain. This TD representation can be further expressed in terms of geometric-optics and boundary-wave components of the scattering mechanisms. Using the proposed line integral representation, the computational complexity is reduced by an order of the scatterer’s linear electric size, compared with the straightforward numerical quadrature of the PO surface integral. An efficient SBR method is developed to analyse the scattering responses from large perfectly conducting objects. By extending frequency/time domain line integral representation to the degenerate case of plane wave excitation, high computational efficiency can be achieved in SBR method under plane wave illumination. In addition, through vector point/dipole source piecewise approximation of antenna pattern, line integral representation can be used in SBR method under arbitrary antenna illumination. The proposed SBR method in conjunction with line integral technique is more efficient in near field calculation of complex targets than conventional SBR method combining surface integral technique. ? Numerical implementation of the method of truncated wedge equivalent edge current (TW-EEC) is presented and be extended to the calculation of near field diffraction. This method compensates the error in near field calculation of only considering the physical optics contribution. The contributions of the TW-EEC component and the physical optics component have exactly the same line integral form, so they can be combined to generate a complete near field line integral computing framework. ? A new automatic target recognition system based on bistation radar echo signal is proposed using deep learning method. This method can avoid the complex process of matched filtering in SAR images generation. The corresponding training and testing database is obtained by means of our proposed near-field line-integral method. The method of deep learning is adopted to avoid the complicated preposition and feature extraction processes of the echo signal which transform the raw input into a representation. Thus the recognition process can be greatly simplified.
学术讨论
主办单位时间地点报告人报告主题
东南大学 2015.3 李文正楼北614 Dr. Bo Zhao 场-路-粒子及多尺寸联合仿真技术
东南大学 2015.4 李文正楼北614 Prof. Levent Sevgi Electromagnetic Diffraction Modeling and Simulation
东南大学 2015.5 李文正楼北614 Prof. Jian Ming Jin From FETD to DGTD for CEM
东南大学 2016.3 无线谷5111 范湉湉 Fast Physical Optics (FPO) Algorithm for High Frequency Scattering
东南大学 2016.6 无线谷5111 范湉湉 Near Field Equivalent Edge Currents for Truncated Wedge Strips
东南大学 2016.9 无线谷5111 范湉湉 Physical Optics Scattering Under Gaussian Beam Illumination
东南大学 2017.4 无线谷5111 范湉湉 Convolutional Neural Networks for SAR Images Classification
东南大学 2017.11 无线谷5111 范湉湉 Automatic target recognition from raw radar echoes with deep learning
     
学术会议
会议名称时间地点本人报告本人报告题目
IEEE 2015.10 上海 Efficient Evaluation of Near-Field PO Scattering from PEC Objects Illuminated by Arbitrary Far-Field Sources
IEEE 2016.11 New Delhi, India Uniform Line-Integral Formulations for the Evaluation of High-Frequency Near-Field Scattering from Arbitrary PEC Targets
     
代表作
论文名称
Numerically Efficient Line-Integral Representation of Physical-Optics Scattered Field: The Case of P
Time-Domain Line-Integral Representations of Physical-Optics Scattered Fields
Singularity-Free Contour-Integral Representations for Physical-Optics Near-Field Backscattering Prob
Efficient evaluation of near-field PO scattering from PEC objects illuminated by arbitrary far-field
 
答辩委员会组成信息
姓名职称导师类别工作单位是否主席备注
方大纲 正高 教授 博导 南京理工大学
顾长青 正高 教授 博导 南京航空航天大学
程强 正高 教授 博导 东南大学
蒋卫祥 正高 教授 博导 东南大学
马慧锋 正高 教授 博导 东南大学
      
答辩秘书信息
姓名职称工作单位备注
蒋忠进 副高 副教授 东南大学