多层快速多极子算法中的两步插值技术

多层快速多极子算法(MLFMA)在快速多极子算法(FMM)的基础上按多层聚集、层间转移和多层扩散的思路以达到优化矩阵向量积的运算的目的,其中多层聚集和多层扩散过程,随着层数递增,角谱积分采样点数逐层递增,为了快速计算角谱积分,需要采用插值技术和反插值技术以提高计算效率.应用两步插值技术替代传统的单步插值技术,大幅提高了多层快速多极子层间插值反插值操作的计算效率,对于应用普通个人计算机求解特大电大尺寸问题,具有重要意义.     

模式匹配与积分方程在HPM模拟中的应用

针对具有电大尺寸、厚介质罩且馈源采用特定模式激励的高功率微波(HMP)传输及辐射结构研究了一种新型的电磁建模技术。将模式匹配方法与积分方程方法进行混合,构建了电磁模型的方程组,采用多层快速多极子技术、预条件器等进行求解加速,最终形成了可对电大尺寸HPM传输、辐射系统进行高效全波电磁仿真的技术。以电大变张角喇叭馈源、功率合成天线、波束波导及波束波导天线作为实例,构建了几何及电磁模型,并进行了包括远区方向图、近区功率密度分布在内的数值模拟,对该技术的正确性和通用性进行了验证。结果表明,该技术边界拟合准确,内存消耗低,且对馈电模式能予以准确反映,适用于HPM传输发射的高效高精度模拟。     

机载多天线系统电磁兼容性研究

为了在有限计算机资源条件下快速解决机载平台中多天线系统的电磁兼容(EMC)问题,基于电场积分方程(EFIE),实现自动分层的多层快速多极子方法(MLFMA).针对天线问的近场耦合特性,采用任意天线间的隔离度代替传统隔离度,使得隔离度计算结果更加合理.同时采用BiCGStab(l)结合近场预条件方法进行求解,进一步提高MLFMA的综合效率.最后实例计算并分析波音747飞机模型上多根超短波天线的辐射方向图和宽频带内天线间的隔离度.数值结果表明方法准确有效.     

有耗半空间上方导体目标电磁散射的多层快速多极子算法

提出基于高阶拟合的多层快速多极子算法,用于分析有耗半空间上方导体目标的电磁散射.采用高阶拟合方法将远区相互作用单元的半空间空域格林函数用级数表示,对高阶拟合项作加法定理展开,进而应用多层快速多极子算法.该方法具有分组小,内存需求低,易于实现的优点.用数值算例验证其正确性与有效性.     

快速计算地面上车辆目标的电磁散射特性

为快速计算地面上车辆目标的电磁散射特性,实现适用于半空间环境的多层快速多极子方法.对于近区项,利用严格的半空间并矢格林函数进行计算;对于远区项,首先将半空间格林函数分解为直射项和反射项,然后采用实镜像方法计算反射项.方法物理概念清晰,与目前常用的离散复镜像方法相比,具有良好的收敛性和稳定性,大大降低计算时间和内存需求.实例计算并分析地面上汽车模型的电磁散射特性.     

MLFMA用于三维电大复杂腔体电磁散射特性的快速计算

基于迭代物理光学法的多层快速多极子算法可高效、快速求解电大腔体的散射特性,给出混合算法的迭代公式.对工程上常见的弯曲较缓慢的S弯腔体,通过合适的分段结构化分组,避免两组之间存在的部分面元遮挡的情况,应用广义互易积分处理带有复杂终端的腔体.数值计算结果表明,所用方法准确并能极大地提高计算速度.     

含腔电大尺寸导体目标电磁散射的一体化数值模拟

应用等效原理，通过引入口面上等效磁流将含腔导电目标电磁散射简化为腔内、外两个等效 问题. 腔内问题分段求解并应用级联法获得口面等效导纳矩阵；腔内外的耦合关系应用近似 边界元方法描述并由此获得口面等效磁流；最后，这一具有混合源的腔体内外一体化散射问 题则应用所提出的广义混合场积分方程方法建立电磁模型，并用多层快速多极子方法实现高 效数值求解. 实例计算结果与测试结果具有很好的一致性. 关键词： 含腔目标 电磁散射 混合场积分方程 数值分析     

有耗介质层上多导体传输线的电磁耦合时域分析方法

目前,针对空间电磁场作用有耗介质层上传输线的电磁耦合,仍缺乏有效的数值分析方法.因此,本文提出一种高效的时域混合算法,很好地解决了有耗介质层上传输线电磁耦合建模难的问题.首先,对经典传输线方程进行改进,推导了适用于有耗介质层上多导体传输线电磁耦合分析的修正传输线方程.然后,结合时域有限差分方法和相应插值技术,求解修正传输线方程,获得多导线及其端接负载上的电压和电流响应,并实现空间电磁场辐射与多导线瞬态响应的同步计算.最后,通过相应计算实例的数值模拟,与CST软件的仿真结果进行对比,验证了时域混合算法的正确性和高效性.     

二维声学多层快速多极子边界元及其应用

与模型自由度的平方成正比的存储量和计算量,使传统边界元无法应用到大型模型的计算。为此,发展了一种二维声学多层快速多极子边界元算法。通过二维Helmholtz核函数展开理论的简要介绍,推导了源点矩计算、源点矩转移、源点矩至本地展开转移、本地展开转移公式,并详细描述了二维声学快速多极子边界元算法的具体实现步骤。使用快速傅里叶插值进行源点矩和本地展开系数的多层传递。采用对角左预处理方法,改善边界方程的条件数,减少迭代求解次数。最后通过数值算例,验证了所发展的二维声学快速多极子算法的正确性和高效性。      

三维声学多层快速多极子边界元及其应用

快速多极子边界元算法可以加速矩阵和向量乘法运算, 将传统边界元算法的计算量和内存占用量分别降为O(N log²N)和O(N), 适用于大型声学模型模拟计算. 本文发展了一种基于Burton-Miller方程的三维多层声学快速多极子边界元算法. 将新的自适应树状算法应用到对角形式的快速多极子边界元算法, 并使用最新提出的解析式源点矩计算公式, 进一步提高了快速多极子边界元的计算效率. 绝对软球体在内部共振频率处的散射声场计算, 验证了所发展算法在共振频率处求解的正确性. 与Bapat所提供的程序在多脉动球体辐射声场计算精度的比较, 验证了算法及程序在大型模型声学计算中的准确性, 同时显示了其求解的高效性. 最后, 将该算法用于车内声场及水下声学探测的分析计算.     

复相介质平稳随机场热传导方程的多尺度模型与计算

本文讨论复相介质(含孔洞)的热传导问题,其中几何形状,大小和材料分布形态构成一平稳随机场,针对这一问题,提出多尺度模型和算法,并给出数值实验结果。     

Light scattering theories and computer codes

In aerosol science today light scattering simulations are regarded as an indispensable tool to develop new particle characterization techniques or in solving inverse light scattering problems. Light scattering theories and related computational methods have evolved rapidly during the past decade such that scattering computations for wavelength sized nonspherical scatterers can be easily performed. This significant progress has resulted from rapid advances in computational algorithms developed in this field and from improved computer hardware.In this paper a review of the recent progress of light scattering theories and available computational programs is presented. We will focus on exact theories and will not cover approximate methods such as geometrical optics. Short outlines of the various theories are given alongside with informations on their capabilities and restrictions.     

Comparisons of the discrete-dipole approximation and modified double interaction model methods to predict light scattering from small features on surfaces

Two numerical methods to model light scattering from illuminated features on surfaces are presented. The discrete-dipole approximation (DDA) method is considered, as well as the modified double interaction method (MDIM). The DDA method models electromagnetic scattering of continuous features using discrete dipoles placed on a lattice structure. Sommerfeld integral terms are used to model dipole/surface interaction in the near-field. The MDIM method first computes scattering from the features based in free space using other methods such as Mie theory or other standard light scattering codes (including DDA). The surface interaction is modeled as a first approximation by means of a geometrical shadowing effect and the Fresnel coefficients. Comparisons of the methods will be shown for light scattering from spherical features. The material properties of dielectric and metallic materials will be considered and the feature sizes will be varied. The prediction accuracy and computational requirements of each method will be investigated. For most cases, the studies will show that the DDA method is more accurate than the MDIM method for dielectric materials since the modeling of the feature and surface electromagnetic interaction is more accurate; however, the modified double interaction method may be advantageous over the discrete-dipole approximation method for metallic features because of lesser computational times and memory requirements.     

Reduced-order modeling of time-dependent reflectance profiles from purely scattering media

Due to the widespread existence and importance of foam, inverse techniques for characterizing industrial foams are of interest. An essential element in an inverse method used to characterize a foam layer is a model of the time-dependent reflectance of a laser pulse. Monte Carlo methods may be used to accurately model reflectance, but these methods are computationally expensive. Computationally efficient methods based on the diffusion approximation have been developed, but this approach is not sufficiently accurate in many cases of interest. Therefore, a computationally efficient and robust method is desirable. This paper presents a computationally efficient method for modeling the time-dependent reflectance of a laser pulse from a non-absorbing, scattering plane layer that is based on reduced-order modeling techniques. The accuracy of the proposed method is demonstrated by comparing reflectance profiles for randomly selected foam layer properties with corresponding profiles that were generated from Monte Carlo simulations.     

Geometrical characterization techniques for microlens made by thermal reflow of photoresist cylinder

Microlens characterization is a prerequisite for improving fabrication process, and for satisfying the end user needs. In this paper we explore techniques to characterize geometrical properties of microlens made by thermal reflow: viz. microlens profile; radius of curvature; microlens height; contact angle and focal length. The geometrical characterization is done using techniques such as contact profilometry, scanning electron microscopy (SEM), optical microscopy, white light confocal microscopy and fluorescence confocal microscopy. All the above techniques are studied and compared, keeping in mind the characterization requirements of polymer microlens made by thermal reflow technique.     

CUDA-based four-path method with application to the EM scattering of a two-layer canopy

The long-term goal of this paper is to develop a robust simulator to study the EM scattering from vegetation. In an effort to overwhelm the intensive computational burden results from large sampling numbers, we decided to utilize the Graphics Processing Unit (GPU), which has been greatly developed in recent years. In this paper, Compute Unified Device Architecture (CUDA) is combined with the four-path method to predict the EM scattering properties from scatterers which are sampled by using the Monte Carlo method in a two-layer canopy model. Obviously, a speedup of 77.8 times could be readily obtained in comparison with the original serial algorithm on a Core(TM) i5 CPU with the help of a GTS250 GPU as a coprocessor.     

Frequency band structure and absorption predictions for multi-periodic acoustic composites

This article introduces a computational framework for studying frequency band structure and absorption behavior in multi-periodic acoustic composite structures. Herein, multi-periodic acoustic composite structures are defined as periodically-layered acoustic media wherein each layer is composed of periodically-repeated fluid unit cells, especially those arising from the study of porous materials. Hence, at least two periodic scales (microscopic and mesoscopic, respectively) comprise the macroscopic acoustic composite media. Exploitation of the multi-periodicity allows for efficient generation of dispersion and absorption curves via the conventional multi-scale asymptotic method (for homogenizing the mesoscale) coupled to the acoustic transfer matrix methods (for the macroscale). The combined computational framework results in a single analysis procedure for evaluating complex dispersion relationships and acoustic absorption. The dispersion curves can be used to reveal frequency stop bands wherein the wave vector is highly imaginary—i.e., plane waves experience rapid attenuation. They can also be used to reinterpret classical absorption curves. The framework is applied to four multi-periodic acoustic composite structures in order to demonstrate the framework's utility and to reveal novel properties, particularly those which can be influenced by design of the mesoscopic unit cell.     

基于GADF变换和多尺度CNN的哈密瓜表面农药残留可见-近红外光谱判别方法

针对哈密瓜表面农药残留化学检测方法成本高且具有破坏性等问题,探索了可见-近红外（Vis-NIR）光谱技术对农药残留定性判别的可行性。以哈密瓜为载体,百菌清和吡虫啉农药为研究对象,采集哈密瓜表面无残留、百菌清和吡虫啉残留的可见-近红外漫反射光谱,利用格拉姆角场（GAF）将一维光谱数据转换为二维彩色图像,构建GAF图像数据集。设计一种包含Inception结构的多尺度卷积神经网络模型用于哈密瓜表面农药残留种类判别,包括1层输入层、3层卷积层、1层融合层、1层平坦层、2层全连接层和1层输出层。模型测试混淆矩阵结果表明,格拉姆角差场（GADF）变换对哈密瓜表面农药残留的可见-近红外光谱表达能力较强。此外,构建AlexNet、VGG-16卷积神经网络（CNN）模型和支持向量机（SVM）、极限学习机（ELM）机器学习模型与提出的多尺度CNN模型进行性能对比。结果表明,3种CNN模型对哈密瓜表面有无农药残留的判别效果较好,综合判别准确率均高于SVM和ELM模型。对比3种CNN模型性能,多尺度CNN模型的性能最佳,训练耗时为14 s,综合判别准确率为98.33%。多尺度CNN模型结构利用多种小尺寸滤波器组合（1×1,3×3和5×5）和并行卷积模块,能够捕获不同层次和尺度的特征,通过级联融合模式进行深度特征融合,提高了模型的特征提取能力。与传统深度CNN模型相比,在保证计算复杂度不变的情况下,多尺度CNN模型的精度得到了有效提高。实验结果表明,GADF变换结合多尺度CNN模型可以有效进行光谱数据解析,利用可见-近红外光谱技术可以实现哈密瓜表面农药残留的定性判别。研究结果为大型瓜果表面农药残留的快速无损检测技术的研发提供了理论参考。     

An improvement to computational efficiency of the drain current model for double-gate MOSFET

As a connection between the process and the circuit design, the device model is greatly desired for emerging devices, such as the double-gate MOSFET. Time efficiency is one of the most important requirements for device modeling. In this paper, an improvement to the computational efficiency of the drain current model for double-gate MOSFETs is extended, and different calculation methods are compared and discussed. The results show that the calculation speed of the improved model is substantially enhanced. A two-dimensional device simulation is performed to verify the improved model. Furthermore, the model is implemented into the HSPICE circuit simulator in Verilog-A for practical application.     

Surface elasticity and size effect on the vibrational behavior of silicon nanoresonators

Predominance of nano-scale effects observed in material behavior at small scales requires implementation of new simulation methods which are not merely based on classical continuum mechanic. On the other hand, although the atomistic modeling methods are capable of modeling nano-scale effects, due to the computational cost, they are not suitable for dynamic analysis of nano-structures. In this research, we aim to develop a continuum-based model for nano-beam vibrations which is capable of predicting the results of molecular dynamics (MD) simulations with considerably lower computational effort. In this classical-based modeling, the surface and core regions are taken to have different mechanical properties, where core atoms are assumed to have macroscale properties whereas surface layer is showing a different elastic modulus from the core components. By estimating physical parameters of proposed classical model using molecular dynamics results and the genetic algorithm, calibrated classical Euler–Bernoulli and Timoshenko beam models are developed. The results demonstrates that a Timoshenko beam model incorporating surface effects and having calibrated parameters, is able to provide almost the same results as molecular dynamics method which can be used to predict the vibrational behavior of nano-beams at different scales from nano to macro.