Hybrid modeling of microwave devices using multi‐kernel support vector regression with prior knowledge

This article proposes a support‐vector hybrid modeling method of microwave devices when only a small number of measurements are available. In this method, a hybrid model of microwave device has been obtained by combining a coarse model and a support‐vector model, where the coarse model is complemented by a support‐vector model capable of correcting the difference between the measurements and the coarse model. The support‐vector model was developed using a novel algorithm. In the algorithm, multi‐kernel and prior knowledge from a calibrated simulator were incorporated into the framework of the linear programming support vector regression by utilizing multiple feature spaces and modifying the optimization formulation. The experimental results from two microwave devices show that the hybrid modeling can enhance the physical meaning of the support‐vector model and improve the modeling accuracy for a small dataset, and that the proposed algorithm shows great potential in some applications where sufficient experimental data is difficult and costly to obtain, but the prior knowledge from a simulation model is available. The hybrid modeling is suited to a microwave computer‐aided design tool or an automatic tuning robot. © 2014 Wiley Periodicals, Inc. Int J RF and Microwave CAE 25:219–228, 2015.     

Parametric modeling of microwave filter using combined MLS‐SVR and pole‐residue‐based transfer functions

In this article, an advanced technique is developed to combine multi‐output least‐squares support vector regression (MLS‐SVR) and pole‐residue‐based transfer function models for microwave filter parametric modeling. MLS‐SVR is trained to learn the relationship between the length of tuning screws and the pole/residues of the transfer function, where MLS‐SVR is an effective method to cope with the multi‐output case unlike the traditional approach. Traditional approach treats the different outputs separately in the multi‐output case and it cannot model the relation between different outputs. Another important element for modeling is feature parameters. Extracted feature parameters have an important influence on the accuracy of modeling. For the purpose of establishing a more accurate model, the complex system poles and residues from Y‐parameters are chosen as the outputs of modeling, which are obtained by vector fitting (VF). Then we give a solution to obtaining pole/residues extracted by VF when the filter is in high detuned. After the proposed modeling process, trained model can be used to provide an accurate and fast prediction of the behavior of microwave filter with the length of tuning screws as variables, and model the electromagnetic simulation (or actual) microwave filter tuning. The methodology is applied to a narrow band coaxial‐resonator filter modeling, and more accurate results are achieved compared with the other methods. An example is presented to illustrate the efficiency of the proposed method.     

Coordinated formation pattern control of multiple marine surface vehicles with model uncertainty and time-varying ocean currents

This paper presents a data-driven modeling and optimization method for cavity filters, according to a limited amount of measurement data. In the method, a model that reveals the effect of mechanical structure on electrical performance of cavity filters is firstly developed by an improved algorithm, which can increase the modeling accuracy of small data set by incorporating multi-kernel and prior knowledge into the framework of linear programming support vector regression. Then, an approach to optimize the structure of cavity filters is formulated by using the developed data-based model, and the obtained results can assist the fabrication of the same filter in the future. Some experiments from a synthetic example and a practical application of cavity filter have been carried out, and the experimental results confirm the effectiveness of the method. The model is particularly suited to a computer-aided manufacturing of volume-producing filters, and the proposed algorithm shows great potential in some applications where the experimental data are very few and the prior knowledge is available.     

Influence and tuning of tunable screws for microwave filters using least squares support vector regression

This article presents an approach that can analyze the influence of tunable screws and perform a computer‐aided tuning for microwave filters. In the approach, a machine‐learning model that reveals the influence of tunable screws on the filter response is first developed by least squares support vector regression, according to some data from the tuning experience of filters. Then a computer‐aided tuning procedure based on the model is proposed, and the obtained adjusting amount of tunable screws can assist an unskilled operator to perform a fast and accurate tuning. The approach is validated by some experiments and the results confirm the effectiveness. The approach is particularly suitable to the computer‐aided tuning of volume‐producing filters. © 2010 Wiley Periodicals, Inc. Int J RF and Microwave CAE, 2010.     

集成先验知识的多核线性规划支持向量回归

为了解决工程中数据样本较少情况下的准确建模问题, 提出了一种集成先验知识的多核线性规划支持向量回归算法. 该算法首先通过修改优化目标和不等式约束条件, 把来自仿真模型具有偏差的先验知识数据集成到现有的线性规划支持向量回归的学习框架中. 然后, 引入多核到集成先验知识的线性规划支持向量回归中以实现复杂规律的准确建模. 最后, 将算法推广到多输入多输出的数据建模中. 仿真案例以及在天线和滤波器的实际应用表明: 该算法求解简单, 具有较好的模型稀疏和准确性.     

Tuning model for microwave filter by using improved back‐propagation neural network based on gauss kernel clustering

Given the difficulty of a single model in dealing with complex systems. In this study, we propose a tuning model that uses a probabilistic fusion of sub‐optimal back‐propagation neural network based on the Gauss kernel clustering. This study focused mainly three aspects of work compared with the traditional tuning model. First, the calculation of the coupling matrix of scattering parameters is achieved by solving polynomial coefficients after eliminating the inconsistent phase shift and resonant cavity loss. Second, the best clustering center and a number were obtained by mapping the scattered data to high‐dimensional space, and the prediction of multi‐output variables were realized by sub‐model probability fusion. Third, an improved shuffled frog leaping algorithm was introduced to optimize the initial weights of the back‐propagation neural network, and a differential operation significantly improved the diversity of the population and the searchability of the algorithm. Finally, the experiment of nine‐order cross‐coupled filters shows that the proposed method has a better capability to train the weights and thresholds, which improves the generalization performance of the system.     

Reduced‐cost microwave filter modeling using a two‐stage Gaussian process regression approach

A technique for the reduced‐cost modeling of microwave filters is presented. Our approach exploits variable‐fidelity electromagnetic (EM) simulations, and Gaussian process regression (GPR) carried out in two stages. In the first stage of the modeling process, a mapping between EM simulation filter models of low and high fidelity is established. The mapping is subsequently used in the second stage, making it possible for the final surrogate model to be constructed from training data obtained using only a fraction of the number of high‐fidelity simulations normally required. As demonstrated using three examples of microstrip filters, the proposed technique allows us to reduce substantially (by up to 80%) the central processing unit (CPU) cost of the filter model setup, as compared to conventional (single‐stage) GPR—the benchmark modeling method in this study. This is achieved without degrading the model generalization capability. The reliability of the two‐stage modeling method is demonstrated through the successful application of the surrogates to surrogate‐based filter design optimization. © 2014 Wiley Periodicals, Inc. Int J RF and Microwave CAE 25:453–462, 2015.     

New multi‐standard dual‐wideband and quad‐wideband asymmetric step impedance resonator filters with wide stop band restriction

New multi‐standard wide band filters with compact sizes are designed for wireless communication devices. The proposed structures realize dual‐wideband and quad‐wideband characteristics by using a new skew‐symmetrical coupled pair of asymmetric stepped impedance resonators, combined with other structures. The first and second dual‐wideband filters realize fractional bandwidths (FBW) of 43.2%/31.9% at the central frequencies (CF) of 1.875/1.63 GHz, and second bandwidths of 580 MHz/1.75 GHz at CF of 5.52/4.46 GHz, respectively. The proposed quad‐band filter realizes its first/second/third/fourth pass bands at CF 2.13/5.25/7.685/9.31 GHz with FBW of 46.0%/11.4%/4.6% and 5.4%, respectively. The wide pass bands are attributed to the mutual coupling of the modified ASIR resonators and their bandwidths are controllable by tuning relative parameters while the wide stop band performance is optimized by the novel interdigital cross coupled line structure and parallel uncoupled microstrip line structure. Moreover, the quad band is generated by introducing the novel defected rectangle structure. These multi‐standard filters are simulated, fabricated and measured, and measured results agree well with both simulated results and theory predictions. The good in‐band and out‐of‐band performances, the miniaturized sizes and simple structures of the proposed filters make them very promising for applications in future multi‐standard wireless communication.     

基于深度去噪核映射的长期预测模型

针对最小二乘支持向量机核函数结构较浅对其长期预测模型精度的限制,采用深度学习中逐层特征提取的思想,提出基于深度去噪核映射的最小二乘支持向量机长期预测模型.该模型通过深度核网络的逐层变换,将样本数据映射到深度特征空间,从而有效提高其长期预测的精度.此外,为了提高模型对含高噪声数据的拟合性能,将去噪算法融入深度核网络的训练过程中,并通过反向传播算法对核网络参数进行整体微调.标准数据集及实际工业数据的仿真实验结果表明,所提方法能够有效提取数据中蕴含的特征信息,提高预测模型的精度.     

Efficient multi‐fidelity design optimization of microwave filters using adjoint sensitivity

A simple and robust algorithm for computationally efficient design optimization of microwave filters is presented. Our approach exploits a trust‐region (TR)‐based algorithm that utilizes linear approximation of the filter response obtained using adjoint sensitivity. The algorithm is sequentially executed on a family of electromagnetic (EM)‐simulated models of different fidelities, starting from a coarse‐discretization one, and ending at the original, high‐fidelity filter model to be optimized. Switching between the models is determined using suitably defined convergence criteria. This arrangement allows for substantial cost reduction of the initial stages of the optimization process without compromising the accuracy and resolution of the final design. The performance of our technique is illustrated through the design of a fifth‐order waveguide filter and a coupled iris waveguide filter. We also demonstrate that the multi‐fidelity approach allows for considerable computational savings compared to TR‐based optimization of the high‐fidelity EM model (also utilizing adjoint sensitivity). © 2014 Wiley Periodicals, Inc. Int J RF and Microwave CAE 25:178–183, 2015.     

Substrate integrated folded waveguide controllable mixed electric and magnetic coupling structure and its application to millimetre‐wave pseudo‐elliptic filters

A substrate integrated folded waveguide (SIFW) controllable mixed electric and magnetic coupling structure is proposed and investigated, which is suitable for millimeter‐wave pseudo‐elliptic filter applications. The proposed structure is built by a meandered slot‐line (MSL) combined with an inductive via‐hole in a SIFW cavity, which can support two closely‐spaced resonator modes and achieves a mixed coupling. By adjusting the physical sizes of the mixed coupling structure, the electric and magnetic coupling can be separately controlled and a controllable transmission zero (TZ) can be produced below/above the passband. Furthermore, detailed analysis of the SIFW cavity mode spectrum and the mixed coupling characteristics are presented. Finally, direct‐coupled and cross‐coupled fourth‐order pseudo‐elliptic filters are designed, fabricated and measured to verify the proposed mixed coupling structure. The experimental results show that the proposed filters have good performance, such as high frequency selectivity, compact size, wide upper stopband, and easy integration with other planar circuits.     

基于改进PSO算法和LS-SVM的短期电力负荷预测

针对电力负荷的小样本、非线性、高维数和局部极小点等问题,提出采用最小二乘支持向量机方法建模,以历史负荷、温度、湿度等数据作为输入量,对短期电力负荷进行预测;针对最小二乘支持向量机在建模中存在的参数选取问题,采用一种根据种群多样性信息来指导初始种群选取和避免粒子早熟收敛现象的改进粒子群优化算法来优化最小二乘支持向量机的惩罚因子和核参数。仿真结果表明,基于改进粒子群优化算法和最小二乘支持向量机的短期电力负荷预测方法较最小二乘支持向量机预测方法、基于基本粒子群优化算法和最小二乘向量机的预测方法具有更好的预测精确度。     

Computer aided design for dual‐band filters with complex source and load impedances

This article presents a computer aided technique to synthesize the filtering function for dual‐band microwave filters with complex source and load impedances. Based on the exchange‐Remez algorithm and the power wave renormalization theory, the characteristic function is first obtained with assigned ripple level at the first passband. Two additional transmission zeros are then optimally allocated to control the ripple level at the second passband. By directly modifying the coupling matrix elements in the first and last stages, a simplified method is also presented when the loading impedance is frequency‐independent. Three synthesized examples with both controlled in‐band and out‐of‐band responses are presented. Finally, LC multi‐coupled circuits are also provided to validate the effectiveness of the proposed method.     

基于核方法的非线性时间序列预测建模

提出了一种基于核的非线性时间序列预测建模方法。对非线性时间序列的相空间进行重构以确定其嵌入维数,并提出一种基于核主成分分析的非线性时间序列相空间重构方法,针对时间序列的时序特征,采用一种加权的支持向量回归模型对时间序列预测建模。在不同基准数据集上的实验结果表明,与通常的基于普通支持向量回归的建模方法相比,该文所提出的预测建模方法具有较高的精度,说明所提方法对非线性时间序列的预测建模是有效的。     

Compact tri‐band BPF applying novel multi‐mode microstrip resonator

This article presents a novel multi‐mode microstrip resonator. Using the even‐odd‐mode method, its resonance characteristics are analyzed and the design graphs are given. Each mode equivalent circuit is a λ/4 stepped impedance resonator (SIR), so the proposed resonator has a compact size and the higher harmonics can be tuned in a wide range. Stub–stub coupling is introduced to split two identical modes and produce two transmission zeros (TZs). Then a tri‐band filter operating at 1.5, 2.4, and 3.8 GHz is designed using the proposed resonator. The three center frequencies and bandwidths can be independently controlled. By tuning the impedance and length ratios of the stubs, wide upper stopband is achieved. Finally, the designed filter is fabricated and measured, and the measured results agree well with the simulated ones. © 2016 Wiley Periodicals, Inc. Int J RF and Microwave CAE 26:559–564, 2016.     

Dual‐band band pass filter using stub‐loaded open‐loop resonators with wide controllable bandwidths

Two dual‐band band pass filters (BPF) using stub‐loaded open‐loop (SLOL) resonator are presented in this article. A novel coupling tuning method by changing the relative coupling position of the resonators is proposed to control the bandwidth of each passband in a wide range. Transmission zeros are created to improve the selectivity by source‐load coupling. Because of the large ratio of two bandwidths, a novel dual‐band matching method is proposed to match the different load impedances at two passband frequencies to the same source impedance. Hence, relax the fabrication requirement of gap. The proposed dual‐band band pass filter is designed and fabricated. The measured 3 dB fractional bandwidths (FBWs) of two 2.45/5.25 GHz dual‐band BPFs are 6.5%/14.5% and 9.8%/5.5%, respectively. The results are in good agreement with the simulation. © 2013 Wiley Periodicals, Inc. Int J RF and Microwave CAE 24:367–374, 2014.     

Efficient design of SIW filters with knowledge‐embedded space mapping technique

This article presents an optimization technique for the design of substrate‐integrated waveguide (SIW) filters using knowledge‐embedded space mapping. An effective coarse model is proposed to represent the SIW filter. The proposed coarse model can be analyzed in the available commercial software ADS. The embedded knowledge includes not only formulas but also extracted design curves, which help to build the mapping between the coarse and fine models. The effectiveness of the proposed method is demonstrated through a design example of a six‐pole SIW filter. © 2012 Wiley Periodicals, Inc. Int J RF and Microwave CAE, 2012.     

一种贝叶斯证据框架下支持向量机建模方法的研究

研究贝叶斯证据框架下标准支持向量机和最小二乘支持向量机的估计算法和建模方法，给出一种高斯桉支持向量机估计算法的参数选择和调整方法，将贝叶斯证据框架下支持向量机的建模方法应用于非线性系统的辨识，仿真结果表明，该方法对于工业过程建模是十分有效的。     

K近邻优化估计的SAR图像建模与目标检测算法

在非均匀杂波环境下的合成孔径雷达(synthetic aperture radar, SAR)图像背景建模问题中,针对非参量建模算法Parzen窗估计严重依赖于窗宽设置及最优核函数选择的问题,提出一种基于K近邻优化的概率密度函数估计算法,解决因固定近邻数而导致估计不准确甚至不能估计的问题.该算法不需要图像的任何先验知识,且无需考虑窗宽的设置及最优核函数的选择问题.与Parzen窗估计、K分布和G⁰分布的对比实验表明,所提出的K近邻优化估计算法可以实现对单峰、多峰甚至不规则图像数据的准确建模,优于K分布和G⁰分布;同时,对图像首尾数据的处理优于Parzen窗估计.实验结果验证了所提出方法对SAR图像杂波建模的精确性、鲁棒性和简便性,以及全局恒虚警率目标检测的有效性.     

基于核相似性差异最大化的支持向量机参数选择算法

针对支持向量机(SVM)参数选择问题,通过分析SVM近似网络模型及分类原理,提出一种基于核相似性差异最大化的高斯核参数快速选择算法(MSD)。同时,将MSD算法与基于交叉验证的参数搜索算法相结合,构成一种复合SVM参数选择算法(MSD-GS),实现核参数与正则化参数的快速优选。UCI数据的仿真实验表明该算法具有参数选择准确、简便快速、无需数据先验知识等优点,参数选择效果甚至优于遍历式指数网格搜索算法。优选出的参数组合能够使SVM具有较高的泛化性能。