RBF神经网络的混合微粒群学习算法

在分析RBF神经网络的结构特点基础上, 定义一个布尔向量L作为网络的结构参数, 与原来RBF神经网络的隐节点参数集一起构成了新的RBF网络隐节点参数集{c,σ, L}, 并给出了一个新的RBF网络输入输出关系表达式;采用一种混合协同微粒群算法同时对RBF网络拓扑结构和隐层节点参数进行优化设计,并将输出线性参数集分离后采用最小二乘法进行优化设计,简化了优化空间,加速了算法的收敛速度.     

粒子群优化算法的发展趋势

分析优化算法的发展历程,指出粒子群优化算法(PSO)是基于群体智能的一种算法,简单易行,可调参数少,研究广泛且发展迅速.结合图形给出算法的个体极值和整体极值的搜优运动过程.研究总结算法的研究现状及特点,认为PSO还需要完善和继续研究.提出将算法应用于复杂的约束优化、随机优化与最优控制问题是算法应用研究的方向,并指出对该算法完整的收敛性分析是算法成熟的标志.     

基于双极偏好控制的多目标粒子群优化算法

考虑双极偏好信息对粒子群的控制作用,提出一种使用双极偏好——正偏好和负偏好引导粒子群向 Pareto 前沿偏好区域进化的方法．根据TOPSIS 决策法思想,将外部种群粒子与正负偏好点的相对贴近度排序作为 外部种群管理和全局最优解更新策略;根据贴近度值确定解集的分布度;选取6 种不同类型的多目标测试函数进行 算法模拟,从世代距离、空间测度和超体积测度3 个指标与基于单极偏好的多目标粒子算法进行性能比较．结果显 示,基于双极偏好控制的多目标粒子群算法的收敛性和综合性能更优秀．     

基于量子粒子群优化的在线航迹规划

现代战场中,环境信息是变化的,飞行器很难预先获得精确的全局环境信息,因此要求无人飞行器具有实时的航迹规划能力,采用量子粒子群优化算法,将约束条件和搜索算法相结合,有效解决了简单粒子群算法在高维空间中易陷入局部最优点的问题;同时,根据地形障碍、敌方防御雷达、防空火力等威胁以及禁飞区的分布情况,引入最小威胁面的概念,利用B-Spline插值逼近最小威胁面中的三维航迹在二维水平面内的投影,从而将三维曲线的规划问题简化为二维平面中控制点的寻优问题,简化了问题复杂度,提高了计算效率.仿真结果表明该方法可以满足在线航迹规划的要求.     

改进的粒子群算法

针对基本粒子群算法容易陷入局部最优点,进化后期速度慢等缺点,设计了一种新的粒子群算法,将基本粒子群算法粒子行为基于个体极值点转化为个体自身极值与其他某一个个体极值的加权平均值,而全局极值点转化为群体中优秀个体极值的加权平均值。数值仿真实验表明,新算法比PSO具有更好的收敛性,能更快地找到问题的最优解。     

Cryo-EM image alignment based on nonuniform fast Fourier transform

In single particle analysis, two-dimensional (2-D) alignment is a fundamental step intended to put into register various particle projections of biological macromolecules collected at the electron microscope. The efficiency and quality of three-dimensional (3-D) structure reconstruction largely depends on the computational speed and alignment accuracy of this crucial step. In order to improve the performance of alignment, we introduce a new method that takes advantage of the highly accurate interpolation scheme based on the gridding method, a version of the nonuniform fast Fourier transform, and utilizes a multi-dimensional optimization algorithm for the refinement of the orientation parameters. Using simulated data, we demonstrate that by using less than half of the sample points and taking twice the runtime, our new 2-D alignment method achieves dramatically better alignment accuracy than that based on quadratic interpolation. We also apply our method to image to volume registration, the key step in the single particle EM structure refinement protocol. We find that in this case the accuracy of the method not only surpasses the accuracy of the commonly used real-space implementation, but results are achieved in much shorter time, making gridding-based alignment a perfect candidate for efficient structure determination in single particle analysis.     

考虑电-氢-热多能互补的微网多目标优化配置研究

氢储能具有储能容量大、储存时间长、清洁无污染、实现多种能源网络互联互补和协同优化等诸多优点,有望成为推动分布式能源发展和提升终端能源利用效率的重要支撑技术。为了提高独立型微网供电可靠性及可再生能源利用率,本文建立了考虑电-氢-热多能互补的独立微网多目标优化配置模型,并基于模拟退火的粒子群算法对目标问题进行求解。最后,通过东北某地独立微网优化配置算例,基于MATLAB平台验证了所提多能互补配置方案较传统电储能配置方案负荷失电率降低了3.18%,可再生能源利用率提高了8.37%,所提配置方案可有效促进可再生能源消纳,保证独立微网的供电可靠性。     

Reconstructing design parameters of a rectangular resonator via peak signal-to-noise ratio and global optimization algorithms

This article proposes an approach for reconstructing physical parameters of a sample in a rectangular resonator during microwave radiation, knowing a priori, its electric field distribution. The inverse problem was solved using two global optimization algorithms and the peak signal-to-noise ratio (PSNR) criterion. First, the Self-regulated Fretwidth Harmony Search algorithm (SFHS) identified suitable resonant frequencies for a given configuration. Next, the unified Particle Swarm Optimization (UPSO) optimized said configuration. Together, they became a maximization strategy of the PSNR through a dual optimization process. Results showed the ability of the approach for estimating the height of each sample block and the resonating frequency of the cavity. This process takes longer to finish as a higher PSNR is demanded (mainly due to the aforementioned dual optimization). Even so, it allows for more similar electric field distributions between both, the direct and inverse problems.     

Erosion coupled dynamic mesh analysis for location of maximum erosion in 90-degree bends for gas-solid flows

The progressive change in the surface geometry of the component due to erosive wear affects the correct estimation of erosive wear performance and service life of the components handling particulate flows. The current study focuses on determining the change in the location of higher erosion on the bend surface during the pneumatic conveying of solids with continuous geometric modification due to erosive wear. Computational fluid dynamics (CFD) based erosion-coupled dynamic mesh methodology is adopted to simulate the time-dependent surface modification of the 90° bend geometry due to erosive wear. Available experimental data are used to validate the numerical results. Further, the erosion distribution and the location of the maximum erosion for different flow velocities, particle sizes, and bend radius ratios with the increase in solid throughput are investigated. It has been found that the modification in the bend geometry due to erosion influences the location of the maximum erosion. The increase in thickness loss due to erosion increases the variation in the location of the maximum erosion. Furthermore, an equation for predicting the location of maximum erosion of bend geometry is obtained based on the bend radius ratio and the thickness loss.     

Direct numerical simulations for assessment of gas-solid drag models in two-dimensional random arrays of particles

The momentum exchange between the phases plays a vital role in modelling of gas–solid flows and it is mathematically described by drag models. However, no consensus exists on which drag model gives the most accurate prediction of the drag force, and, despite the increase in available computing power, the same drag models are used in two-dimensional and three-dimensional simulations. In this study, direct numerical simulations of gas flow through multiple random configurations of static monodisperse particles are performed. The variations of solid volume fraction and particle Reynolds number are in the ranges of 0.05–0.4 and 13.7–136.9, respectively. The drag force exerted on particles is calculated and properly averaged. Based on the simulation results, thirteen drag models are compared and correction factors are introduced using the stochastic gradient descent algorithm. The correction factors provide a simple adjustment for the models to be used in 2D modelling.