基于THz时域混沌特征的煤粉细度检测方法的研究

煤粉气力输送的细度检测对磨煤机工作状态的最优控制具有重要的意义。传统的检测方法多采用抽检取样法,通过分样筛等设备检测样品的细度,耗时长且操作复杂。国内外对细度地快速检测也有部分研究成果,但所测粉体浓度须较低,且设备稳定性还有待提高。太赫兹时域光谱技术（THz-TDS）是一种新型的无损检测技术,其低能性、选择透过性、相干性等特点使它具备其他光谱测量方法没有的优势。国内外对太赫兹辐射与颗粒介质相互作用的研究表明,太赫兹波对颗粒介质的细度具有显著敏感性,因此通过太赫兹波检测煤粉细度具有可行性。太赫兹波在高浓度颗粒介质的传播可以被看成是一个非线性动力过程,这个过程包含了复杂的非线性动力学效应,导致光谱测量结果具有混沌特征。将非线性动力系统的概念应用到太赫兹时域光谱信号的分析中,将太赫兹时域光谱信号视为由复杂非线性动力系统产生的时间序列进行特征分析。实验中将煤样研磨并筛分为<38.5,55~74,74~88,88~105和105~200 μm六种细度,并将煤粉与HDPE混合后压制成样品片。分别提取了的煤粉样品太赫兹时域光谱信号的功率谱熵、小波能量熵、盒维数、关联维数、偏度和峭度作为太赫兹时域光谱的混沌特征,通过比较发现这些混沌特征与细度变化具有一定的相关性,从视觉上可以大致区分出细度范围,但无法进行定量分析。支持向量机常用来解决小样本和非线性的分类问题,但是需要选择合适的参数才能建立较为准确的预测模型。文中引入粒子群算法来优化支持向量机建模参数选择。将上述提取的混沌特征向量作为粒子群算法优化的支持向量机的输入变量,以分样筛筛孔作为回归目标,对所测量煤粉细度建立回归模型。实验结果表明利用混沌特征建立的回归模型对<38.5和38.5~55 μm样品的预测结果要逊色于消光谱建模的回归结果,认为这是因为煤粉细度小,太赫兹波在样品中传播时与煤粉颗粒相互作用也比较弱,时域信号的混沌特征表现不明显所导致。对55~74,74~88,88~105和105~200 μm煤粉样品细度的预测结果要明显优于频域消光谱建立的模型,特别是74~88和105~200 μm样品,校正集均方根误差相对于消光谱分别下降了29.48%和26.14%,预测集误差分别下降了88.62%和56.86%。从预测结果整体上来看,采用混沌特征建模的预测结果与目标细度的相关系数为0.9618,消光谱建模的预测结果相关系数仅为0.78。混沌特征建模的均方根预测误差仅为9.52,消光谱建模的均方根预测误差为24.48。同时采用混沌特征的建模时间相对于消光谱的建模时间下降了43.19%。研究结果为太赫兹时域光谱技术在高浓度煤粉气力输送细度检测上的应用提供了科学依据和参考。     

基于PSO-BP模型的加筋双模量矩形薄板弹性模量反演

为了利用少量观测点高效反演材料的弹性模量,本文利用基于粒子群算法优化PSO(Particle Swarm optimization)的误差逆向传播网络BP(back propagation)建立反演模型,利用粒子群算法对误差逆向传播网络的参数进行寻优。以解析法求解双模量矩形薄板在不同弹性模量时的挠度作为训练样本,输入四个观测点的挠度值,利用PSO-BP模型对板的弹性模量进行反演。结果表明,PSO-BP模型可以建立挠度与弹性模量的联系,PSO算法可以提高BP模型的精度,加筋双模量矩形薄板的三个弹性模量的最大残差分别为39.052 kPa, 73.513 kPa和64.207 kPa,最大相对误差分别为1.722%,3.681%和3.637%。本模型可为工程实践提供参考和指导。     

季节调整的PSO-SVR模型及其在旅游客流量预测中的应用——以海南省三亚市为例

准确的旅游客流量预测对旅游目的地做好事前准备工作至关重要.然而旅游客流量具有明显的非线性和季节性特征,采取季节调整方法对样本数据进行预处理,消除季节性的影响,可以提高客流量预测的准确性.同时SVR(支持向量回归机)是一种良好的机器学习方法,非常适合预测研究,辅以PSO(粒子群算法)选取合适的回归参数可以获得更加精确的预测结果.提出了一种考虑季节影响并通过PSO优化SVR模型的旅游客流量预测模型,并以海南省三亚市为例进行了实证研究.研究结果表明,季节调整的PSO-SVR模型预测精度明显高于SVR、季节调整的SVR和PSO-SVR模型,是进行旅游客流量预测的有效工具.     

Cure kinetic parameter estimation of thermosetting resins with isothermal data by using particle swarm optimization

The heuristic algorithms have shown to be a powerful tool in parameter estimation. Among these algorithms, particle swarm optimization (PSO) has become a method whose application has been increasing quickly. In the present work a new way for parameter estimation from cure kinetic model of polymeric resin using a differential-algebraic approach is shown. The PSO was applied to minimize the least squares function and to find the parameters from an autocatalytic model for describing cure kinetics of thermosetting resins. The isothermal data were obtained at four temperatures: 318, 333, 348 and 363 K. Three parameter estimation procedures were compared for finding a parameter set for all temperatures simultaneously. In the first one, called classical method, a curing rate was calculated with experimental values of the degree of cure and the temperature. In the second and third methods, the curing rate was obtained from the integration of a differential-algebraic system and the main difference between them is the objective function and the way to determine the ultimate reaction heat. All methods showed good results; however, the third method was more accurate than the others. The confidence regions of all parameters were found and they were used to give us indication whether the parameters estimated here by different methods are statistically different.     

A novel optimal path-planning and following algorithm for wheeled robots on deformable terrains

An immense body of research has focused on path-planning and following of wheeled robots in unstructured surfaces. Nonholonomic robots traveling over deformable terrains together with complex operating conditions, however, pose further challenges in terms of a higher demand for robustness and optimality. In this paper, a Chaos-enhanced Accelerated Particle Swarm Optimization (CAPSO) algorithm is employed for planning an optimal path of a wheeled robot, so as to ensure shortest path from the starting point to the target location together with safety through guaranteed avoidance of collisions with static and dynamic obstacles. The fundamental terramechanics concepts are employed to derive essential forces and moments acting on the wheeled robot. Subsequently, a kineto-dynamic model of the robot is developed for designing a novel robust control algorithm based on an exponential-integral-sliding mode (EISMC) scheme and a RBF-NN approximator. The results revealed that the proposed algorithm is responsive and robust to withstand adverse effects of structured and unstructured uncertainties by using the designed adaptation law according to the Lyapunov stability theorem. The effectiveness of the proposed algorithm is also validated against several reported frameworks.     

Nonlinear hybrid system and parameter identification of microbial fed-batch culture with open loop glycerol input and pH logic control

In this paper, we study the fed-batch fermentation of glycerol by Klebsiella pneumoniae with open loop glycerol input and pH logic control using a flow of alkali as manipulated variable. A nonlinear hybrid system is developed to describe this process. We prove the finiteness of switching numbers of the system in terms of bounded variation and explore the existence and uniqueness of the solutions. Additionally, a parameter identification problem is proposed and an asynchronous parallel particle swarm optimization (PSO) algorithm is constructed to solve it. Numerical results show the effectiveness of the algorithm and reveal that the proposed model could describe the fed-batch process properly.     

Solving the semi-desirable facility location problem using bi-objective particle swarm

In this paper, a new model for the semi-obnoxious facility location problem is introduced. The new model is composed of a weighted minisum function to represent the transportation costs and a distance-based piecewise function to represent the obnoxious effects of the facility. A single-objective particle swarm optimizer (PSO) and a bi-objective PSO are devised to solve the problem. Results are compared on a suite of test problems and show that the bi-objective PSO produces a diverse set of non-dominated solutions more efficiently than the single-objective PSO and is competitive with the best results from the literature. Computational complexity analysis estimates only a linear increase in effort with problem size.     

基于混合粒子群优化算法的DV-Hop定位算法

针对无线传感器网络节点定位技术中DV-Hop算法的不足,利用混合粒子群优化算法对DV-Hop算法的位置估计进行校正,提出了一种CCPDV-Hop算法,该方法在不需要任何额外硬件设备和通信开销基础上,将未知节点定位问题抽象为高维最优化问题,并利用混合粒子群优化算法进行求解。仿真实验结果表明,改进的DV-Hop算法与传统方法相比,定位误差显著下降,定位精度和鲁棒性都有明显提高。     

基于粒子群优化的网络动态可靠性分析

时变性是节点和网络的典型特征。针对时变可靠性问题主要是基于上穿率法进行分析,传统的上穿率法分析方法通过时间离散,然后采用FORM法计算其瞬时可靠度,进而计算上穿率。对于功能函数非线性程度较高时,FORM求解精度降低。本文采用pso算法改进上穿率法,首先在离散时间之后,通过pso算法求解瞬时可靠度指标,然后计算上穿率进行时变可靠性分析,并通过实例与上穿率法及Monte-Carlo方法进行对比,本文提出的方法可以提高计算精度。     

基于RBF网络的断路器电流自适应保护研究

针对传统断路器电流保护方法存在受系统运行方式影响、整定困难、智能化低等问题,本文提出了基于RBF的断路器电流自适应保护算法,并给出了算法的模型。该算法融合了RBF神经网络的故障检测和电流自适应保护。首先通过RBF网络检测负载线路的电流故障,然后用电流自适应算法进行保护。在对神经网络进行训练时,利用PSO算法对RBF神经网络的参数进行优化以此来提高网络的泛化能力和学习能力;然后采用优化后的PSO-RBF神经网络对电流故障进行诊断。实验表明,该算法较大地提高了断路器智能化管理水平。