系统可靠性冗余最优分配问题的神经网络解法

本文提出了一种基于神经网络的系统可靠性冗余最优分配方法。模拟计算结果表明，本文提出的方法是实用和有效的。     

神经网络和证据理论融合的故障诊断方法研究

针对传统故障诊断方法存在的诊断准确率不高的问题，利用数据融合原理，将神经网络和证据理论进行有机结合，使两者优势互补，提出了神经网络和证据理论融合的故障诊断方法。通过简化网络结构提高了局部诊断网络的诊断能力，并使证据理论的基本可信度分配不再完全依赖专家进行的主观化赋值，从而实现了赋值的客观化。充分利用各种故障的冗余和互补信息，显著提高了故障诊断的准确率。在材料试验机伺服系统的液压泵上进行了模拟故障诊断试验，验证了该方法的有效性。     

基于小波神经网络信息融合动密封故障诊断研究

针对动密封中唇形密封的综合故障问题，应用小波神经网络技术加混合式数据融合方法，将数据级、特征级和决策级故障诊断数据融合在一起，对动密封中唇形密封的缺陷进行智能诊断，描述了小波神经网络的建模过程，探讨了通过多源互补信息减少故障诊断系统不确定性的优化方法。结果表明，采用混合式融合结构可以通过多源互补以及冗余信息来提高诊断系统的鲁棒性；采用小波神经网络信息融合的诊断方法，可有效诊断动密封中唇形密封存在的综合故障问题。     

冗余约束绳驱并联机器人的张力分配研究

提出一种改进的二次规划方法求解冗余约束绳驱并联机器人的绳索张力分配问题.建立一般冗余约束绳驱并联机器人运动学模型,对机构进行受力分析,得到一般冗余约束绳驱并联机器人的动力学方程.结合力封闭的思想,引入参考力的概念,改进传统的二次规划方法,得到了适用于冗余约束绳驱并联机器人的绳索张力分配优化模型.以八绳六自由度绳驱并联机...     

基于RBF网络的冗余扫查机器人逆运动学研究

利用径向基函数（RBF）神经网络逼近能力较优和收敛速度较快的特点将运动学逆解过程转换为神经网络权值的训练过程。在训练RBF网络时，采用正交最小二乘（OLS）算法来确定网络中心，并将正解结果作为训练样本，实现了冗余扫查机器人运动学逆解的计算，避免了传统方法的繁琐公式推导及数值病态问题。     

柔性机器人协调操作的冗余驱动

针对具有柔性关节和柔性臂的机器人协调操作刚性负载时,其逆动力学问题存在冗余驱动这种情况,提出了一种载荷分配方法。将这种分配方法与基于绝对坐标并以期望轨迹为实际边界条件的逆动力学模型相结合,可以有效地解决冗余驱动问题。文中就两个具有柔性关节和柔性臂的３Ｒ机器人协调操作刚性负载为例,验证了本方法的有效性。     

基于粗集-神经网络的智能混合压缩机故障诊断系统的研究

利用粗集的约简功能消除压缩机故障样本数据中冗余信息，将粗集与神经网络相结合，构建了一个基于粗集一神经网络的智能混合压缩机故障诊断系统，实现了粗集对神经网络在压缩机故障诊断中的优化。     

基于mRMR的发电机DCS信号神经网络趋势预测方法

针对神经网络方法在发电机信号趋势预测过程中如何合理选择网络输入节点的问题,提出了一种基于最大相关和最小冗余(mRMR)算法的神经网络输入信号选取准则。该方法研究了发电机分布式控制系统(DCS)监测数据的特点,采用了mRMR算法从原始特征集合中选择了与被描述对象具有最大相关性,且特征集元素间冗余量最小的特征子集作为网络输入,进而有效地提高了网络模型对输入输出间非线性函数关系的拟合精度。研究结果表明,当对某电厂DCS信号进行分析时,与直接利用神经网络进行趋势预测的准确性相比,该方法预测准确性高、泛化能力好,具有良好的工程适用性。     

基于BP神经网络的冗余机械臂逆运动学分析

常见的七自由度冗余机械臂逆运动学解法较为繁杂,且对不同构型机械臂的逆运动学解法通用性较差。为了找到一种通用的七自由度机械臂逆运动学求解方法,建立了神经网络模型,选择了合适的激励函数、隐藏层神经元数量、神经网络层数和学习速率等模型参数。设计轨迹跟踪实验对神经网络模型进行验证,实验数据证明该方法有效且精度较高,是一种可行的冗余机械臂逆运动学求解方法。     

基于遗传算法优化的神经网络及其在机械故障诊断中的应用

针对神经网络结构的设计基本上依赖于人的经验的缺点,提出了一种利用遗传算法优化神经网络的算法。该算法结合了神经网络的快速并行性和遗传算法的全局搜索性,同时对神经网络的连接权和结构进行了优化,剔除整个网络冗余节点和冗余连接权,提高了网络的处理能力。将该方法应用于磨机故障诊断,能够对故障模式进行精确的识别, 取得了很好的效果。     

An emotional learning-based fuzzy inference system for improvement of system reliability evaluation in redundancy allocation problem

A major challenge in reliability based optimal design problems is the evaluation of system reliability given a system design. Reliability evaluator is a tool, a model, or a system that analyses the reliability of a system given a specified system design. Reliability evaluator usually is embedded in reliability optimization model and it can be seen as a computational engine that can provide optimization model with the value of its objective function. Enhancing performance of the reliability evaluator in optimization models is important as the overall performance of optimization model is significantly affected by its embedded reliability evaluator. The purpose of this paper is to present an approach to improve the accuracy and generalization ability of evaluating system reliability in redundancy allocation problem. The main idea is to employ emotional learning-based fuzzy inference system (ELFIS) to improve performance of reliability evaluator. A series–parallel case reliability–redundancy allocation is considered and the proposed ELFIS is validated by comparison of its results with those of multi-layer perceptron neural network and adaptive network-based fuzzy inference system. Normalized mean squared error and mean absolute percentage error show that ELFIS can bring better accuracy in evaluating the system reliability for those reliability-based optimal designs.     

基于复杂网络理论的多项目实施资源优化配置研究

针对多项目实施过程中的资源优化配置问题,首先建立了基于复杂网络理论的多项目实施资源配置复杂网络模型,并结合Pajek分析软件对模型参数进行分析研究;其次提出了资源受限下的多项目实施资源配置算法;最后以多项目实施中资源配置为例,结合资源优化配置算法,以资源配置效益最大化为目标函数,对资源受限下的多项目实施资源优化配置问题进行研究,验证了该方法的可行性和有效性.     

粒子群及局部搜索算法在串并联系统结构优化中的应用

将粒子群算法和局部搜索算法相结合,用于解决串并联系统的冗余分配问题。介绍串并联系统结构,确定该冗余分配问题以系统最小费用为优化目标,以系统可用度不能低于某一确定值为约束条件。应用概率生成函数(UGF)方法计算系统可用度,将粒子群算法和局部搜索算法相结合进行优化求解,并给出迭代过程。通过实例对优化迭代过程进行具体说明。     

A new fuzzy optimal allocation of detuned passive filters based on a Nonhomogeneous Cuckoo Search Algorithm considering resonance constraint

Passive filters are used as one of the effective solutions to mitigate harmonics and improve power quality in electrical networks. In this paper, a new fuzzy approach is proposed for the allocation of detuned passive filters based on a Nonhomogeneous Cuckoo Search Algorithm (NoCuSa). In this method, a resonance index is inserted in the problem formulation to avoid being in a resonance condition after the allocation of passive filters. In this regard, the candidate locations for the installation of passive filters are first selected based on a sensitivity analysis. Then, the values and tuning orders of the passive filters are optimized by using the proposed algorithm for single- and multi-load levels while applying fixed and switched passive filters. In the simulations, different scenarios for optimal allocation of passive filters are investigated and compared with optimal allocation of capacitors. Finally, the fuzzy problem model is implemented on an IEEE 69-bus network by NoCuSa and compared with different optimization algorithms. The results demonstrate an improvement in the final annual net benefit by applying NoCuSa in comparison with other algorithms. In addition, another comparison made between the proposed method and those implemented on IEEE 69-bus system shows the efficiency of the proposed method.     

Bi-objective redundancy allocation problem for a system with mixed repairable and non-repairable components

Traditionally, in the redundancy allocation problem (RAP), two general classes of optimization problems are considered; reliability optimization and availability optimization. Contrary to reliability optimization, fewer researchers have studied availability optimization to find out the optimal combination of components type and redundancy levels for each subsystem in a system for maximizing (or minimizing) the objectives. In each problem it is assumed that either the entire components are repairable or they are non-repairable. However, in real world situations, systems usually consist of both repairable and non-repairable components. In this paper a new Mixed Integer Nonlinear Programming (MINLP) model is presented to analyze the availability optimization of a system with a given structure, using both repairable and non-repairable components, simultaneously. To find the solution of the introduced MINLP, an efficient Genetic Algorithm (GA) is also developed. Furthermore, to show the efficiency of the proposed GA, a numerical example is presented. Experimental results demonstrate that the proposed GA has a better performance compared to one of the most recommended algorithm in the literature.     

Output-feedback super-twisting control for line-of-sight angles tracking of non-cooperative target spacecraft

Without the line-of-sight (LOS) angles rate information, this paper investigates the LOS angles tracking problem of non-cooperative target in chaser’s body frame with the external disturbance force and torque via chaser’s control torque. By integrating with the attitude dynamics of the chaser, a novel coupled LOS-based relative motion model is firstly established, which reveals the redundancy relationship between the LOS angles motion with two Degree-of-Freedom (DOF) and three dimensional control torque. More specially, the LOS angles tracking control problem is formulated as an output-feedback control problem of an uncertain nonlinear system with the actuator redundancy. As a stepping-stone, a fourth order high order sliding mode observer (HOSMO) is proposed to estimate the system state and uncertain terms. A combination of modified super twisting algorithm (STA) with nonsingular fast terminal sliding mode (NFTSM) and control allocation is proposed, the main novelty of modified STA is that the NFTSM is introduced to replace the linear sliding mode (LSM), and the original STA cannot be applied directly, a modified STA is proposed, which can guarantee the fast finite-time convergence. Finally, simulations are conducted to show fine performance of the proposed control scheme.     

产品回收多级逆向物流网络优化设计模型

为降低逆向物流成本,改善客户产品回收便利性,优化逆向物流网络中产品回收的渠道选择决策,在原有正向物流网络基础上,提出了包含初始收集点、集中回收中心和回收处理工厂三种物流设施的多级逆向物流网络.以总逆向物流成本最小为目标,兼顾各客户群的产品回收便利性,建立了0-1混合整数规划模型,实现逆向物流网络的选址/分配问题.应用遗传算法求解该模型.模拟算例表明,该算法有效且可行,为企业进行逆向物流网络设计提供可参考的模型和求解算法.     

遗传算法在串-并联冗余系统可靠性优化设计中的应用

串-并联冗余系统可靠性的最优设计是可靠性技术中一个十分重要的问题。讨论了遗传算法及其在串-并联冗余系统可靠性优化设计中的应用。串-并联冗余系统具有冗余单元和可选设计，其可靠性优化设计是一高度非线性、具有多极值点、有约束的整数规划问题，传统的解决方法往往对于连续实数域有效可行，而对于解决此类多变量，离散型整数域取值的复杂的优化问题则往往很困难。由于遗传算法是一类全局收敛算法，它用点群进行寻优，而不是用一个单点进行寻优，具有隐含并行算法的特点，从而能够得到全局最优解。所给出的实例说明本文的算法是有效可行的。