四旋翼ESO的RBF神经网络PID控制器研究

四旋翼飞行器因存在参数不确定性和环境干扰,会出现姿态不稳定的问题,而传统的PID控制对四旋翼的姿态稳定及机动性达不到控制需求.为此,提出了一种扩张状态观测器(ESO)的RBF神经网络PID控制器.首先,利用ESO的扩张特性和非线性函数对扰动进行估计和补偿,减少系统的误差;其次,将ESO对系统输出的估计值作为RBF神经网络的输入,使梯度信息更加精确,能够更好地优化增量PID的参数;最后,该神经网络的激励函数取高斯基函数,利用RBF神经网络的自适应性、自学习能力对模型控制参数进行调整.Matlab仿真实验表明,在未知干扰环境下,ESO的RBF神经网络PID控制器能够明显提高系统的抗干扰能力,且具有较小的超调量及较好的鲁棒性.     

基于RBF神经网络的半主动悬架滑模控制

针对悬架参数的不确定性,提出一种基于径向基函数(RBF)神经网络的滑模控制方法。根据滑模变结构控制理论设计了两自由度半主动悬架系统的滑模控制器,采用极点配置法确定滑模切换面参数,应用比例切换的控制方法和等速趋近率确定控制律,采用RBF神经网络优化算法优化了滑模控制器。运用MATLAB/simulink进行仿真,结果显示,与被动悬架相比,基于RBF神经网络的滑模半主动控制具有良好的控制效果,显著地改善了车辆的行驶平顺性。     

RBF神经网络在薄膜厚度控制系统中的应用

针对双向拉伸薄膜厚度控制系统中,经典的PID控制器难以达到理想控制效果的问题,设计一个基于RBF神经网络整定的PID控制器。该控制器由RBF网络对系统进行在线辨识,找到最佳控制参数,并将其反馈给PID控制器,从而实现控制器参数的在线调整。仿真结果表明,本控制器具有超调量小,响应速度快和自适应性强等优点,可有效提高双向拉伸薄膜厚度控制系统的性能。     

具有输入饱和的纯反馈系统有限时间跟踪控制

针对具有输入饱和及输出约束的非线性纯反馈系统,提出了有限时间自适应神经跟踪控制方法.利用有限时间控制理论、Barrier Lyapunov函数以及径向基函数(RBF)神经网络设计出新颖的虚拟和实际输入信号,解决了具有输入饱和及输出约束的非线性纯反馈系统的有限时间控制问题,同时,确保系统在满足输入饱和及输出约束的条件下,系统的输出在有限时间内跟踪上参考信号,并且系统的跟踪误差在有限时间内被限制在原点的小邻域内.最后,仿真实验阐明了所设计控制器的有效性.     

模糊神经网络在电厂给水加药系统中的应用

火电厂的给水加药系统具有时滞、非线性等特点,并且难以建立数学模型。模糊控制对非线性、时滞、难以建立数学模型等系统有较好的控制效果和鲁棒性。神经网络在自学习、自适应和容错性方面表现出了良好的效果。将这两种智能算法融合,使系统在更好利用经验规则实现较好控制效果的同时能根据环境的变化不断地学习,提高系统的自适应能力和鲁棒性。文章设计了基于T-S模糊神经网络的加药控制器,并通过仿真对比检验了T-S模糊神经网络静态性能和动态性能以及对于非线性和时滞的控制效果。     

采用自抗扰控制的三电机同步协调系统

自抗扰控制器自问世以来被用于多种应用场合,但其有着参数多,难整定的弊端,文中依据RBF神经网络结构简单、逼近能力强的优势提出了基于RBF神经网络的复合自抗扰参数调节器,利用RBF神经网络获得自抗扰参数的在线调整信息,且将其应用于三电机同步系统,经过实验验证,该调节器能实时调节自抗扰参数且能改善控制性能,具有实用性。     

粒子群RBF神经网络电力电子电路故障诊断

为检测和诊断电力电子电路中的故障,获得更高的诊断精确度,提出粒子群算法优化RBF神经网络的故障诊断方法.与基本RBF神经网络相比,粒子群RBF神经网络可以提高系统的收敛速度和精度.把通过特征提取获得的电力电子电路故障特征量作为神经网络的输入,利用训练好的粒子群优化后的RBF神经网络进行故障诊断.仿真结果表明,实际输出与期望输出基本吻合,具有良好的分类效果,能够提高诊断精确度,对于电力电子电路的故障诊断是一种有效的方法.     

基于MFO优化RBF神经网络的舵机控制与仿真

针对某型号火箭发动机摆动机构舵机系统在非线性时变复杂条件下传统的PID控制精度低、适应能力差的缺点,通过对RBF神经网络算法和MFO算法的研究,并结合传统PID控制器,设计了MFO-RBF神经网络控制器。通过对系统仿真可以看出,同传统PID相比舵机最大偏角反馈值从22.2°优化到19.78°相移滞后减小了9°,系统的频率响应大大提高,阶跃响应超调量明显减少,上升时间从27 ms减小到11 ms,并且大大减小了稳态误差。研究表明,和传统的PID相比,MFO优化的RBF神经网络PID在位置环的控制效果上有明显提升。     

基于自抗扰控制器的变风量空调解耦控制

变风量空调是一个多输入多输出的系统,由于具有多个回路,各个回路之间相互影响,具有强耦合、非线性和大时滞等特点。常规PID控制存在稳定性与精确性的矛盾。文中根据自抗扰控制原理,分析了变风量空调系统的耦合关系,通过扩张状态观测器实现对系统总扰动的估计和补偿,实现解耦控制。文中将其用于变风量空调末端的主回路温度控制系统。仿真实验结果表明,与PID控制相比,自抗扰控制器的超调量、时滞等得到了明显改善。     

基于自抗扰控制器的变风量空调解耦控制

变风量空调是一个多输入多输出的系统,由于具有多个回路,各个回路之间相互影响,具有强耦合、非线性和大时滞等特点.常规PID控制存在稳定性与精确性的矛盾.文中根据自抗扰控制原理,分析了变风量空调系统的耦合关系,通过扩张状态观测器实现对系统总扰动的估计和补偿,实现解耦控制.文中将其用于变风量空调末端的主回路温度控制系统.仿真实验结果表明,与PID控制相比,自抗扰控制器的超调量、时滞等得到了明显改善.     

基于神经网络的非线性预测在恒温系统中的应用

针对恒温系统的非线性、大延时等特点,提出了基于RBF神经网络的恒温系统的预测模型,采用非线性预测控制、引入模型误差项和改进后的PI控制器,并且在Matlab7.2/Simulink这个平台下构建了整个系统的仿真模型。仿真结果表明：所建立的模型系统具有良好的动静态性能且增强了系统的抗干扰性。     

Neural-network-based predictive learning control of ram velocity in injection molding

In this paper, we develop a predictive learning controller for ram velocity of injection molding based on neural networks. We first introduce a model of describing the injection molding, including the time horizon and the batch index. The feedback control plus biased function is proposed for controlling this plant. More specifically, a radial basis function (RBF) network is used to approximate the biased function based on the time horizon. The weights in the RBF are determined by a predictive control scheme based on the batch index. For this algorithm, relevant convergence is investigated. Simulation results reveal that the proposed control can achieve our claims.     

基于RBF神经网络的永磁同步伺服电机控制系统

针对永磁同步电机控制系统,建立其磁场定向控制数学模型。运用增量式数字PID的方法实现对PMSM的传统PID控制策略。在此基础上,借助RBF神经网络的学习能力,进行PID控制器参数的自适应整定,进一步改善PID控制器的性能。同时,为提高RBF网络性能,采用粒子群算法对网络进行优化。仿真表明,与传统PID控制比较,基于RBF的PID控制系统能提高PID控制器的性能,改善了PMSM控制系统的收敛速度和跟踪精度。     

基于动态神经网络在线辨识的PID控制

针对工业控制领域中复杂非线性时变系统和传统RBF神经网络辨识PID控制的不足,提出了一种基于聚类结合算法的动态RBF神经网络在线辨识PID自适应控制方法.通过优化的动态RBF辨识神经网络更好地描述了控制对象的动态行为,获得PID参数在线调整信息,实现系统的智能控制.仿真结果表明,与常规RBF神经网络辨识的PID控制方法相比该方法具有较高的控制精度,较快的系统响应,较强的适应性和鲁棒性.     

Dynamical neural network organization of the visual pursuit system

The central nervous system is a parallel dynamical system which connects sensory input with motor output for the performance of visual tracking. This paper applies elementary control system tools to extend dynamical neural network models to the visual smooth pursuit system. Observed eye position responses to target motions and characteristics of the plant (eye muscles and orbital mechanics) place dynamical constraints on the interposed neural network controller. In the process of constructing a model for the controller, we show two previous pursuit system models, using efference copy and feedforward compensation, are equivalent from an input-output standpoint. We introduce a controller model possessing a potentially highly parallel implementation and offer an example with supporting neural firing rate data. Changes in time delays or other system dynamics are expected to lead to compensatory adaptive changes in the controller. A scheme to noninvasively simulate such changes in system dynamics was developed. Actual physiologic data of adaptive responses to increased time delay is presented as an example of the utility of this parallel controller. Compensatory changes in our parallel controller model are easily predicted. These results suggest a productive interaction between neural network modeling, neurophysiology, and control systems engineering.     

Neural control of posture during small magnitude perturbations: effects of aging and localized muscle fatigue

This study investigated the effects of aging and localized muscle fatigue on the neural control of upright stance during small postural perturbations. Sixteen young (aged 18-24 years) and 16 older (aged 55-74 years) participants were exposed to small magnitude, anteriorly-directed postural perturbations before and after fatiguing exercises (lumbar extensors and ankle plantar flexors). A single degree of freedom model of the human body was used to simulate recovery kinematics following the perturbations. Central to the model was a simulated neural controller that multiplied time-delayed kinematics by invariant feedback gains. Feedback gains and time delay were optimized for each participant based on measured kinematics, and a novel delay margin analysis was performed to assess system robustness. A 10.9% longer effective time delay ( p = 0.010) was found among the older group, who also showed a greater reliance upon velocity feedback information (31.1% higher differential gain, p = 0.001) to control upright stance. Based on delay margins, older participants adopted a more robust control scheme to accommodate the small perturbations, potentially compensating for longer time delays or degraded sensory feedback. No fatigue-induced changes in neural controller gains, time delay, or delay margin were found in either age group, indicating that integration of this feedback information was not altered by muscle fatigue. The sensitivity of this approach to changes with fatigue may have been limited by model simplifications.     

基于时滞不确定Lorenz混沌同步的保密通信

针对一类含保密信息的时延不确定Lorenz混沌系统,提出了基于径向基神经网络的同结构同步控制方案,在混沌系统同步的基础上,有效地恢复出隐藏的多路明文信号。利用神经网络的良好逼近能力对时延Lorenz混沌系统设计鲁棒同步控制器,实现了同步误差的收敛。当同步误差收敛时,混沌系统所传输的隐藏信号则可以正确恢复。仿真结果表明,文章所给出的同步控制器可以在5s内实现时延不确定Lorenz混沌系统同步,并能恢复出多路明文信号。     

一种改进的RBF整定PID及其仿真实现

文中在分析RBF神经网络整定PID算法优缺点的基础上,给出了一种采用遗传模拟退火算法来优化网络结构和权值参数的RBF神经网络,将改进的RBF神经网络用于整定PID控制,并给出了相应的仿真测试例子。仿真实验结果表明,与采用梯度法优化网络权值等参数的RBF神经网络相比,给出的优化算法能更好地辨识控制系统,具有通用性好、调节精度高、在抑制超调量能力强等优点。     

Trajectory Switching Control of Robotic Manipulators Based on RBF Neural Networks

In this paper, we discuss the trajectory switching neural control problem for the switching model of a serial n-joint robotic manipulator. The key feature of this paper is to provide the dual design of the control law for the developed adaptive switching neural controller and the associated robust compensation control law. RBF Neural Networks (NNs) are employed to approximate unknown functions of robotic manipulators and a robust controller is designed to compensate the approximation errors of the neural networks and external disturbance. Via switched multiple Lyapunov function method, the adaptive updated laws and the admissible switching signals have been developed to guarantee that the resulting closed-loop system is asymptotically Lyapunov stable such that the joint position follows any given bounded desired output signal. Finally, we give a simulation example of a two-joint robotic manipulator to demonstrate the proposed methods and make a comparative analysis.