Global stability in switched recurrent neural networks with time-varying delay via nonlinear measure

In this paper, based on switched systems and recurrent neural networks (RNNs) with time-varying delay, the model of switched RNNs is formulated. Global asymptotical stability (GAS) and global robust stability (GRS) for such switched neural networks are studied by employing nonlinear measure and linear matrix inequality (LMI) techniques. Some new sufficient conditions are obtained to ensure GAS or GRS of the unique equilibrium of the proposed switched system. Furthermore, the proposed LMI results are computationally efficient as it can be solved numerically with standard commercial software. Finally, three examples are provided to illustrate the usefulness of the results.     

有限空间爆炸静态压力的温度补偿方法

为改善压阻式压力传感器的温度漂移特性,构建了基于遗传算法和小波神经网络的压力传感器温度补偿模型。针对小波神经网络收敛速度慢且易陷入局部最优解的问题,采用遗传算法对小波神经网络的连接权值、伸缩参数和平移参数进行优化。基于压力传感器的标定数据,分别采用BP神经网络、小波神经网络和遗传小波神经网络对其进行温度补偿研究,结果表明：遗传小波神经网络兼容了小波分析的时频局部特性和神经网络的自学习能力,表现出良好的收敛速度和补偿精度,经补偿后传感器的输出值更接近于标定值,其最大误差由−17.44 kPa变至0.38 kPa,最大相对误差由−14.0%变至0.38%。将该模型应用于有限空间爆炸静态压力的温度补偿中,取得了较好的实际应用效果。

     

STUDY ON PREDICTION METHODS FOR DYNAMIC SYSTEMS OF NONLINEAR CHAOTIC TIME SERIES＊

The prediction methods for nonlinear dynamic systems which are decided by chaotic time series are mainly studied as well as structures of nonlinear self-related chaotic models and their dimensions. By combining neural networks and wavelet theories, the structures of wavelet transform neural networks were studied and also a wavelet neural networks learning method was given. Based on wavelet networks, a new method for parameter identification was suggested, which can be used selectively to extract different scales of frequency and time in time series in order to realize prediction of tendencies or details of original time series. Through pre-treatment and comparison of results before and after the treatment, several useful conclusions are reached:High accurate identification can be guaranteed by applying wavelet networks to identify parameters of self-related chaotic models and more valid prediction of the chaotic time series including noise can be achieved accordingly.     

遗传优化小波神经网络在组合导航系统中的应用

实时性是组合导航系统的一个重要指标，而神经网络的优化学习问题是决定网络效率的关键技术。遗传优化小波神经网络不仅继承了小波分析良好的局部性及其神经网络的学习和推广能力，而且具有遗传算法全局寻优的特点，是多层前向神经网络学习的一种理想算法。将它应用于组合导航系统中并进行了仿真，结果表明，该算法能够根据实际情况自适应确定网络结构，实时性好，精度与常规方法相当。     

Dynamic control for permanent magnet synchronous generator system using novel modified recurrent wavelet neural network

Because permanent magnet synchronous generator (PMSG) system driven by permanent magnet synchronous motor (PMSM) based on wind turbine emulator (WTE) is a nonlinear and time-varying system with high complication, an accurate dynamic model of the PMSG system directly driven by WTE is difficult to establish for the linear controller design. In order to conquer this difficulty and improve the robustness of dynamic system, the PMSG system controlled by the online-tuned parameters of the novel modified recurrent wavelet neural network (NN)-controlled system is proposed to control output voltages and powers of controllable rectifier and inverter in this study. First, a closed-loop PMSM-driven system based on WTE is designed for driving the PMSG system to generate output power. Second, the rotor speeds of the PMSG, the voltages, and currents of the two power converters are detected simultaneously to yield maximum power output. In addition, two sets of the online-tuned parameters of the modified recurrent wavelet NN controllers in the controllable rectifier and inverter are developed for the voltage-regulating controllers in order to improve output performance. Finally, some experimental results are verified to show the effectiveness of the proposed novel modified recurrent wavelet NN controller for the power output of the PMSG system driven by WTE.     

Stability criteria for periodic neural networks with discrete and distributed delays

In this paper, the stability analysis problem is dealt with for a class of periodic neural networks with both discrete and distributed time delays. Both global asymptotic and exponential stabilities are considered. The existence of the periodic solutions of the addressed neural networks is briefly discussed. Then, by constructing different Lyapnuov--Krasovskii functionals and using some analysis techniques, several new easy-to-test sufficient conditions are derived, respectively, for checking the globally asymptotic stability and globally exponential stability of the delayed neural networks. These results are useful in the design and applications of globally exponentially stable and periodic oscillatory neural circuits for recurrent neural networks with mixed time delays. A simulation example is provided to demonstrate the effectiveness of the results obtained.     

Global exponential stability for quaternion-valued recurrent neural networks with time-varying delays

Nonlinear Dynamics - In this paper, we employ a novel method for solving the problem of the global exponential stability of quaternion-valued recurrent neural networks (QVNNs) with time-varying...     

无网格动力分析的循环卷积神经网络代理模型

在无网格动力分析中,除了无网格形函数本身构造复杂引入的计算成本,还需要逐步递推求解每个时间步的动力响应,因而计算效率较为低下.本文通过研究无网格离散数据与机器学习训练样本、无网格动力分析递推计算过程与循环卷积神经网络序列信息传递模式之间的本征联系,构建了与无网格法相匹配的循环卷积神经网络设计方法,进而提出了一种无网格动...     

弹性关节空间机械臂级联智能滑模控制

谐波减速器和力矩传感器等柔性元件因其独特性能而广泛应用在空间机器人关节系统中,以获取高减速比．但同时这些柔性元件的存在为空间机械臂系统引入了关节柔性,使得对其稳定控制变得更为复杂．基于此,文中讨论了基于自适应回归小波神经网络(Self-Recurrent Wavelet Neural Networks, SRWNN)的弹性关节空间机械臂系统动力学建模及级联智能滑模控制．首先,利用级联系统理论及第二类拉格朗日方法推导出了由外环刚性臂子系统和内环关节电机转子子系统组成的系统级联动力学模型;其次,为两个子系统分别设计了内、外环自适应滑模回归小波神经网络控制．外环控制算法以期望轨迹为控制量,而其控制信号作为抑制弹性关节振动的内环控制算法的控制量,整个控制系统由内、外环控制系统叠加而成;而后,基于Lyapunov稳定性理论证明了整个控制系统的稳定性并设计了自适应回归小波神经网络的各权值参数在线学习算法．所提的控制算法有效地消除了模型不确定的影响,避免了复杂的求导计算和角加速度可测的要求,同时,控制系统设计过程中未涉及惯常奇异摄动双时标分解操作,在理论上适合任意大小的关节柔性刚度．最后,系统对比仿真试验证明了所提的级联智能控制算法优于惯常基于奇异摄动法和基于柔性铰补偿奇异摄动法的控制方案．     

The Representational Relation Between Environmental Structures and Neural Systems: Autonomy and Environmental Dependency in Neural Knowledge Representation

In this paper it will be shown that in neural systems with a recurrent architecture, the traditional concepts of knowledge representation cannot be applied any more; no stable representational relationship of reference can be found. That is why a redefinition of the relationship between the states of the environment and the internal representational states is proposed. Studying the dynamics of recurrent neural systems reveals that the goal of representation is no longer to map the environment as accurately as possible to the representation system (e.g., to symbols). It is suggested that it is more appropriate to look at neural systems as physical dynamical devices embodying the (transformation) knowledge for sensorimotor integration and for generating adequate behavior enabling the organism's survival. As an implication the representation is determined not only by the environment, but highly depends on the organization, structure, and constraints of the representation system as well as the sensory/motor systems which are embedded in a particular body structure. This leads to a system relative concept of representation. By transforming recurrent neural networks into the domain of finite automata, the dynamics as well as the epistemological implications become more clear. In recurrent neural systems a type of balance between the autonomy of the representation and the environmental dependence/influence emerges. This not only affects the traditional concept of knowledge representation, but has also implications for the understanding of semantics, language, communication, and even science.     

Stimulus-induced sequential activity in supervisely trained recurrent networks of firing rate neurons

Nonlinear Dynamics - In this work, we consider recurrent neural networks of firing rate neurons supervisely trained to generate multidimensional sequences of given configurations. We study...     

A fuzzy wavelet neural network-based approach to hypersonic flight vehicle direct nonaffine hybrid control

A direct nonaffine hybrid control methodology is proposed for a generic hypersonic flight models based on fuzzy wavelet neural networks (FWNNs). The addressed strategy extends the previous indirect nonaffine control approaches stemming from simplified models of affine formulations. To cope with nonaffine effects on control design, analytically invertible models are constructed and then novel hybrid controllers are developed directly using nonaffine models. Furthermore, by employing FWNNs to devise adaptive terms, inversion errors are canceled via fuzzy neural approximations. In addition, robust terms are designed to achieve larger stable region in comparison with earlier work using Lyapunov synthesis. Finally, numerical simulation results from a hypersonic flight vehicle model are given to clarify the efficiency of the proposed direct nonaffine control scheme in the presence of parametric uncertainties.     

基于神经网络的偏微分方程求解方法研究综述

神经网络作为一种强大的信息处理工具在计算机视觉,生物医学,油气工程领域得到广泛应用,引发多领域技术变革.深度学习网络具有非常强的学习能力,不仅能发现物理规律,还能求解偏微分方程.近年来基于深度学习的偏微分方程求解已是研究新热点.遵循于传统偏微分方程解析解、偏微分方程数值解术语,本文称用神经网络进行偏微分方程求解的方法为...     

基于转角模态和小波神经网络的连续梁损伤识别研究

基于小波奇异性检测原理和神经网络非线性映射能力,结合结构基本模态参数,提出了一种结合小波神经网络与结构转角模态的损伤识别方法．首先,建立三跨连续梁的有限元模型获取结构模态参数,并对其进行Mexihat小波变换,通过系数图突变点判断结构损伤位置．然后,将小波系数模特征向量作为BP神经网络的输入,分别研究了该方法在单损伤和多损伤工况下的识别能力．最后将不同工况下神经网络预测值与结构实际损伤程度进行对比,得到单处损伤预测误差平均值为0.22%,多处损伤预测误差平均值分别为0.22%和0.18%,结果表明该方法在结构损伤识别方面的有较高有效性及精确度．     

Robust synchronization of uncertain fractional-order chaotic systems with time-varying delay

This paper presents a new technique using a recurrent non-singleton type-2 sequential fuzzy neural network (RNT2SFNN) for synchronization of the fractional-order chaotic systems with time-varying delay and uncertain dynamics. The consequent parameters of the proposed RNT2SFNN are learned based on the Lyapunov–Krasovskii stability analysis. The proposed control method is used to synchronize two non-identical and identical fractional-order chaotic systems, with time-varying delay. Also, to demonstrate the performance of the proposed control method, in the other practical applications, the proposed controller is applied to synchronize the master–slave bilateral teleoperation problem with time-varying delay. Simulation results show that the proposed control scenario results in good performance in the presence of external disturbance, unknown functions in the dynamics of the system and also time-varying delay in the control signal and the dynamics of system. Finally, the effectiveness of proposed RNT2SFNN is verified by a nonlinear identification problem and its performance is compared with other well-known neural networks.     

神经网络混沌运动的控制--CM方法

基于压缩映射的混沌控制方法——CM方法被应用到小的离散神经网络,通过一个外部输入的小干扰,稳定混沌轨道嵌入在混沌吸引子内的某一不稳周期轨上。利用闭回路对技术估计欲稳定周期轨的近似位置。给出二维和三维神经网络的典型例子,通过数值模拟显示CM方法控制离散神经网络混沌行为的简单和有效性。     

Exponential state estimation for delayed recurrent neural networks with sampled-data

In this paper, the sampled-data state estimation problem is investigated for a class of recurrent neural networks with time-varying delay. Instead of the continuous measurement, the sampled measurement is used to estimate the neuron states, and a sampled-data estimator is constructed. By converting the sampling period into a bounded time-varying delay, the error dynamics of the considered neural network is derived in terms of a dynamic system with two different time-delays. Subsequently, by choosing an appropriate Lyapunov functional and using the Jensen??s inequality, a sufficient condition depending on the sampling period is obtained under which the resulting error system is exponentially stable. Then a sampled-data estimator is designed in terms of the solution to a set of linear matrix inequalities (LMIs) which can be solved by using available software. Finally, a numerical example is employed to demonstrate the effectiveness of the proposed sampled-data estimation approach.     

结构损伤的小波分形神经网络检测

用神经网络进行结构损伤检测、分析的有效性在很大程度上取决于训练样本的好坏。小波变换在时域和频域都具有表征信号局部特征的能力,小波包分析利用可以伸缩和平移的可变视窗能够聚焦到信号的任意细节,因此对有损伤的结构的非线性动力特性能进行有效的分析。利用分形几何方法不依赖于系统的数学模型的特点,将分形维数与小波分析相结合,建立了结构损伤的小波分形神经网络检测方法。研究结果表明,结构不同状态下的振动信号的各频段分形维数有明显的不同,可以将振动信号的各频段分形维数作为结构损伤检测的特征量,并用神经网络将结构的不同状态模式识别出来。     

基于前向线性预测算法的光纤陀螺零漂的神经网络建模

在详细分析光纤陀螺零漂的基础上,提出了先用滤波算法对光纤陀螺信号进行预处理,然后采用RBF神经网络对滤波后的信号进行建模的方法。针对光纤陀螺信号特点分别采用FLP算法、小波滤波算法、解相关变步长LMS自适应滤波算法对其进行了预处理,比较三种滤波方法,小波滤波算法效果优于其它两种预处理方法,但针对基于预处理后的陀螺信号采用RBF神经网络进行建模时,小波滤波预处理后的信号在建模精度上却是最差的,而对FLP算法滤波后的信号进行RBF建模,建模精度提高了两个数量级。结果表明:基于FLP算法的RBF神经网络在光纤陀螺中的建模是有效的,可大大提高建模的精度。     

地震作用下高层建筑结构的分散神经网络振动控制研究

针对地震作用下高层建筑振动神经网络控制问题,将神经网络理论与分散控制理论相结合,提出分散神经网络振动控制方案,并应用于高层结构地震反应振动控制中。利用多层前馈神经网络建立结构模型,预测结构的振动响应。基于NARMA-L2的神经自校正控制系统设计BP神经网络控制器,研究分散神经网络振动控制效果,并与神经网络集中控制进行比较。对某20层Benchmark结构模型进行数值模拟分析,结果表明,本文提出的分散神经网络振动控制方法简化了神经网络的结构,可有效控制结构振动和消除时滞;同时,相对于集中控制的单一失效,本文方法的可靠性更强且可以保证振动控制系统的实时响应。