Echo state networks based predictive model of vinyl chloride monomer convention velocity optimized by artificial fish swarm algorithm

Conversion rate in the Polyvinylchloride (PVC) polymerizing process has a certain influence on the molecular weight of PVC, porosity, absorption rate of plasticizer, vinyl chloride monomer (VCM) residue and thermal stability. Therefore, a predictive model based on echo state networks (ESN) method optimized by the artificial fish swarm algorithm (AFSA) is proposed to predict the conversion velocity. Firstly, the hot balancing mechanisms of polymerizer and the influenced factors of convention rate of VCM are analyzed in details. Then the auxiliary variables of the predictive model kernel are selected by using the kernel principal component analysis method for reducing the model dimensionality. Thirdly, the structure parameters of the ESN are optimized by the AFSA to realize the nonlinear mapping between input and output variables of the discussed soft-sensor model. The artificial fish swarm behaviors, such as foraging, swarming, chasing, random, are introduced in details. Finally, simulation results show that the proposed model can significantly enhance the predictive accuracy and robustness of the technical index and satisfy the real-time control requirements of PVC polymerizing production process.     

基于神经网络的聚氯乙烯汽提过程自适应解耦控制

聚氯乙烯汽提过程具有高度非线性和时变性等特点,是一类复杂的非线性工业过程.首先基于动态模糊神经网络建立了数据驱动的聚氯乙烯树脂（PVC）汽提过程的被控对象模型;然后采用一种神经网络分散式解耦控制器对汽提过程进行解耦,得到浆料流量-塔顶温度和蒸汽流量-塔底温度两个单变量系统;最后采用BP神经网络PID控制器对系统进行控制.仿真实验结果验证了所提出集成控制策略的有效性.     

Dynamic fuzzy neural networks-a novel approach to functionapproximation

In this paper, an architecture of dynamic fuzzy neural networks (D-FNN) implementing Takagi-Sugeno-Kang (TSK) fuzzy systems based on extended radial basis function (RBF) neural networks is proposed. A novel learning algorithm based on D-FNN is also presented. The salient characteristics of the algorithm are: 1) hierarchical on-line self-organizing learning is used; 2) neurons can be recruited or deleted dynamically according to their significance to the system's performance; and 3) fast learning speed can be achieved. Simulation studies and comprehensive comparisons with some other learning algorithms demonstrate that a more compact structure with higher performance can be achieved by the proposed approach.     

Research on prediction of traffic flow based on dynamic fuzzy neural networks

Combining the advantages of the neural network and fuzzy system, this paper makes a further research on the dynamic fuzzy neural networks (D-FNN) traffic flow prediction. Instead of being in consistence with growth of the input number, the fuzzy rule number of the D-FNN increases exponentially in the whole training network structure. In particular, this method can establish a required network structure automatically. This method is applied to the traffic flow time series to analyze and compare the predicting performance of the predicting model based on the neural network method and the adaptive neural fuzzy inference system by combining with the chaos theory. The simulation result shows that this method is quite effective and can improve the predicting accuracy.     

基于修剪策略的D-FNN直接逆控制算法研究

神经网络是模拟人脑结构,它具有大规模并行及分布式信息处理能力,但是不能处理和描述模糊信息.模糊系统具有推理过程容易理解,但它很难实现自适应学习的功能.如果结合神经网络与模糊系统,可以取长补短.基于此,本文提出了一种新型动态模糊神经网络（Dynamic fuzzy neural network,D-FNN）学习算法.因为它具有结构和参数同时调整且学习速度快等优点,所以既可以在模糊逻辑系统中包含低级的神经网络学习和计算功能,也可以为神经网络提供高级的类似人的思维和推理的模糊逻辑系统.此外,本文还开发了生物医学工程应用算法程序,针对药物注射系统的直接逆控制案例进行了仿真,结果表明:D-FNN具有实时学习和控制能力强、参数估计和结构辨识同时进行等优点.     

基于D-FNN的开关磁阻无位置传感器的研究

提出了一种基于扩展径向基函数(RBF)神经网络的动态模糊神经网络(D-FNN)的开关磁阻电机无位置传感器控制的新方法。动态模糊神经网络系统以在线采样的相绕组的电流和磁链为输入,以转子位置角度为输出,从而建立起电流和磁链、转子位置角度的非线性映射关系;训练完成后,用D-FNN输出结果取代位置传感器角度信号,实现电机无位置传感器运行。仿真和实验结果表明:由D-FNN获得的角度信号和由位置传感器获得的角度信号相比误差小,电机能够准确换相,且输出转矩波动小,转速曲线平滑,电机在无位置传感器下运行良好。     

Genetic reinforcement learning through symbiotic evolution forfuzzy controller design

An efficient genetic reinforcement learning algorithm for designing fuzzy controllers is proposed in this paper. The genetic algorithm (GA) adopted in this paper is based upon symbiotic evolution which, when applied to fuzzy controller design, complements the local mapping property of a fuzzy rule. Using this Symbiotic-Evolution-based Fuzzy Controller (SEFC) design method, the number of control trials, as well as consumed CPU time, are considerably reduced when compared to traditional GA-based fuzzy controller design methods and other types of genetic reinforcement learning schemes. Moreover, unlike traditional fuzzy controllers, which partition the input space into a grid, SEFC partitions the input space in a flexible way, thus creating fewer fuzzy rules. In SEFC, different types of fuzzy rules whose consequent parts are singletons, fuzzy sets, or linear equations (TSK-type fuzzy rules) are allowed. Further, the free parameters (e.g., centers and widths of membership functions) and fuzzy rules are all tuned automatically. For the TSK-type fuzzy rule especially, which put the proposed learning algorithm in use, only the significant input variables are selected to participate in the consequent of a rule. The proposed SEFC design method has been applied to different simulated control problems, including the cart-pole balancing system, a magnetic levitation system, and a water bath temperature control system. The proposed SEFC has been verified to be efficient and superior from these control problems, and from comparisons with some traditional GA-based fuzzy systems.     

基于模糊子空间聚类的0 阶岭回归TSK 模糊系统

经典数据驱动型TSK模糊系统在利用高维数据训练模型时,由于规则前件采用的特征过多,导致规则的解释性和简洁性下降.对此,根据模糊子空间聚类算法的子空间特性,为TSK模型添加特征抽取机制,并进一步利用岭回归实现后件的学习,提出一种基于模糊子空间聚类的0阶岭回归TSK模型构建方法.该方法不仅能为规则抽取出重要子空间特征,而且可为不同规则抽取不同的特征.在模拟和真实数据集上的实验结果验证了所提出方法的优势.     

一种基于进化聚类的动态TSK模型建模方法

在一种进化聚类算法（ECM）的基础上提出了一种新的动态TSK模糊模型的建模算法,以往许多神经模糊模型都不适用于自适应在线学习,而文章模型能实时地调整模糊规则库及规则参数,具有较强的在线学习能力;仿真结果表明,该方法是有效的。     

A self-organization mining based hybrid evolution learning for?TSK-type fuzzy model design

In this paper, a self-organization mining based hybrid evolution (SOME) learning algorithm for designing a TSK-type fuzzy model (TFM) is proposed. In the proposed SOME, group-based symbiotic evolution (GSE) is adopted in which each group in the GSE represents a collection of only one fuzzy rule. The proposed SOME consists of structure learning and parameter learning. In structure learning, the proposed SOME uses a two-step self-organization algorithm to decide the suitable number of rules in a TFM. In parameter learning, the proposed SOME uses the data mining based selection strategy and data mining based crossover strategy to decide groups and parental groups by the data mining algorithm that called frequent pattern growth. Illustrative examples were conducted to verify the performance and applicability of the proposed SOME method.     

神经网络模糊规则抽取及其应用

从功能的观点出发,提出了一种基于粒子群优化算法的神经网络模糊规则抽取方法。该方法利用所要抽取模糊规则的表达形式,设计了规则的粒子三段表示方式,在粒子群算法优化过程中,采用两种更新方法,即离散化方法和连续化方法。该方法不依赖于具体的网络结构和训练算法,可以方便地应用于各种回归型神经网络。仿真实验表明,该方法可以抽取出保真度较高的符号规则。在实际应用中,采用模糊规则抽取算法,从丙烯腈反应器软测量模型中所得到的规则,提供了一种参数调节的指导性策略。     

基于级联网络的电力系统短期负荷预测

电力系统短期负荷预测对电力系统运行设计具有十分重要的意义。因此，在分析了电力负荷运行曲线的基础上，提出了一种基于级联模糊神经网络的预测模型。该模型采用基于神经网路理论的模糊模型参数辨识方法，很适合于复杂系统的模糊预测和控制。详细地对输入量的选择和学习算法进行了分析。实例表明，此方法具有町靠、鲁棒性好和快速等特点，优于神经网络电力负荷预报方法。     

基于参数调整的动态模糊神经网络算法

模糊逻辑与神经网络结合形成的模糊神经网络同时具有易于表达人类知识、存储与学习分布信息的优点,基于此,提出一种基于参数调整的动态模糊神经网络算法。采用扩展卡尔曼滤波器法将全局算法划分为线性和非线性部分,线性参数由最小二乘法和滤波器法决定,非线性参数由训练样本和启发式法直接决定,线性和非线性参数可进行实时更新。仿真结果表明,该算法能保证更简洁的结构和更短的学习时间。     

基于T-S模型二级倒立摆的逐级模糊控制

针对二级倒立摆系统,提出了一种逐级模糊控制策略.该种方法基于Takagi-Sugeno模型,利用分级思想,设计逐级模糊控制规则,实现二级倒立摆系统的稳定控制.仿真结果表明,该种模糊控制策略规则数较少,控制精度高,比较适合于解决控制快速、多变量、非线性不稳定系统时遇到的规则组合爆炸问题,具有较强的实用性.     

基于自组织模糊RBF网络的连续空间Q学习

针对连续空间下的强化学习控制问题,提出了一种基于自组织模糊RBF网络的Q学习方法．网络的输入为状态,输出为连续动作及其Q值,从而实现了“连续状态—连续动作”的映射关系．首先将连续动作空间离散化为确定数目的离散动作,采用完全贪婪策略选取具有最大Q值的离散动作作为每条模糊规则的局部获胜动作．然后采用命令融合机制对获胜的离散动作按其效用值进行加权,得到实际作用于系统的连续动作．另外,为简化网络结构和提高学习速度,采用改进的RAN算法和梯度下降法分别对网络的结构和参数进行在线自适应调整．倒立摆平衡控制的仿真结果验证了所提Q学习方法的有效性．     

Modeling and advanced control method of PVC polymerization process

Based on data driven modeling theory, PVC polymerization process modeling and intelligent optimization control algorithm is studied. Firstly, a multi-T–S fuzzy neural networks soft-sensing model combining the principal component analysis (PCA) and fuzzy c-means (FCM) clustering algorithm is proposed to predict the convention rate and velocity of Vinyle Chloride Monomer (VCM). The proposed hybrid learning algorithm utilizing the harmony search (HS) and least square method is used to adjust the model premise parameters and consequent parameters. Secondly, the generalized predictive control (GPC) algorithm of polymerizer temperature based on segmental affine is proposed. According to dynamic equation of polymerizer temperature deduced by heat balance mechanism, the segmental affine model is built by temperature and convention rate of the polymerizer. Then linear matrix inequality (LMI) method is used to design the controller. Finally, simulation results and industrial application show the validity and feasibility of the proposed control strategy.     

基于增强型算法并能自动生成规则的模糊神经网络控制器

给出了一种基于增强型算法并能自动生成控制规则的模糊神经网络控制器RBFNNC（reinforcements based fuzzy neural network comtroller）。该控制器能根据被控对象的状态通过增强型学习自动生成模糊控制规则,RBFNNC用于倒立摆小车平衡系统控制的仿真实验表明了该系统的结构及增强型学习算法是有效和成功的。     

Data-driven fuzzy clustering based on maximum entropy principle and PSO

To identify the optimum fuzzy rule base is the major difficulty in designing fuzzy model. To design optimum fuzzy rule base, which is traditionally achieved by tedious trial and error process, from numerical data, a novel data-driven fuzzy clustering method based on maximum entropy principle (MEP) and particle swarm optimization (PSO) is proposed. In this algorithm, the memberships of output variables are inferred by maximum entropy principle, and the centers of fuzzy rule base are optimized by PSO. Comparing with the method that designing fuzzy rule base only by PSO or other evolutionary computation methods, the number of parameters to be optimized decreased greatly, and the computation cost declined. To check the effectiveness of the suggested approach, three examples for modeling are examined comparing with the method only using PSO. The performance of the identified fuzzy models is demonstrated.     

多粒度融合的模糊规则系统图像特征学习

当前最流行的图像特征学习方法是深度神经网络,该类方法无需人工参与即可自动地通过特征学习提取高效的特征,用于分类识别等任务。然而,深度神经网络图像特征抽取方法目前也面临着诸多挑战,其有效性严重依赖大规模的数据,且通常被视为黑盒模型,解释性较差。针对上述挑战,以基于模糊规则推理的TSK模糊系统(TSK-FS)为基础,提出了一种适用于不同规模数据集且易于理解的特征学习方法——多粒度融合的模糊规则系统图像特征学习算法。该方法通过基于规则的TSK-FS抽取图像特征,因而特征学习过程是可以利用规则进行解释的。其次,多粒度扫描也使得其特征学习能力进一步提升。在不同规模的图像数据集上进行了充分的实验,实验结果表明该方法在图像数据集上具有较好的有效性。     

基于模糊分类的模糊神经网络辨识方法及应用

基于改进的T-S模型，提出一种自适应模糊神经网络模型(AFNN)，给出了网络的连接结构和学习算法。基于竞争学习算法的模糊分类器确定系统的模糊空间和模糊规则数，并得出每个样本对每条规则的适用程度。利用卡尔曼滤波算法在线辨识删的后件参数。AFNN结构简洁，逼近能力强，能够显著提高辨识精度，并且在线辨识的模糊模型简单有效。将该AFNN用于非线性系统的模糊辨识和化工过程连续搅拌反应器(CSTR)的建模中，仿真结果验证了该方法的有效性，表明该网络能够实现复杂非线性系统的建模，而且建模精度高、收敛速度快。可当作复杂系统建模的一种有效手段。