基于正则化 AdaBound 的区间二型模糊 神经网络软测量建模

针对复杂化工过程中存在强非线性、多变量耦合、参数时变及大时滞等因素,导致监测变量软测量精度不高的问题,提 出了一种基于正则化 AdaBound 的区间二型模糊神经网络(RAIT2FNN) 软测量建模方法。 首先为了解决区间二型神经网络 (IT2FNN)结构难以确定的问题,提出了一种采用激励强度和相似度定义增长和删减指标的自组织产生规则的算法。 该算法利 用激励强度的大小决定是否产生规则,并根据相似度进行规则的删减从而确定了区间二型模糊神经网络的结构。 其次,本文提 出正则化和 AdaBound 相结合的算法对 RAIT2FNN 模型相关参数进行修正,使得不同参数具有有界的自适应学习速率。 最后将 RAIT2FNN 作为软测量模型应用于环己烷无催化氧化过程尾氧浓度预测问题中。 实验结果为测试时间为 0. 008 2,训练 RMSE 为 0. 018 2,测试 RMSE 为 0. 009 6,表明 RAIT2FNN 作为软测量模型具有预测及时且预测精度较高的优点。     

非线性广义预测控制算法及其仿真研究

为了实现对非线性广义预测控制算法的仿真研究,采用神经网络建立非线性预测模型,在其工作点附近线性化,并对线性化模型进行广义预测控制,利用单片输出BP网络所辨识的非线性模型进行分析,提出了一种基于非线性广义预测前馈补偿控制律的补偿算法,改善了线性化所带来的模型失配误差.最后通过Matlab进行了仿真.仿真结果验证了算法的有效性.     

基于BP神经网络的离心泵的性能预测

针对离心泵性能实测过程繁琐费用昂贵以及鉴于神经网络在非线性系统映射的优越性,提出利用神经网络中的BP网络模型来预测离心泵的性能,并用Matlab7对一系列的离心泵进行性能预测.预测结果表明:神经网络可以实现对离心泵性能的精确预测并用于实际应用.     

一种高速数控机床电主轴表面温度智能预测方法

电主轴是高速数控机床核心功能部件,电主轴损坏基本是电主轴发热引起的.电主轴温度场具有复杂的非线性特征,神经网络在处理非线性系统温度预测方面得到了广泛的研究,神经网络与传统模型相比具有更好的适时预报性和持久性.论文利用遗传算法优化BP神经网络建立电主轴表面温度预测模型.预测结果表明,未优化的BP神经网络与遗传神经网络预测误差相对比,遗传神经网络对电主轴表面温度预测具有更高的预测精度和稳定性.     

基于人工神经网络研究油田产量预测方法

准确预测油田未来原油产量对油田的开发和生产管理具有重要的指导意义.对于具有非线性、不确定和开放特性的多变量系统进行预测,使用传统的统计学方法或静态模型预测通常不能满足精度要求.这里把BP神经网络、模糊神经网络以及基于神经网络的组合预测方法应用于多变量系统.采用了两层预测系统第一层包含两个神经网络;第二层是把第一层的两个网络输出进行组合.研究了3种不同的组合算法平均法、最小平方回归法和前馈神经网络法.实验结果表明,采用组合方法比采用单一的预测方法的预测精度有了进一步的提高,特别是应用人工神经网络(即BPNN)的组合预测优于其他的预测方法,具有较好的适应性.     

基于组合模型的滚动轴承故障预测研究

针对轴承故障预测可使用的样本数据少、特征参数信息贫乏且呈现非线性、不确定性等特点,提出一种基于改进灰色GM(1,1)和遗传算法优化的BP神经网络的组合预测模型。首先,根据各单一模型在当前时段的预测误差,通过最小二乘法确定出在未来时段中两种单一模型的权重,然后将预测结果进行加权求和,得到最终的组合模型预测值。该模型既能实现灰色GM(1,1)模型处理小样本的轴承振动数据预测的目标,也能发挥BP神经网络解决非线性拟合问题的优势。最后,将组合模型与各单一模型进行实例数据分析,结果表明组合模型的预测精度为96.63%,比上述子模型的预测结果分别提高了7.84%和6.13%。     

基于BP神经网络的磨削温度预测

磨削温度是评价磨削过程的一个重要指标,利用BP神经网络良好的非线性映射功能,以磨削用量(砂轮线速度、工作太速度和磨削深度)为输入,以磨削温度为输出,建立了磨削温度的BP神经网络预测模型。并通过仿真验证了模型的正确性,为磨削温度的预测提供了一个简单可行的方法。     

BP神经网络在双燃料发动机排放预测中的应用

运用BP神经网络(Back Propagation Network)的自学习以及非线性逼近能力,对双燃料发动机排气中CO、HC、NO_x和碳烟的浓度进行拟合和预测。搭建神经网络模型,通过采集双燃料发动机排气浓度数据对神经网络模型进行训练和验证。当BP神经网络训练过程中样本和模型计算值的线性相关系数R大于0.9,且用于验证的数据和模型运算值误差在可忽略范围内,则所建的神经网络模型能够有效预测双燃料发动机的排气浓度。训练结果显示,CO、HC、NO_x和碳烟浓度的模型计算值和实测值线性相关系数R都大于0.9,说明神经网络具有较强的拟合能力;验证结果显示,预测值和实测值的相对平均误差都小于10%,能够满足实际需求。结果表明,运用神经网络模型能够有效预测双燃料发动机的排放。     

经济视角下质量预测的BP神经网络优化方法研究

对过程质量利用前馈(BP)神经网络进行预测,通过对网络参数和训练样本的优化来克服BP神经网络的缺陷.仿真结果表明网络的预测性能具有较好的可信度,而且较之开发新的网络更为成熟,同时能降低质量控制成本.     

基于径向基函数神经网络的发动机磨损预测分析

针对BP神经网络算法的不足,利用径向基函数(RBF)神经网络建立设备的磨损预测模型,对光谱分析数据进行实例仿真,并与BP网络模型进行对比研究.仿真结果表明,该模型预测精度高,训练时间短,大大优于BP神经网络模型.     

基于NARMAX模型的阀控非对称缸神经网络预测控制

针对非对称缸位置跟踪控制精度较差,提出了一种基于非线性自回归平均滑动离散模型(NARMAX)和量子粒子群算法的神经网络预测控制策略(QPSO-NNMPC)。利用NARMAX模型表示阀控非对称缸的动态模型,使用粒子群算法优化BP神经网络(PSO-BP)对阀控非对称缸系统在线预测,使用量子粒子群算法(QPSO)对目标函数非线性优化。仿真结果表明,在不同频率期望信号与变干扰力情况下,该控制策略具有良好的跟踪效果和鲁棒性。     

微粒群优化倒立摆神经网络控制器设计研究

将微粒群算法和多层前馈神经网络相结合,提出了一种利用微粒群算法代替BP算法训练多层前馈神经网络权值,以实现神经网络控制的方法,并对非线性模型的辨识问题和一级直线倒立摆的控制问题进行了仿真研究。仿真实验表明：微粒群算法在神经网络控制及非线性模型辨识方面效果良好,具有良好的应用前景。     

基于BP神经网络的SU-8光刻胶工艺参数优选研究

SU-8是一种性能优异的厚胶,广泛应用于高深宽比的MEMS微结构中。本文首先用正交试验研究了前烘时间、曝光剂量、后烘时间以及显影时间对SU-8光刻胶图形尺寸精度的影响,得到了优化的工艺组合。在此基础上,运用BP神经网络对试验数据进行分析处理,预测了较正交试验分析结果更为优化的工艺组合,并用试验验证了其正确性。结果表明,经正交试验数据训练过的BP神经网络,很好地映射了工艺参数与优化指标之间的复杂非线性关系,此时应用BP神经网络对工艺参数进行优选研究能够得到更全面、准确的结果。     

Internal leakage rate prediction and failure diagnosis of buried pipeline ball valve based on valve cavity pressure detection

— Ball valve is a key fluid control equipment used extensively in oil and gas pipelines. The online detection and failure diagnosis of the internal leakage of the ball valve is of great significance to ensure the safety operation of natural gas transmission pipelines. This paper proposes a prediction method of the internal leakage rate and a diagnosis method of the failure mode of the buried pipeline ball valve based on valve cavity pressure detection. Firstly, the valve cavity pressure signal generated by the internal leakage of the ball valve is detected by the pressure sensor, and the valve cavity pressure signal is denoised by wavelet threshold denoising. Then, the back propagation (BP) neural network has the disadvantage of unstable learning ability, so the BP neural network is optimized by chaos sparrow search optimization algorithm (CSSOA-BP). Finally, the prediction model of the ball valve internal leakage rate and the diagnosis model of the ball valve failure mode are established by using CSSOA-BP neural network and the characteristic parameters of the valve cavity pressure signal. To verify the performance of the prediction model and the diagnosis model of CSSOA-BP neural network, the predictive results and diagnostic results are compared with those of the sparrow search algorithm optimization BP (SSA-BP) neural network and BP neural network. The experimental results show that the maximum prediction error of CSSOA-BP neural network is the smallest, which is 13.6%. The accuracy of the diagnostic results of CSSOA-BP neural network is the highest, which is 83.3%. It indicates that the proposed method can achieve better predictive results of the ball valve internal leakage rate and more accurate diagnostic results of the ball valve failure mode.     

A QP Artificial Neural Network inverse kinematic solution for accurate robot path control

In recent decades, Artificial Neural Networks (ANNs) have become the focus of considerable attention in many disciplines, including robot control, where they can be used to solve nonlinear control problems. One of these ANNs applications is that of the inverse kinematic problem, which is important in robot path planning. In this paper, a neural network is employed to analyse of inverse kinematics of PUMA 560 type robot. The neural network is designed to find exact kinematics of the robot. The neural network is a feedforward neural network (FNN). The FNN is trained with different types of learning algorithm for designing exact inverse model of the robot. The Unimation PUMA 560 is a robot with six degrees of freedom and rotational joints. Inverse neural network model of the robot is trained with different learning algorithms for finding exact model of the robot. From the simulation results, the proposed neural network has superior performance for modelling complex robot’s kinematics.     

Soft sensor modeling of chemical process based on self-organizing recurrent interval type-2 fuzzy neural network

This study introduces a novel self-organizing recurrent interval type-2 fuzzy neural network (SRIT2FNN) for the construction of a soft sensor model for a complex chemical process. The proposed SRIT2FNN combines interval type-2 fuzzy logic systems (IT2FLSs) and recurrent neural networks (RNNs) to improve the modeling precision. The Gaussian interval type-2 membership function is used to describe the antecedent part of the SRIT2FNN fuzzy rule, and the consequent part is of the Mamdani type with an interval random number. An adaptive optimal clustering number of fuzzy kernel clustering algorithm based on a Gaussian kernel validity index (GKVI-AOCN-FKCM) is developed to determine the structure of the SRIT2FNN and fuzzy rule antecedent parameters, and the parameter learning of SRIT2FNN used the gradient descent method. Finally, the proposed SRIT2FNN is applied to the soft sensor modeling of ethylene cracking furnace yield in a typical chemical process. Comparisons between the SRIT2FNN and conventional fuzzy neural network (FNN) and interval type-2 fuzzy neural network (IT2FNN) are made via simulation experiments. The results show that the proposed SRIT2FNN performs better than the conventional FNN and IT2FNN.     

Indirect predictive type-2 fuzzy neural network controller for a class of nonlinear input - delay systems

In this paper a new indirect type-2 fuzzy neural network predictive (T2FNNP) controller has been proposed for a class of nonlinear systems with input-delay in presence of unknown disturbance and uncertainties. In this method, the predictor has been utilized to estimate the future state variables of the controlled system to compensate for the time-varying delay. The T2FNN is used to estimate some unknown nonlinear functions to construct the controller. By introducing a new adaptive compensator for the predictor and controller, the effects of the external disturbance, estimation errors of the unknown nonlinear functions, and future sate estimation errors have been eliminated. In the proposed method, using an appropriate Lyapunov function, the stability analysis as well as the adaptation laws is carried out for the T2FNN parameters in a way that all the signals in the closed-loop system remain bounded and the tracking error converges to zero asymptotically. Moreover, compared to the related existence predictive controllers, as the number of T2FNN estimators are reduced, the computation time in the online applications decreases. In the proposed method, T2FNN is used due to its ability to effectively model uncertainties, which may exist in the rules and data measured by the sensors. The proposed T2FNNP controller is applied to a nonlinear inverted pendulum and single link robot manipulator systems with input time-varying delay and compared with a type-1 fuzzy sliding predictive (T1FSP) controller. Simulation results indicate the efficiency of the proposed T2FNNP controller.     

云遗传算法在木材干燥过程建模中的应用

木材干燥过程具有强耦合、非线性的特点,而且木材本身的结构也具有复杂性和多样性,所以很难建立起精确的数学模型。本文利用云遗传算法优化BP神经网络的权值建立木材干燥基准模型,并与BP网络干燥基准模型做比较。实验表明优化后的干燥基准模型提高了收敛速度和误差精度。     

Synchronization of chaotic systems and identification of nonlinear systems by using recurrent hierarchical type-2 fuzzy neural networks

This paper proposes a novel approach for training of proposed recurrent hierarchical interval type-2 fuzzy neural networks (RHT2FNN) based on the square-root cubature Kalman filters (SCKF). The SCKF algorithm is used to adjust the premise part of the type-2 FNN and the weights of defuzzification and the feedback weights. The recurrence property in the proposed network is the output feeding of each membership function to itself. The proposed RHT2FNN is employed in the sliding mode control scheme for the synchronization of chaotic systems. Unknown functions in the sliding mode control approach are estimated by RHT2FNN. Another application of the proposed RHT2FNN is the identification of dynamic nonlinear systems. The effectiveness of the proposed network and its learning algorithm is verified by several simulation examples. Furthermore, the universal approximation of RHT2FNNs is also shown.