神经网络方法在自相关过程控制中的应用

将传统休哈特控制图应用于自相关过程控制时,会引发大量虚发报警.本文将使用时间序列模型模拟自相关过程并将神经网络方法引入自相关过程控制中.以神经网络特有的模式识别技术,对自相关过程中均值发生突变的情况进行监控,取得了良好效果.     

平稳自相关过程的EWMA控制图

讨论了对平稳自相关过程中出现的较小波动进行监控的一种方法.采用自回归移动平均(ARMA)模型对平稳自相关过程进行适当的拟合,通过计算残差的方法消除过程中的自相关要素,并在此基础上提出对于均值和方差出现的较小波动进行监控的指数加权移动平均(EWMA)控制图的构造.通过与其它几种方法的比较来说明该方法在监控平稳自相关过程时有更好的效率.     

自相关过程的修正均值控制图在MAP中的应用

传统控制图的应用前提是来自过程的观测值彼此独立,但通过自动化测试设备获得的校准值通常存在自相关现象,违背独立性假定.结合新型计量保证方案应用的工程实际,对观测值的自相关参数做出辨识和估计,建立一阶自回归时间序列模型,修正了传统控制图的控制限,使之适用于自相关过程.实例分析表明,当过程观测值存在自相关时使用常规控制图将得到错误的控制限设置,使用修正均值控制图则可以正确判断过程是否受控.     

卷积门控循环残差网络在轴承故障诊断中的研究

针对滚动轴承故障特征较难提取及许多深度学习方法因模型简单而导致准确率偏低的问题,提出一种基于残差网络的门控循环网络(GRU),该算法可以减少时序信息的丢失及解决由于网络较深而出现性能下降的问题。该模型包含2个卷积层、2个GRU层、1个残差块以及1个输出层,其先利用具有强大特征提取能力的卷积神经网络(CNN)提取轴承振动信号中的信息,然后将提取到的信息输入GRU中以保证时序信息不丢失,再通过残差模块解决神经网络深度较深问题,最后通过输出层输出结果。结果表明：该方法可以一次性诊断多种轴承的不同位置及不同尺寸的故障,且对比其他深度学习网络,该算法具有更高的准确性。     

基于GRU神经网络的UWB定位算法研究

针对复杂环境下的室内精确定位需求,采用基于双边双向测距的超宽带定位方案。为解决非视距情况下超宽带测距误差大而引起定位精度下降的问题,考虑到超宽带测距具备时间序列预测问题的特点,引入了循环门控单元(Gate Recurrent Unit, GRU)来搭建神经网络,并设计不同隐藏层数、结构来验证其有效性。实验结果表明,相比于LS(Least Square)、 UKF(Unscented Kalman Filter)定位算法,该GRU神经网络定位算法的均方根误差指标平均降低了30.81%、 21.51%,定位效果更好。     

输入层神经元数对自相关过程控制的影响

由于将传统休哈特控制图应用于自动化连续生产过程,经常会引发大量虚发报警,而使用神经网络方法对存在相关性的连续生产过程进行研究时,可取得良好效果.使用时间序列模型模拟自相关过程中的均值变动,在以BP神经网络对自相关过程进行监控的基础之上,通过神经网络识别率的变化趋势分析,对输入层神经元数对于神经网络识别率的影响进行分析研究,以便使BP神经网络的识别率得到优化.     

统计过程控制中基于数据深度的非参数变点控制图

为了解决多元非正态分布情况下的过程控制问题,提出基于数据深度的变点控制图,并对构建该控制图检验统计量的具体方法及控制流程进行了详细描述。为了检验该控制图的控制效果,采用服从二元伽马分布的样本数据对其进行了验证,并设置位置参数偏移范围为0.2至1.0,变点为14、24、34,几种情况分别检验其控制效果。数据仿真的结果表明:偏移越大,检测效果越好;偏移量小于0.7时,变点越大,检测效率越高;而当变点大于0.7时变点对检测效果的影响不明显。偏移量在0.1至0.4的范围内,变点越大,检测效果越好,但是这种边际效果在减小。     

基于深度残差学习的彩色图像去噪研究

目的 当噪声存在时,尤其是等级相对较大的噪声,会导致彩色图像的视觉质量下降,为了有效去除噪声的同时使去噪后的图像有更好的视觉效果,提出一种基于深度残差学习的彩色图像去噪方法。方法 首先设计由多个残差单元模块组成的残差层,然后在每个残差单元模块之间添加跳跃连接,构成由噪声图像到去噪图像的非线性映射,并优化残差单元个数,使网络能学习到更多的图像细节特征,以提升网络的去噪性能,同时将每个残差单元模块中的激活函数提到卷积层前面,以加速网络收敛。结果 与常用去噪算法相比,文中方法在Kodak24和CBSD100数据集上的主观视觉打分MOS值以及客观指标(PSNR和SSIM)上,较其他方法有更好的效果。结论 提出的基于深度残差学习的彩色图像去噪方法能有效去除图像中的噪声,尤其是较严重的噪声,并取得了良好的视觉效果,表明该方法具有良好的去噪性能。     

基于神经网络的SPC/EPC整合过程监测方法研究

为解决统计过程控制(SPC)/工程过程调整(EPC)整合引起的传统SPC控制图监测异常扰动效率低的问题,提出了采用神经网络技术监测SPC/EPC整合过程的策略,并对神经网络模型结构和参数设置进行分析,构建过程输入、过程输出及两者的协方差为输入参数,异常扰动发生与否为输出参数的3层神经网络模型。为验证该方法的性能,进行了大量的比较实验:即对相同的样本,分别采用Shewhar图、CUSUM图和上述神经网络模型进行监测。实验结果表明:神经网络模型能准确监测幅度大于2的阶跃扰动和大于2的过程漂移,平均运行步长(ARL)为1;传统SPC监测技术只能较准确地(监测率大于90%)监测幅度大于5的阶跃扰动和大于2的过程漂移,ARL大于2。与传统监测方法相比,该方法能快速有效地监测异常扰动的发生。     

控制图自动分析系统

描述了一种控制图自动分析系统,该系统具有自动计算各统计量与控制界限、绘图以及根据国标GB/T4091-2001对控制图进行自动分析等功能.通过该系统的应用实例,证明其具有很强的实用性.     

Deep recurrent neural network‐based residual control chart for autocorrelated processes

With the growth of automation in process industries, there is correlation in the process variables. Deep learning has achieved many great successes in image and visual analysis. This paper concentrates on developing a deep recurrent neural network (RNN) model to characterize process variables at vary time lags, and then a residual chart is developed to detect mean shifts in autocorrelated processes. The experiment results indicate that the RNN‐based residual chart outperforms other typical methods (eg, autoregressive [AR]‐based control chart, back propagation network [BPN]‐based residual chart). This paper provides guideline for deep learning technique employed as an effective tool in autocorrelated process control.     

An AR‐Sieve Bootstrap Control Chart for Autocorrelated Process Data

There are two major approaches in dealing with autocorrelated process data in process control, that is, residual‐based approaches and methods that modify control limits to adjust for autocorrelation. We proposed a methodology for constructing control charts for autocorrelated process data using the AR‐sieve bootstrap. The simulation study illustrates the relative advantage of the AR‐sieve bootstrap control chart with respect to the in‐control and out‐of‐control run length and false alarm rate. The proposed methodology works even for small sample sizes and conditions of the near nonstationarity of the generating process. The proposed AR‐sieve bootstrap control chart presents the advantage of being distribution‐free for certain class of linear models as well as the tracking of actual process observations instead of model residuals, thus facilitating the implementation during actual plant operations. Copyright © 2011 John Wiley & Sons, Ltd.     

面向未知自相关过程的Bootstrap控制图设计

针对自动化生产中,通过拟合自回归滑动平均模型(ARMA)建立残差控制图监控未知自相关过程数据时,存在误报率高的问题,提出一种基于Bootstrap的方法,通过重构样本,对原始数据建立非参数控制图。在考虑不同的模型系数、偏移大小、样本个数及残差分布类型的情况下,通过蒙特卡洛模拟,比较传统残差控制图和新控制图的平均运行链长(ARL),证明新控制图提高了对过程偏移的灵敏度,降低了误报率。实际应用中,新的Bootstrap控制图在仅获取一组Phase-I阶段的受控数据样本下即可生成,受所取样本个数的影响较小,且直接用于监控原始数据,适用范围广,操作简便。     

A Two‐Sided Cumulative Sum Chart for First‐Order Integer‐Valued Autoregressive Processes of Poisson Counts

Count data processes are often encountered in manufacturing and service industries. To describe the autocorrelation structure of such processes, a Poisson integer‐valued autoregressive model of order 1, namely, Poisson INAR(1) model, might be used. In this study, we propose a two‐sided cumulative sum control chart for monitoring Poisson INAR(1) processes with the aim of detecting changes in the process mean in both positive and negative directions. A trivariate Markov chain approach is developed for exact evaluation of the ARL performance of the chart in addition to a computationally efficient approximation based on bivariate Markov chains. The design of the chart for an ARL‐unbiased performance and the analyses of the out‐of‐control performances are discussed. Copyright © 2012 John Wiley & Sons, Ltd.     

Using Neural Networks to Detect and Classify Out‐of‐control Signals in Autocorrelated Processes

This paper presents an artificial neural network model for detecting and classifying three types of non‐random disturbances referred to as level shift, additive outlier and innovational outlier which are common in autocorrelated processes. To the best of our knowledge, this is the first time that a neural network has been considered for simultaneous detection and classification of such non‐random disturbances. An AR (1) model is considered to characterize the quality characteristic of interest in a continuous process where autocorrelated observations are generated over time. The performance of the proposed procedure is evaluated through the use of a numerical example. Preliminary results indicate that the procedure can be used effectively to detect and classify unusual shocks in autocorrelated processes. Copyright © 2003 John Wiley & Sons, Ltd.     

Single Variables Control Charts: an Overview

Control charts are widely used in industries to monitor a process for quality improvement. When dealing with variables data, we usually employ two control charts to monitor the process location and spread. We give an overview of the control charts proposed in the last decade or so in an effort to use only one chart to simultaneously monitor both process location and spread. Two approaches have been advocated for using one control chart for process monitoring. One approach plots two quality characteristics in the same chart while the other uses one plotting variable to represent the process location and spread. Copyright © 2006 John Wiley & Sons, Ltd.     

An Effective Double Sampling Scheme for the c Control Chart

In the present paper is developed a statistical process control inspection procedure based on a new simple‐to‐implement and effective double sampling scheme for the c control chart, aimed at the minimization of the number of inspected observation units and warranting fixed levels for the type I and II error risks. In particular, the formulations of the false alarm risk α, the power P of the chart, and the expected number of inspected observation units for the developed inspection procedure are given, whereas a macro of Microsoft Excel is adopted to solve the tackled problem. In order to illustrate the application of the developed approach and to investigate on the influence of several operating parameters, numerical examples are carried out and the related considerations are given. Finally, by comparing the performance of the developed inspection procedure with that of the related classic c chart scheme, meaningful reduction of the number of the inspected observation units can be achieved by adopting the proposed approach. Copyright © 2013 John Wiley & Sons, Ltd.     

A Bootstrap Control Chart for Weibull Percentiles

The problem of detecting a shift of a percentile of a Weibull population in a process monitoring situation is considered. The parametric bootstrap method is used to establish lower and upper control limits for monitoring percentiles when process measurements have a Weibull distribution. Small percentiles are of importance when observing tensile strength and it is desirable to detect their downward shift. The performance of the proposed bootstrap percentile charts is considered based on computer simulations, and some comparisons are made with an existing Weibull percentile chart. The new bootstrap chart indicates a shift in the process percentile substantially quicker than the previously existing chart, while maintaining comparable average run lengths when the process is in control. An illustrative example concerning the tensile strength of carbon fibers is presented. Copyright © 2005 John Wiley & Sons, Ltd.