A predictive maintenance system for multi-granularity faults based on AdaBelief-BP neural network and fuzzy decision making

Predictive maintenance of production equipment is a prerequisite to ensure safe and reliable manufacturing operations. To avoid unexpected shutdown and even casualties caused by faults during production, it is paramount to design an effective predictive maintenance decision system for production equipment. Most of the related research works concentrate on early warn of specific faults but neglect the differentiations of the fault severity. To address the issue, this paper presents an intelligent predictive maintenance system for multi-granularity faults of production equipment based on the AdaBelief-BP (back propagation) neural network and the fuzzy decision making. The characteristics of the system presented in this paper include: (1) The proposed system implements a two-stage framework, integrating the functions of fault type prediction and fault degree prediction, which can provide comprehensive fault information throughout production lifecycles; (2) On the maintenance solution identification stage, fuzzy logic-based decision making is carried out to determine appropriate maintenance solutions based on the practical vague boundary of fault severity. In the system, the design of the AdaBelief-BP neural network can achieve a higher convergence rate and a better generalization capability as well. Meanwhile, to the best of our knowledge, in this research, it is the first time to use the migration of the fuzzy membership degree as the indicator of the changing condition of fault severity to facilitate more accurate maintenance solution identification. To verify the effectiveness of the system, comparison experiments with some popular algorithms are conducted. Benchmarking results show that the developed system can achieve higher prediction accuracy as well as higher efficiency than the comparative methods.     

基于扩展T-S模型的PSO神经网络在故障诊断中的应用

针对现实故障现象具有模糊性和非线性的特点,提出了一种利用自适应扩展T-S(Takagi-Sugeno)模糊模型的PSO(Particle Swarm Optimization)算法和神经网络相结合的新型智能结构化算法来进行故障诊断的新方法.首先通过自适应的高斯函数来更改基本T-S模糊模型中的隶属度函数,进而使用扩展的T-S模糊模型来调整PSO算法的参数.然后使用该PSO算法作为神经网络的学习训练算法来进行训练.最后将此算法用于齿轮箱实测故障诊断.诊断结果显示均方误差提高了0.1981%.通过不同模型的诊断结果比较,表明本方法便捷、高效,为解决故障诊断问题提供了一条新途径.     

基于建康管理技术的机载计算机智能故障诊断方法

为了满足飞机机载电子设备以状态监控为基础的视情维修保障策略,提升设备可维护性,提出了一种基于在线检测、故障预测、辅助决策的健康监控管理故障诊断方法,支持对机载电子设备的健康状态进行预测和评估。通过划分机载电子设备子功能的敏感威胁区域,对这些区域设计专门的威胁预警监控电路,进行功能危害监控,建立推理监控模型对监控电路故障进行预警监控,结合辅助决策的方式对预警到的故障进行定位,实现对电子设备的智能故障诊断。通过FMEA的分析与故障注入测试验证,该预警电路、推理模型和辅助决策能有效的预测故障及定位,具有较高的故障预测覆盖率,可提高机载计算机的维修性、降低维修时间,在电子设备视情维修策略上具备工程应用价值。     

基于优化的支持向量机的机械设备多故障诊断模型

提出了一种基于遗传算法优化支持向量机的故障诊断模型.它利用遗传算法对支持向量机同时对传统的时域特征参量子集和核参数同时优化,以达到选择最优的设备故障主导特征参数组合的目的,实现对机器不同类型故障的识别.对齿轮故障诊断的结果表明它有效提高了多分类支持向量机的故障分类准确性.     

新型电子装备软件故障诊断方法研究

操作程序(软件)已成为新型电子装备的重要的组成部分．软件中存在的缺陷影响电子装备的战术与技术性能。为了检测和诊断电子装备操作软件故障．针对软件特点和故障特征．建立了软件故障诊断模型，分析了故障隔离方法，提出了基于故障报告信息的诊断策略。实践证明．该方法较好地解决了电子装备操作软件的故障诊断问题。     

基于大数据的电力企业资产组维修策略优化研究

汇总过去若干年的电力设备故障数据,运用大数据分析方法,把故障预测技术引入到预防性维修的实践中,提出一种基于大数据的预防性维修策略。首先,根据由状态检测信息得到剩余寿命的预测结果,以预防性维修时的剩余寿命为阀值制定预防性维修策略。然后,根据更新过程理论,建立以电力设备的预防性维修阀值和预测间隔期为优化变量,综合考虑电力设备维修成本、客户满意度、电量销售、停电损失、维修时机选择等约束条件呢,以电力设备平均维修费用最小和电力设备可用度最大为优化目标的预防性维修优化模型。采用人群搜索算法进行优化求解,得到系统最佳的预防性维修阀值和维修预测间隔期。最后,通过引入算例,对所建模型优化仿真求解,得到电力设备最佳的预测周期,在保证电力设备可用度的同时,使电力设备的平均维修费用最小,验证了所建模型的可行性和有效性,从而提高电力企业的整体效益。     

基于免疫神经网络的航空设备故障预测研究

为解决武器装备全寿命周期费用高、经济可承受性差的难题,开展航空装备的故障预测技术研究。采用免疫算法改进隐含层激励函数得到免疫神经网络,用以进行装备特征参数的跟踪预测,结果表明免疫算法改进的神经网络可在故障前3小时实现预测,较BP网络性能有较大改善。     

民航涡扇发动机故障诊断的人工神经网络方法仿真研究

民用航空发动机故障诊断的方法很多，而各种方法在具体操作过程中有着不同的作用和意义。本文提出一种依靠人工神经网络方法研究民航发动机各种故障征兆影响因素的作用程度，并利用其仿真数据分析，找出具体工况过程中作用相对较大的因素。从而在适时与视情维修过程中提出合理的建议，以便为相应维修方法的实施、改进和研究应用提供可靠的依据。     

A data-driven two-stage maintenance framework for degradation prediction in semiconductor manufacturing industries

To reduce the production costs and breakdown risks in industrial manufacturing systems, condition-based maintenance has been actively pursued for prediction of equipment degradation and optimization of maintenance schedules. In this paper, a two-stage maintenance framework using data-driven techniques under two training types will be developed to predict the degradation status in industrial applications. The proposed framework consists of three main blocks, namely, Primary Maintenance Block (PMB), Secondary Maintenance Block (SMB), and degradation status determination block. As the popular methods with deterministic training, back-propagation Neural Network (NN) and evolvable NN are employed in PMB for the degradation prediction. Another two data-driven methods with probabilistic training, namely, restricted Boltzmann machine and deep belief network are applied in SMB as the backup of PMB to model non-stationary processes with the complicated underlying characteristics. Finally, the multiple regression forecasting is adopted in both blocks to check prediction accuracies. The effectiveness of our proposed two-stage maintenance framework is testified with extensive computation and experimental studies on an industrial case of the wafer fabrication plant in semiconductor manufactories, achieving up to 74.1% in testing accuracies for equipment degradation prediction.     

基于改进一维卷积和双向长短期记忆神经网络的故障诊断方法

针对工业领域中故障诊断数据存在时序性和夹杂强噪声的特点导致的收敛速度慢以及诊断精度低的问题,提出了一种基于改进一维卷积和双向长短期记忆（1DCNN-BiLSTM）神经网络融合的故障诊断方法。该方法包括故障振动信号的预处理、特征的自动提取以及振动信号的分类。首先,采用自适应白噪声的完整经验模态分解（CEEMDAN）技术对原始振动信号进行预处理;其次,构建1DCNN-BiLSTM双通道模型,将处理后信号输入双向长短期记忆（BiLSTM）神经网络模型和一维卷积神经网络（1DCNN）模型两个通道,从而对信号的时序相关性特征、局部空间的非相关性特征和弱周期性规律进行充分提取;然后,针对信号夹杂强噪声的问题,对压缩与激励网络（SENet）模块进行改进并将其作用于两个不同的通道;最后,输入全连接层将双通道提取的特征进行融合并借助Softmax分类器实现对设备故障的精确识别。使用凯斯西储大学轴承数据集进行实验,结果表明改进后的SENet模块同时作用于1DCNN通道和stacked BiLSTM通道,1DCNN-BiLSTM双通道模型在保证快速收敛的情况下有最高诊断精度96.87%,优于传统单通道模型,有效提高了机械设备故障诊断效率。     

基于循环神经网络的电信行业容量数据预测方法

在电信运维的容量预测过程中,存在容量指标和部署业务种类繁多的问题。现有研究未考虑指标数据类型的差异,对所有类型的数据使用同种预测方法,使得预测效果参差不齐。为了提升指标预测效率,提出一种指标数据类型分类方法,利用该方法将数据类型分为趋势型、周期型和不规则型。针对其中的周期型数据预测,提出基于双向循环神经网络（BiRNN）的周期型容量指标预测模型,记作BiRNN-BiLSTM-BI。首先,为分析容量数据的周期特征,提出一种忙闲分布分析算法;其次,搭建循环神经网络（RNN）模型,该模型包含一层BiRNN和一层双向长短时记忆网络（BiLSTM）;最后,充分利用系统忙闲分布信息,对BiRNN输出的结果进行优化。与传统的三次指数平滑、差分自回归移动平均（ARIMA）模型和反向传播（BP）神经网络模型进行比较的实验结果表明,在统一日志数据集和分布式缓存数据集上,提出的BiRNN-BiLSTM-BI模型的均方误差（MSE）分别比对比模型中表现最优的模型降低了15.16%和45.67%,可见预测准确率得到了很大程度的提升。     

一种基于容错的感知数据回归模型研究

传感器主要用于对外部环境进行监测,然而当传感器发生故障时监测结果会出现误差。为了提高传感器发生故障时系统的容错能力,提出了一种容错的感知数据回归模型。首先,对最小二乘和岭回归两种线性回归模型进行分析,并分析了线性回归模型的相关统计量;然后,分析了部分传感器发生故障时系统的相关统计量,并以此为基础分析了协变量矩阵的上下界;最后,依据协变量矩阵定义了故障指标,并将优化模型转化为同时最小化故障指标和均方误差的问题。实验表明,提出的容错回归模型与传统的最小二乘法和岭回归方法相比具有更小的预测误差,因而当传感器发生故障时所提模型具有更好的健壮性。     

工程机械远程故障监诊系统

工程机械由于其结构复杂,使用范围广,使得故障诊断与维护工作难度加大。针对这种情况设计一种工程机械的远程监控诊断系统,为分布在不同地理位置上的机械设备提供及时、有效、全面的诊断维护服务,缩短设备维修时间,提高设备利用率。系统基于ARM体系结构的嵌入式系统作为车载终端,通过GPRS网络与远程服务器交互。采用多传感器信息融合技术和故障树分析相结合增加了故障预测的准确率。     

A signal-processing tool for non-destructive testing of inaccessible pipes

The design of Non-Destructive-Testing systems for fault detection in long and not accessible pipelines is an actual task in the industrial and civil environment. At this purpose, the diagnosis based on the propagation of guided ultrasonic waves along the pipes offers an attractive solution for the fault identification and classification. The authors studied this problem by means of suitable Artificial Neural Network models. Numerical techniques have been used to simulate the guided wave propagation in the pipes. In particular, the finite element method has been used to model different kinds of pipes and faults, and to obtain several returning echoes containing the faults information. Torsional wave modes have been used as excitation waves. The obtained signals have been processed in order to reduce the data dimensionality, and to extract suitable features. The features selected from the signals can be further processed in order to limit the size of the Neural Network models without loss of information. At this purpose, the principal component analysis has been investigated. Finally, the selected features have been used as input for the neural network models. In this paper, traditional feed-forward, multi-layer perceptron networks have been used to obtain the information on size and location of localized notches.     

Fault detection and diagnosis of pneumatic valve using Adaptive Neuro-Fuzzy Inference System approach

Detection and diagnosis of faults in cement industry is of great practical significance and paramount importance for the safe operation of the plant. In this paper, the design and development of Adaptive Neuro-Fuzzy Inference System (ANFIS) based fault detection and diagnosis of pneumatic valve used in cooler water spray system in cement industry is discussed. The ANFIS model is used to detect and diagnose the occurrence of various faults in pneumatic valve used in the cooler water spray system. The training and testing data required for model development were generated at normal and faulty conditions of pneumatic valve in a real time laboratory experimental setup. The performance of the developed ANFIS model is compared with the MLFFNN (Multilayer Feed Forward Neural Network) trained by the back propagation algorithm. From the simulation results it is observed that ANFIS performed better than ANN.     

基于维修日志的飞机设备故障原因判别方法

在飞机维修与保养过程中,航空维修公司已积累了大量经验性的维修日志数据.合理利用该类维修日志,结合机器学习方法,可以辅助维修人员做出正确的故障诊断决策.首先,针对维修日志的特殊性,提出一种迭代式的故障诊断基本过程;其次,在传统的文本特征提取技术的基础上,基于领域内信息,提出一种基于卷积神经网络（convolution neural network,简称CNN）的小样本文本特征提取方法,在样本量较少的情况下,利用预测目标将字向量作为输入,得到更为充分的文本特征;最后,使用随机森林（random forest,简称RF）模型,结合其他故障特征判别飞机设备的故障原因.卷积神经网络以故障原因为目标,预先对故障现象中的字向量进行训练,从而得到更能反映该领域的文本特征.与其他文本特征提取方法相比,该类方法在小样本数据上得到了更好的效果.同时,将卷积神经网络与随机森林模型应用于飞机设备的故障原因判别,并与其他文本特征提取方式和机器学习预测模型进行对比,说明了该类文本特征提取方式和故障原因判别方法的合理性和必要性.     

基于深度聚合神经网络的网约车需求时空热度预测

为解决服务车辆与乘客间的供需不平衡问题,提升服务车辆的运营效率和利润,同时降低乘客等待时间并改善其对服务平台的满意度,针对差异化结构的多维时空数据,提出一种深度聚合神经网络（DANN）模型用于对网约车需求进行预测。首先,通过综合考虑时间、空间和外部环境等多维影响因素,提出了基于周期的时空变量和基于图像点值的空间变量划分方法;其次,依据数据特点构建了不同的子神经网络结构来分别拟合时间变量、空间变量和环境变量与需求间的非线性关系;然后,提出了多种异类子神经网络的聚合方法以同时捕捉不同结构时空数据的隐含特征;最后,分析了聚合权重的设置方法以获得网络模型的最优性能。实验结果表明,在三个真实数据集上所提模型的R2平均误差仅为9.36%,与卷积长短时记忆网络（FCL-Net）和混合深度学习神经网络（HDLN-Net）模型相比,所提模型的R2分别平均提升了4.6%和5.22%,均方误差（MSE）分别平均降低了27.01%和26.6%。因此,DANN在实际应用中能较大幅度地提升需求预测的准确性,可以作为网约车需求预测的有效手段。     

基于蚁群优化神经网络的AUV舵的故障诊断

针对流线型AUV舵故障,提出了基于Elman神经网络的故障诊断方法。基于蚁群算法优化改进型Elman神经网络,建立了AUV角速度运动模型,通过蚁群算法和梯度下降法对改进型Elman神经网络训练的对比分析,验证了蚁群算法优化的改进型Elman神经网络具有训练速度快,不易陷入最优解等特点。提出了基于角速度残差检测舵故障,再通过定角度航行和定速直航的主动诊断方式,判定舵故障类型的故障诊断方法,探讨了基于角速度残差和角度残差的变化趋势来诊断舵卡死和舵变形故障的故障决策方法。对流线型AUV的舵变形及舵卡死故障进行了水池模拟实验,实验结果验证了所提方法的有效性。     

Application of case-based reasoning in a fault detection system on production of drippers

Injection molding has been a preferred production process in the fabrication of complex components. In this technique not only the injection machine and mold play important roles, but also different process parameters have strong effects on the quality of the final products. The production process might be stopped because of different types of faults on the production line. In this paper, a case-based reasoning (CBR) methodology is employed to implement an intelligent fault detection system for the production of injection molded drippers. This CBR system utilizes similar occurred faults to solve particular new problems. Case retrieval and similarity measurements are defined based on fault occurrence weight of features (fault’s causes). Application and accuracy of the proposed system are experimentally tested and validated through analyzing the current case study. The obtained results indicated that the implemented CBR system is able to detect the faults on the injection molding machine. By utilizing the proposed system machine downtime is reduced, speeded production with high productivity is achieved.     

一起500千伏变压器多维状态监测与故障诊断

为了提升变压器故障诊断水平并有效实现状态检修,本文针对一起500千伏主变压器内部潜伏性故障案例介绍了利用多种在线监测、离线检测等手段与故障诊断方法快速准确查找出变压器故障原因及位置。它通过采用油色谱检测、变压器振动及声音检测、容性设备在线监测、变压器局部放电检测、带电测试等多维状态监测技术开展变压器状态评价及故障诊断,并帮助实现对主变压器内部故障的准确定位和制定状态检修策略。该主变在通过状态监测及故障诊断后实现隐患故障快速准确定位,并通过紧急处理重新恢复正常运行状态,从而避免一起重大设备事故发生。结果表明,应用变压器油溶解气体色谱分析检测技术,可通过排除法准确地判断变压器故障性质和严重程度,它是早期发现变压器潜伏性故障特别有效的方法。同时,采用基于交叉小波的变压器振动信号特征量提取方法分析评价结果表明,变压器铁芯发生接地故障后其振动信号存在大量50赫兹谐波分量。借助振动及声音检测技术,能够有助于变压器故障诊断及准确定位,提高设备状态检修效率。