用于建模、优化、故障诊断的数据挖掘技术

建模、优化、故障诊断是流程工业ＣＩＭＳ技术中的关键技术。传统的建模、优化、故障诊断方法依赖于数学模型仿真或专家经验规则,对于强非线性和非高斯分布噪声的对象存在着知识获取瓶颈。而数据挖掘技术综合运用机器学习、计算智能（人工神经网、遗传算法）、模式识别、数理统计等技术,从大量数据中挖掘和发现有价值和隐含的知识。本文进一步研究了建模、优化、故障诊断的数据挖掘系统,以及规则挖掘、参变量优化、故障诊断建模的     

发电机组状态巡检与故障诊断系统的研究与实现

针对传统故障诊断专家系统实用性弱等缺点，面向现场实际应用，提出了基于状态预警机制并且融自动诊断、状态巡检和常规专家系统诊断为一体的多层次故障诊断系统设计方案。该方案基于数理统计原理，通过正态性和平稳性检验实现状态预警机制；利用故障模型辨识和诊断规则的量化实现自动诊断；通过对诊断规则的总结形成诊断流程以实现故障普查。利用该设计方案，针对实际汽轮机组和水轮机组，开发了相应的实际应用系统，并取得了很好的应用效果。     

基于 LeNet5like 的迁移学习风电机组叶片覆冰故障诊断研究

针对海上风电场和高海拔地区风机机组的叶片覆冰故障模型精度低、建模速度慢等问题,提出一种基于 LeNet5like 的 迁移学习风电机组叶片覆冰故障诊断方法。 首先,整合监控和数据采集系统的记录数据与风机覆冰情况进行预处理,建立训练 数据集;其次,基于改进后的 LeNet5like 网络构建覆冰故障诊断模型,提取数据集中多变量间的相关性特征信息;然后,经网络 参数微调迁移学习对模型进行训练,实现对其他风机覆冰故障诊断模型的快速建立;最后,经实验验证,该模型覆冰故障诊断准 确率为 98. 90% ,较无迁移模块网络训练时间缩短 28 s,提升约 15. 91% ,验证了基于 LeNet5like 的迁移学习风电机组叶片覆冰 故障诊断方法的精确性和快速性。     

基于粗集理论的往复泵泵阀故障诊断方法

提出了一种基于粗集理论的往复泵泵阀故障诊断方法。该方法可以直接从经过小波包处理的泵阀振动信号中提取故障诊断观测，并由此建立基于规则的泵阀故障诊断系统，该系统不仅可以对发生故障的单个泵阀进行诊断，而且还能对同时发生故障的多个泵阀进行诊断，试验结果表明了这种方法的有效性。这种方法的可行性也为其它复杂机械的故障诊断提供了新思路。     

基于随机集理论的并发故障诊断信息融合方法

为了诊断并发故障,提出一种基于随机集理论的信息融合方法.首先构造包含并发故障的论域,并在此论域的超幂集上定义扩展型随机集.基于该随机集和广义集值映射给出证据组合规则的随机集模型,用其构造可以同时适用于单发和并发故障诊断的新型组合规则.此外,根据传感器提供的故障信息构造故障样板模式与待检模式的模糊隶属度函数,利用模糊集的随机集表示以及随机集似然测度,获得两种模式匹配的程度作为待融合的诊断证据.最后通过在电机柔性转子平台上的试验,证明了所提方法可有效地减少单一传感器信息诊断的不确定性,显著提高转子系统故障诊断的精度.     

基于物联网技术的矸石发电机组故障诊断系统设计

为确保矸石发电厂正常运行,降低运行成本,设计一种基于物联网技术的矸石发电机组故障诊断系统。系统硬件部分由下位机和嵌入式网关组成,根据发电机组数据特征分别给出各模块的电路设计方案;系统软件由预处理机、状态服务器和 Web 服务器组成,通过构建物联网传感器基阵,采集发电机组实时监测数据,利用 FCM 算法获得发电机组数据特征模式矢量,确定故障类别和中心矢量,以此实现矸石发电机组故障诊断。实验表明,该系统能够正常进行数据库访问和数据检索等操作,可有效地执行业务规则并完成相关工作。     

Optimization of measuring points for machine tool thermal error based on grey system theory

Optimization of thermal sensors’ placement on machine tools based on grey correlation model of grey system theory is studied. After optimization, the temperature variables in the thermal error’ model are reduced from 16 to 4. It greatly reduces the time for variable searching and modelling and meanwhile it eliminates the coupling problems among temperature variables, so the robustness of the model could be increased and the predicting precision of the model is enhanced. Consequently, the real-time error compensation would be more effective and convenient.     

基于LS-SVM的一次风机振动在线监测及故障预警研究

针对火电厂一次风机运行工况复杂和多状态变量强耦合特性而难以构建设备精确模型问题,将智能数据挖掘方法应用于风机设备故障预警和诊断中。通过对风机典型运行特性进行分析,提出了一种基于最小二乘支持向量机(LS-SVM)的一次风机振动状态估计和故障预警方法。结合山西河曲发电厂1号机组的1#一次风机历史运行数据,应用Matlab对所提出的方法进行了验证和分析。研究结果表明,该预测方法有较高的估计精度,能够及时辨别一次风机在运行中的振动异常,适用于火电厂辅机设备的故障诊断,具有一定的工程应用价值。     

Intelligent Predictive Decision Support System for Condition-Based Maintenance

The high costs in maintaining today’s complex and sophisticated equipment make it necessary to enhance modern maintenance management systems. Conventional condition-based maintenance (CBM) reduces the uncertainty of maintenance according to the needs indicated by the equipment condition. The intelligent predictive decision support system (IPDSS) for condition-based maintenance (CBM) supplements the conventional CBM approach by adding the capability of intelligent condition-based fault diagnosis and the power of predicting the trend of equipment deterioration. An IPDSS model, based on the recurrent neural network (RNN) approach, was developed and tested and run for the critical equipment of a power plant. The results showed that the IPDSS model provided reliable fault diagnosis and strong predictive power for the trend of equipment deterioration. These valuable results could be used as input to an integrated maintenance management system to pre-plan and pre-schedule maintenance work, to reduce inventory costs for spare parts, to cut down unplanned forced outage and to minimise the risk of catastrophic failure.     

基于状态监测和故障诊断的设备管理系统

当前由于状态监测和故障诊断与维修管理的相互脱节,使其应有的作用没有得到充分发挥。针对这种情况,采用了离线和在线监测诊断相结合的状态维修策略,建立了状态维修的工作模型,运用集成定义方法建立了系统的功能模型,进而综合采用网络技术、计算机技术和数据库技术等,开发出了基于状态监测和故障诊断的设备管理系统,为企业成功开展状态维修提供了有效的管理工具。     

基于粗糙集理论的电力变压器故障诊断研究

粗糙集理论是一种较新的数据处理工具，可以有效地分析和处理不完备信息。运用粗糙集理论研究了因各种复杂因素造成的不完备信号模式下电力变压器故障诊断的方法。该方法利用油色谱分析得到的各种气体浓度百分比作为故障分类的条件属性集，考虑各种故障情况，建立决策表。利用决策表的约简方法进行化简，区分关键信号与冗余信号，导出故障诊断规则，从而达到不完备信号模式下快速准确地故障诊断的目的。通过实际应用表明，该方法简单、有效、具有良好的容错性能。     

Development of a guiding system and visual feedback real-time controller for the high-speed self-align optical cable winding

Recently, the demand for the optical cable has been rapidly growing because of the increasing number of internet users and the high speed internet data transmission required. But the present optical cable winding systems have some serious problems such as pile-up and collapse of cables usually near the flange of the bobbin in the process of cables winding. To reduce the pile-up collapse in cable winding systems, a new guiding system is developed for a high-speed self-align cable winding. First, mathematical models for the winding process and bobbin shape fault compensation were proposed, the winding mechanism was analyzed and synchronization logics for the motions of winding, traversing, and the guiding were created. A prototype cable winding systems was manufactured to validate the new guiding system and the suggested logic. Experiment results showed that the winding system with the developed guiding system outperformed the system without the guiding system in reducing pile-up and collapse in high-speed winding. This paper was recommended for publication in revised form by Associate Editor Dae-Eun Kim Chang-woo Lee received a B.S. degree in Mechanical Engineering from Konkuk University in 2001. He received his M.S. and Ph.D. degrees from Konkuk university in 2003 and 2008, respectively. Dr. Lee is currently a researcher at the Flexible Display Roll to Roll Research Center at Konkuk University in Seoul, Korea. Dr. Lee’s research interests are in the area of fault tolerant control, R2R e-Printing line design, and tension-register control. He is the holder of several patents related to R2R e-Printing system. HyanKyoo Kang received the B.S. and M.S degree in 2000 and 2003 respectively from Konkuk University, Seoul, Korea, where he is currently working toward the Ph.D. degree in mechanical design. He took part in the development of an autoalign guiding system for high-speed winding in a cable winding system, a 3-D roll-shape diagnosis method in a steel rolling system, a design of register controller for high-speed converting machine and real-time control design of electronic printing machine. His research topics include register modeling and control for printed electronics and distributed real-time control. Kee-Hyun Shin (S’81-M’02) received the B.S. degree from Seoul National University, Seoul, Korea, and the M.S. and Ph.D. degrees in mechanical engineering from Oklahoma State University (OSU), Still-water. Since 1992, he has been a Professor with the Department of Mechanical and Aerospace Engineering, Konkuk University, Seoul, Korea. For more than 18 years, he has covered several research topics in the area of web handling, including tension control, lateral dynamics, diagnosis of defect rolls/rollers, and fault-tolerant real-time control in the Flexible Display Roll-to-Roll Research Center, Konkuk University, of which he has also been a Director. His research topics include distributed real-time control, embedded control, monitoring, and diagnosis and fault-tolerant control of large-scale systems such as steel plants, film-and-paper-making plants, aircraft, ships, and ubiquitous control of multirobot systems. He is the author of Tension Control (TAPPI Press, 2000) and is the holder of several patents related to R2R e-Printing system.     

Sifting process of EMD and its application in rolling element bearing fault diagnosis

Among the vibration-based fault diagnosis methods for rolling element bearing, the shock pulse method (SPM) combined with the demodulation method is a useful quantitative technique for estimating bearing running state. However, direct demodulation often misestimates the shock value of characteristic defect frequency. To overcome this disadvantage, the vibration signal should be decomposed before demodulation. Empirical mode decomposition (EMD) can be an alternative for preprocess bearing fault signals. However, the trouble with this method’s application is that it is time-consuming. Therefore, a novel method that can improve the sifting process’s efficiency is proposed, in which only one time of cubic spline fitting is required in each sifting process. As a consequence, the time for EMD analysis can be evidently shortened and the decomposition results simultaneously maintained at a high precision. Simulations and experiments verify that the improved EMD method, combined with SPM and demodulation analysis, is efficient and accurate and can be effectively applied in engineering practice. This paper was recommended for publication in revised form by Associate Editor Eung-Soo Shin Hongbo Dong was born in Chaoyang, China, in 1979. He received the B.E. and M.E. degree from Northwestern Polytechnical University in Mechanical Engineering in 2002 and 2005 respectively and received the Ph.D degree from Xi’an Jiaotong University in Mechanical Engineering in 2009. His research interests include fault diagnosis of rotor and bearing system. Bing Li was born in Xuzhou, China, in 1976. He received the B.E. and M.E. degree from Northwestern Polytechnical University in Mechanical Engineering in 1999 and 2002 respectively and received the Ph.D degree from Xi’an Jiaotong University in Mechanical Engineering in 2005. After graduating from Xi’an Jiaotong University, he works as a lecturer in Xi’an Jiaotong University. His research interests include wavelet finite element theory and its application in fault diagnosis.     

热力参数的KPCA-RBF网络建模及传感器故障诊断方法

针对复杂恶劣环境下机组热力参数的数据监测及传感器故障诊断问题,建立了融合机理分析、核主元分析(kernel principle component analysis,简称KPCA)与径向基神经网络(radial basis function,简称RBF)的发电机组热力参数预测及传感器故障检测模型。首先,根据机理分析得到完备的辅助变量集,并利用核主元分析提取辅助变量的特征信息以有效处理发电机组中高维、强耦合的非线性数据;其次,将主元变量集输入径向基神经网络进行学习,实现热力参数的重构;最后,基于预测模型与窗口移动法实现传感器的故障诊断,并对故障数据进行及时修复和准确替换。以燃气轮机排气温度为例进行验证的结果表明,该预测模型具有更高的精度和泛化能力,能在传感器故障发生初期及时发现并识别故障类型,检测效果优良。     

Model-based monitoring and fault diagnosis of fossil power plant process units using Group Method of Data Handling

This paper presents an incipient fault diagnosis approach based on the Group Method of Data Handling (GMDH) technique. The GMDH algorithm provides a generic framework for characterizing the interrelationships among a set of process variables of fossil power plant sub-systems and is employed to generate estimates of important variables in a data-driven fashion. In this paper, ridge regression techniques are incorporated into the ordinary least squares (OLS) estimator to solve regression coefficients at each layer of the GMDH network. The fault diagnosis method is applied to feedwater heater leak detection with data from an operating coal-fired plant. The results demonstrate the proposed method is capable of providing an early warning to operators when a process fault or an equipment fault occurs in a fossil power plant.     

A combined mono- and multi-turbine approach for fault indicator synthesis and wind turbine monitoring using SCADA data

The monitoring of wind turbines using SCADA data has received lately a growing interest from the fault diagnosis community because of the very low cost of these data, which are available in number without the need for any additional sensor. Yet, these data are highly variable due to the turbine constantly changing its operating conditions and to the rapid fluctuations of the environmental conditions (wind speed and direction, air density, turbulence, …). This makes the occurrence of a fault difficult to detect. To address this problem, we propose a multi-level (turbine and farm level) strategy combining a mono- and a multi-turbine approach to create fault indicators insensitive to both operating and environmental conditions. At the turbine level, mono-turbine residuals (i.e. a difference between an actual monitored value and the predicted one) obtained with a normal behavior model expressing the causal relations between variables from the same single turbine and learnt during a normal condition period are calculated for each turbine, so as to get rid of the influence of the operating conditions. At the farm level, the residuals are then compared to a wind farm reference in a multi-turbine approach to obtain fault indicators insensitive to environmental conditions. Indicators for the objective performance evaluation are also proposed to compare wind turbine fault detection methods, which aim at evaluating the cost/benefit of the methods from a production manager’s point of view. The performance of the proposed combined mono- and multi-turbine method is evaluated and compared to more classical methods proposed in the literature on a large real data set made of SCADA data recorded on a French wind farm during four years : it is shown than it can improve the fault detection performance when compared to a residual analysis limited at the turbine level only.     

A knowledge-based diagnosis system for automobile engines

A newly-developed knowledge-based diagnosis system for automobile engines is described in this paper. The system is based on the Hierarchical Diagnosis Principle, suggested by the authors. According to this principle, a complex diagnostic task can be divided into several simple ones and then solved step-by-step. Both deep and shallow knowledge are used in the system, and organised in two different knowledge bases:

基于粗糙集理论的变压器故障诊断专家系统研究

在传统的变压器故障诊断专家系统的基础上 ,引入粗糙集理论以解决专家系统较难获取完备知识的瓶颈问题。该系统从历史故障数据所形成的决策表出发 ,运用粗糙集理论进行约简 ,构建专家系统知识库模型。通过计算规则隶属粗糙度 ,来表示诊断规则的置信程度。利用推理机和故障事例库 ,实现对知识库的动态维护。     

基于支持向量机和顺序前项选择算法的PNN风电机组液压变桨的故障诊断

针对液压变桨距系统的强耦合、非线性,以及液压变桨距故障发生原因复杂、故障单一造成的定位问题,该文提出基于支持向量机和顺序前项选择算法的概率神经网络诊断方法。首先,选取SCADA数据的特征值为输入,桨距角为输出,利用支持向量机进行模型的回归,得出桨距角输出的预测值;接着,将测量值与预测值带入顺序前项选择算法,挖掘和发现特征与故障之间的关系,评估各特征之间的重要性,并选出最好的一组特征集合;最后,建立变桨距概率诊断模型,将所选的数据送到故障诊断模型进行训练,再用所选数据进行测试,定位出变桨距系统的故障原因。实验分析表明:基于支持向量机和顺序前项选择算法的概率神经网络液压变桨距故障诊断方法可以有效地分辨出不同故障,并且诊断的精确度得到了提高。     

Rough set based rule learning and fuzzy classification of wavelet features for fault diagnosis of monoblock centrifugal pump

The fault diagnosis problem is conceived as a classification problem. In the present study, vibration signals are used for fault diagnosis of centrifugal pumps using wavelet analysis. Rough set theory is applied to generate the rules from the vibration signals. Based on the strength of the rules the faults are identified. The different faults considered for this study are: pump at good condition, cavitation, pump with faulty impeller, pump with faulty bearing and pump with both faulty bearing and impeller. However, the classification accuracy is based on the strength and number of rules generated using rough set theory. Wavelet features are computed using Discrete Wavelet Transform (DWT) from the vibration signals and rules are generated using rough sets and classified using fuzzy logic. The results are presented in the form of confusion matrix which shows the classification capability of wavelet features with rough set and fuzzy logic for fault diagnosis of monoblock centrifugal pump.