基于PCA和随机森林的故障趋势预测方法研究

故障预测和健康管理技术(PHM)在现代工程系统中能够在系统具备较高复杂度的情况下,有效保障其可靠性和安全性。在机械故障诊断中对于采集到的原始数据的高维特征量的处理较为复杂,并且在实际应用中趋势预测的精度要求较高,针对该问题提出一种基于主成分分析(PCA)与随机森林算法的轴承故障趋势预测方法。该方法利用PCA对提取的原始轴承数据特征量进行线性降维,并选取其中主成分特征量,输出非线性时间序列数据。原始数据经过PCA处理得到非线性时间序列,将该序列作为随机森林算法的输入进行故障趋势预测,并把预测结果与BP神经网络模型预测的结果进行对比,结果表明随机森林在故障趋势预测上在精度相较于BP神经网络有显著提高,是一种有效的故障趋势预测方法。     

Intelligent engineering asset management system for power transformer maintenance decision supports under various operating conditions

Large sized transformers are an important part of global power systems and industrial infrastructures. An unexpected failure of a power transformer can cause severe production damage and significant loss throughput the power grid. In order to prevent power facilities from malfunctions and breakdowns, the development of real-time monitoring and fault prediction tools are of great interests to both researches and practitioners. This research develops an intelligent engineering asset management system for power transformer maintenance. The system performs real-time monitoring of key parameters and uses data mining and fault prediction models to detect transformers’ potential failure under various operating conditions. Principal component analysis (PCA) and a back-propagation artificial neural network (BP-ANN) are the algorithms adopted for the prediction model. Historical industrial power transformer data from Taiwan and Australia are used to train and test the failure prediction models and to verify the proposed general methodology as comparative case studies. The PCA algorithm reduces the number of the primary dissolved gasses as the key factor values for BP-ANN prediction modeling inputs. The system yields effective predictions when verified using various operating condition data from Australia and Taiwan power companies. The accuracy rates are much higher when compared to the fault prediction results without using PCA. The intelligent system combining PCA and BP-ANN algorithms, developed in this research, can be adopted by asset managers in different regions to develop suitable maintenance and repair strategies for transformer failure preventions.     

A process monitoring system based on the Kohonen self-organizing maps

Process monitoring and fault diagnosis have been studied widely in recent years, and the number of industrial applications with encouraging results has grown rapidly. In the case of complex processes a computer-aided monitoring enhances operators possibilities to run the process economically. In this paper, a fault diagnosis system will be described and some application results from the Outokumpu Harjavalta smelter will be discussed. The system monitors process states using neural networks (Kohonen self-organizing maps, SOMs) in conjunction with heuristic rules, which are also used to detect equipment malfunctions.     

Case studies in process modelling and condition monitoring using artificial neural networks

Increasingly artificial neural networks are finding applications in a process engineering environment. Recently the Department of Trade and Industry in the UK has supported the transfer of neural technology to industry with a £5.7M campaign. As part of the campaign, the University of Newcastle and EDS Advanced Technologies Group have set up a Process Monitoring and Control Club.This paper presents two case studies from the work of the Club. Firstly, the ability of neural networks to provide enhanced modelling performance over traditional linear techniques is demonstrated on real process data. Secondly, the ability of neural networks to capture non-linear system characteristics is exploited in a novel way in a condition monitoring exercise. The process studied in both applications is the melter stage of the BNFL Vitrification Process. The process involves the encapsulation of highly active liquid waste in glass blocks to provide a safe and convenient method of storage.     

A novel method of short-term load forecasting based on multiwavelet transform and multiple neural networks

This paper aims to develop a load forecasting method for short-term load forecasting based on multiwavelet transform and multiple neural networks. Firstly, a variable weight combination load forecasting model for power load is proposed and discussed. Secondly, the training data are extracted from power load data through multiwavelet transform. Lastly, the obtained data are trained through a variable weight combination model. BP network, RBF network and wavelet neural network are adopted as the training network, and the trained data from three neural networks are input to a three-layer feedforward neural network for the load forecasting. Simulation results show that accuracy of the combination load forecasting model proposed in the paper is higher than any one single network model and the combination forecast model of three neural networks without preprocessing method of multiwavelet transform.     

Faulted gear identification of a rotating machinery based on wavelet transform and artificial neural network

In this paper, a condition monitoring and faults identification technique for rotating machineries using wavelet transform and artificial neural network is described. Most of the conventional techniques for condition monitoring and fault diagnosis in rotating machinery are based chiefly on analyzing the difference of vibration signal amplitude in the time domain or frequency spectrum. Unfortunately, in some applications, the vibration signal may not be available and the performance is limited. However, the sound emission signal serves as a promising alternative to the fault diagnosis system. In the present study, the sound emission of gear-set is used to evaluate the proposed fault diagnosis technique. In the experimental work, a continuous wavelet transform technique combined with a feature selection of energy spectrum is proposed for analyzing fault signals in a gear-set platform. The artificial neural network techniques both using probability neural network and conventional back-propagation network are compared in the system. The experimental results pointed out the sound emission can be used to monitor the condition of the gear-set platform and the proposed system achieved a fault recognition rate of 98% in the experimental gear-set platform.     

10kV配网电容器在线监测与故障诊断系统

目前大多数电力电容器状态监测系统存在实时性不足和数据频密度不足的缺点,难以实现准确的在线故障预判,易导致故障处理滞后或误报等不良后果。文章立足于电力电容器运行和维护的实际需求,构建了一套完整的电力电容器故障在线监测系统,并给出了一套采用神经网络融合电容器电流、电容、电阻和电压等信息的故障诊断模型和方法,并在实际中进行了应用。在实际应用中,该系统能及时并准确地对电容器的异常状态和故障特征进行捕捉,避免了故障判断的滞后性,提高了获得数据的准确性,能够提高电网设备的运行和维护效率,提升电网运行可靠性。     

A spiking neural network-based approach to bearing fault diagnosis

Fault diagnosis, with the aim of accurately identifying the presence of various faults as early as possible so at to provide effective information for maintenance planning, has been extensively concerned in advanced manufacturing systems. With the increase of the amount of condition monitoring data, fault diagnosis methods have gradually shifted from the model-based paradigm to data-driven paradigm. Intelligent fault diagnosis approaches which can automatically mine useful information from a huge amount of raw data are becoming promising ways to identify faults of manufacturing systems in the context of massive data. In this paper, the Spiking Neural Network (SNN), as the third generation neural network, is tailored as an intelligent fault diagnosis tool for bearings in rotating machinery. Compared to the perceptron and the back propagation neural network (BPNN) which are respectively the first and second generations of neural networks. SNN, which introduces the concept of time into its operating model can more closely mimic natural neural networks and possesses high bionic characteristics. In the proposed SNN-based approach to bearing fault diagnosis, features extracted from raw vibration signals through the local mean decomposition (LMD) are encoded into spikes to train an SNN with the improved tempotron learning rule. The performance of the proposed method is examined by the CWRU and MFPT datasets, and the experimental results show that the method can achieve a promising accuracy in bearing fault diagnosis.     

基于卷积神经网络的机械故障诊断技术综述

针对传统机械故障诊断方法难以解决人工提取不确定性的问题,提出了大量深度学习的特征提取方法,极大地推动了机械故障诊断的发展。作为深度学习的典型代表,卷积神经网络（CNN）在图像分类、目标检测、图像语义分割等领域都取得了重大的发展,在机械故障诊断领域也有大量文献发表。为了进一步了解利用CNN的方法进行机械故障诊断的问题,首先简单介绍了CNN的相关理论,然后从数据输入类型、迁移学习、预测等方面对CNN在机械故障诊断中的应用进行了归纳总结,最后展望了CNN及其在机械故障诊断应用中的发展方向。     

一种基于时序预报神经网络的故障预报方法及其应用

提出一种基于时序预报神经网络的工业过程故障预报方法,同时给出了描述神经网络预 报和外推能力的表达方式,并以氯碱电解工艺的现场数据验证了这种故障预报方法的有效性. 实验结果表明,该方法可成功地用以实现氯中含氢的24小时预报.     

Connectionist approach to time series prediction: an empirical test

Among the various potential applications of neural networks, forecasting is considered to be a major application. Several researchers have reported their experiences with the use of neural networks in forecasting, and the evidence is inconclusive. This paper presents the results of a forecasting competition between a neural network model and a Box-Jenkins automatic forecasting expert system. Seventy-five series, a subset of data series which have been used for comparison of various forecasting techniques, were analysed using the Box-Jenkins approach and a neural network implementation. The results show that the simple neural net model tested on this set of time series could forecast about as well as the Box-Jenkins forecasting system.     

Adaptative time constants improve the prediction capability of recurrent neural networks

Classical statistical techniques for prediction reach their limitations in applications with nonlinearities in the data set; nevertheless, neural models can counteract these limitations. In this paper, we present a recurrent neural model where we associate an adaptative time constant to each neuron-like unit and a learning algorithm to train these dynamic recurrent networks. We test the network by training it to predict the Mackey-Glass chaotic signal. To evaluate the quality of the prediction, we computed the power spectra of the two signals and computed the associated fractional error. Results show that the introduction of adaptative time constants associated to each neuron of a recurrent network improves the quality of the prediction and the dynamical features of a neural model. The performance of such dynamic recurrent neural networks outperform time-delay neural networks.     

Temporal Bayesian Network of Events for Diagnosis and Prediction in Dynamic Domains

In some domains like industry, medicine, communications, speech recognition, planning, tutoring systems, and forecasting; the timing of observations (symptoms, measures, test, events, as well as faults) play a major role in diagnosis and prediction. This paper introduces a new formalism to deal with uncertainty and time using Bayesian networks called Temporal Bayesian Network of Events (TBNE). In a TBNE each node represents an event or state change of a variable, and an arc corresponds to a causal-temporal relationship. A temporal node represents the time that a variable changes state, including an option of no-change. The temporal intervals can differ in number and size for each temporal node, so this allows multiple granularity. Our approach is contrasted with a Dynamic Bayesian network for a simple medical example. An empirical evaluation is presented for a subsystem of a thermal power plant, in which this approach is used for fault diagnosis and event prediction with good results. The TBNE model can be used for the diagnosis of a cascade of anomalies arising with certain delays; this situation is typical in the diagnosis of medical and industrial processes.     

一种神经网络预测器在传感器故障诊断中的应用

讨论基于神经网络预测器的传感器故障诊断问题.介绍了传感器故障诊断技术的发展,提出了一种基于神经网络在线学习的传感器故障实时诊断的模型.通过比较三种前馈神经网络的预测残差确定网络类型.介绍了网络的学习规则,说明了在线学习的能力.最后,通过电厂高加热器的几个温度传感器的实际数据为例说明了此方法的实效性.     

一种新颖的深度因果图建模及其故障诊断方法

为了实现复杂工业过程故障检测和诊断一体化建模, 提出了一种新颖的深度因果图建模方法. 首先, 利用循环神经网络建立深度因果图模型, 将Group Lasso稀疏惩罚项引入到模型训练中, 自动地检测过程变量间的因果关系. 其次, 利用模型学习到的条件概率预测模型对每个变量建立监测指标, 并融合得到综合指标进行整体工业过程故障检测. 一旦检测到故障, 对故障样本构建变量贡献度指标, 隔离故障相关变量, 并通过深度因果图模型的局部因果有向图诊断故障根源, 辨识故障传播路径. 最后, 通过田纳西?伊斯曼过程进行仿真验证, 实验结果验证了所提方法的有效性.     

Short-term electric power load forecasting using feedforward neural networks

Abstract: This paper presents the results of a study on short‐term electric power load forecasting based on feedforward neural networks. The study investigates the design components that are critical in power load forecasting, which include the selection of the inputs and outputs from the data, the formation of the training and the testing sets, and the performance of the neural network models trained to forecast power load for the next hour and the next day. The experiments are used to identify the combination of the most significant parameters that can be used to form the inputs of the neural networks in order to reduce the prediction error. The prediction error is also reduced by predicting the difference between the power load of the next hour (day) and that of the present hour (day). This is a promising alternative to the commonly used approach of predicting the actual power load. The potential of the proposed method is revealed by its comparison with two existing approaches that utilize neural networks for electric power load forecasting.     

Pygmalion: ESPRIT II project 2059, neurocomputing

An overview is given of Pygmalion, which aims to promote European industry's application of neural networks and develop `standard' computational tools for their programming and simulation. A complete environment for developing algorithms and applications will demonstrate the network capabilities expected from their properties of massive parallelism, fault tolerance, adaptivity, and learning. Key real-world applications in image processing and speech processing and a small application in acoustic signals were selected to demonstrate the potential of neural networks for various industrial problems. In image processing, remote data sensing and factory inspection were investigated. In speech processing, the foundations were laid for an automatic speech recognition system by developing efficient learning algorithms for the basic building blocks     

基于进化神经网络的短期电网负荷预测算法

本文提出了一种基于进化神经网络的短期电网负荷预测算法。该算法使用改进的人工蜂群算法与BP神经网络融合生成进化神经网络,然后使用改进的人工蜂群算法对进化神经网络的偏置和权重进行优化。该算法将火电历史负荷数据作为输入,使用进化神经网络训练预测模型,预测未来一段时间内的电网负荷。首先,获取历史负荷数据。然后,将获取到的数据输入到进化神经网络模型中进行训练。在训练过程中,采用了改进的人工蜂群算法对进化神经网络对神经网络的权重和偏置进行优化,提高模型的预测精度。人工蜂群算法作为一种全局搜索算法,可以有效地探索模型参数空间,找到最优的模型参数组合,从而提高模型的预测精度。为了验证所提出的负荷预测方法的有效性,我们使用了火电网负荷数据进行了测试。实验结果表明本文提出的进化神经网络在短期电网负荷预测方面表现出了良好的预测精度和实用性。与传统的预测方法相比,该算法的预测误差更小,预测结果更加准确可靠。     

Neural networks and traditional time series methods: a synergisticcombination in state economic forecasts

Ever since the initial planning for the 1997 Utah legislative session, neural-network forecasting techniques have provided valuable insights for analysts forecasting tax revenues. These revenue estimates are critically important since agency budgets, support for education, and improvements to infrastructure all depend on their accuracy. Underforecasting generates windfalls that concern taxpayers, whereas overforecasting produces budget shortfalls that cause inadequately funded commitments. The pattern finding ability of neural networks gives insightful and alternative views of the seasonal and cyclical components commonly found in economic time series data. Two applications of neural networks to revenue forecasting clearly demonstrate how these models complement traditional time series techniques. In the first, preoccupation with a potential downturn in the economy distracts analysis based on traditional time series methods so that it overlooks an emerging new phenomenon in the data. In this case, neural networks identify the new pattern that then allows modification of the time series models and finally gives more accurate forecasts. In the second application, data structure found by traditional statistical tools allows analysts to provide neural networks with important information that the networks then use to create more accurate models. In summary, for the Utah revenue outlook, the insights that result from a portfolio of forecasts that includes neural networks exceeds the understanding generated from strictly statistical forecasting techniques. In this case, the synergy clearly results in the whole of the portfolio of forecasts being more accurate than the sum of the individual parts.     

Neural networks and statistical techniques: A review of applications

Neural networks are being used in areas of prediction and classification, the areas where statistical methods have traditionally been used. Both the traditional statistical methods and neural networks are looked upon as competing model-building techniques in literature. This paper carries out a comprehensive review of articles that involve a comparative study of feed forward neural networks and statistical techniques used for prediction and classification problems in various areas of applications. Tabular presentations highlighting the important features of these articles are also provided. This study aims to give useful insight into the capabilities of neural networks and statistical methods used in different kinds of applications.