基于BP网络的电力机车充电机故障诊断研究

充电机作为电力机车的一个重要装置,直接影响电力机车的安全运行.文章以BP神经网络故障诊断理论为基础,提出了以充电机输入和输出以及晶闸管触发信号为神经网络输入的特征参数,建立了充电机的23种故障模式和特征量之间非线性关系.仿真及实验结果表明了该诊断模型和算法的可行性以及有效性,同时特征参数的有效提取提高了诊断的精确性.     

Sensorless speed measurement using current harmonic spectralestimation in induction machine drives

This paper proposes a sensorless speed measurement scheme that improves the performance of transducerless induction machine drives, especially for low-frequency operation. Speed-related harmonics that arise from rotor slotting and eccentricity are analyzed using digital signal processing. These current harmonics exist at any nonzero speed and are independent of time-varying parameters, such as stator winding resistance. A spectral estimation technique combines multiple current harmonics to determine the rotor speed with more accuracy and less sensitivity to noise than analog filtering methods or the fast Fourier transform. An on-line initialization routine determines machine-specific parameters required for slot harmonic calculations. This speed detector, which has been verified at frequencies as low as 1 Hz, can provide robust, parameter-independent information for parameter tuning or as an input to a sensorless flux observer for a field-oriented drive. The performance of the algorithm is demonstrated over a wide range of inverter frequencies and load conditions     

电子设备的故障单元诊断系统设计

为故障精确定位而设计的电子设备故障单元诊断系统,利用VXI测试仪器为硬件平台,设计了基于征兆的模糊识别技术为软件推理平台。系统能根据需要对征兆数据进行获取,再利用基于VXI bus的模糊波形识剐技术进行故障的合理识别、推理和判断,将电子设备故障定位在最小可更换单元。     

基于DGA支持向量机的变压器故障诊断

针对基于DGA的变压器故障诊断方法在实际操作中存在的不足,提出两种解决方案:基于粒子群优化支持向量机的变压器故障诊断、基于差分进化支持向量机的变压器故障诊断。通过分析两种方案的算法原理建立支持向量机的变压器故障诊断模型,从而完成参数的优化,对得到的最优参数进行验证,获取最优的支持向量机模型。在Matlab软件平台上进行仿真实验,结果证明,采用基于粒子群优化支持向量机的变压器故障诊断结果获取的变压器故障诊断率较高;基于差分进化支持向量机的变压器故障诊断方法的误判率较低,全局寻优能力较好,相比于粒子群优化算法,差分进化支持向量机的优化精度更高。     

基于改进神经网络的机床刀具故障诊断

针对机床刀具的故障诊断系统进行研究,使用智能人工神经网络算法建立诊断模型。为了提高神经网络模型的训练效率,避免网络陷入局部最优解,使用一种改进的量子神经网络,将附加动量与自适应学习速度方法融合,提高网络收敛效率。使用五轴联动铣床进行刀具故障诊断识别。对声发射信号进行特征提取,使用总振铃技术、总能量、有效电压、事件计数、重心频率、均方根频率以及频率标准方差作为网络的输入向量,判别刀具为新刀、轻微磨损或严重磨损。实验结果表明,使用的改进的量子神经网络的效率以及识别准确度均高于常规BP神经网络。     

Fault diagnosis and its prediction in wireless sensor networks using regressional learning to achieve fault tolerance

Due to the wide range of critical applications and resource constraints, sensor node gives unexpected responses, which leads to various kind of faults in sensor node and failure in wireless sensor networks. Many research studies focus only on fault diagnosis, and comparatively limited studies have been conducted on fault diagnosis along with fault tolerance in sensor networks. This paper reports a complete study on both 2 aspects and presents a fault tolerance approach using regressional learning with fault diagnosis in wireless sensor networks. The proposed method diagnose the different types of faulty nodes such as hard permanent, soft permanent, intermittent, and transient faults with better detection accuracy. The proposed method follows a fault tolerance phase where faulty sensor node values would be predicted by using the data sensed by the fault free neighbors. The experimental evaluation of the fault tolerance module shows promising results with R² of more than 0.99. For the periodic fault such as intermittent fault, the proposed method also predict the possible occurrence time and its duration of the faulty node, so that fault tolerance can be achieved at that particular time period for better performance of the network.     

Improved diagnosis using enhanced fault dominance

Defect diagnosis can benefit from fault dominance relations to reduce the set of defect candidate sites. This paper presents new fault dominance collapsing operators that further reduce the set of candidates considered during the initial phase of diagnosis. In contrast to existing dominance-based methods which operate on pairs of faults, the proposed method operates on sets of faults. Fault-related entities are generated to guide the diagnosis process. The proposed collapsing operators can be used to accelerate effect-cause diagnosis. Experimental results demonstrate that the proposed method achieves a higher collapsing ratio than existing methods.     

Solid-state control of electric drives

A tutorial review of the dc and ac electric-drive field is presented. The goal is to present fundamental concepts, principle concerns, and key developments in electric-drive technology. Principles of ac and dc power converters and ac and dc motors are presented. Then the combination of the converter and motor to provide a complete drive system is discussed along with drive-system characteristics and methods for analyzing performance. Finally, some application guidelines for both ac and dc systems are given.     

Comparison of machine learning and traditional classifiers in glaucoma diagnosis

Glaucoma is a progressive optic neuropathy with characteristic structural changes in the optic nerve head reflected in the visual field. The visual-field sensitivity test is commonly used in a clinical setting to evaluate glaucoma. Standard automated perimetry (SAP) is a common computerized visual-field test whose output is amenable to machine learning. We compared the performance of a number of machine learning algorithms with STATPAC indexes mean deviation, pattern standard deviation, and corrected pattern standard deviation. The machine learning algorithms studied included multilayer perceptron (MLP), support vector machine (SVM), and linear (LDA) and quadratic discriminant analysis (QDA), Parzen window, mixture of Gaussian (MOG), and mixture of generalized Gaussian (MGG). MLP and SVM are classifiers that work directly on the decision boundary and fall under the discriminative paradigm. Generative classifiers, which first model the data probability density and then perform classification via Bayes' rule, usually give deeper insight into the structure of the data space. We have applied MOG, MGG, LDA, QDA, and Parzen window to the classification of glaucoma from SAP. Performance of the various classifiers was compared by the areas under their receiver operating characteristic curves and by sensitivities (true-positive rates) at chosen specificities (true-negative rates). The machine-learning-type classifiers showed improved performance over the best indexes from STATPAC. Forward-selection and backward-elimination methodology further improved the classification rate and also has the potential to reduce testing time by diminishing the number of visual-field location measurements.     

Induction machine fault detection using SOM-based RBF neural networks

A radial-basis-function (RBF) neural-network-based fault detection system is developed for performing induction machine fault detection and analysis. Four feature vectors are extracted from power spectra of machine vibration signals. The extracted features are inputs of an RBF-type neural network for fault identification and classification. The optimal network architecture of the RBF network is determined automatically by our proposed cell-splitting grid algorithm. This facilitates the conventional laborious trial-and-error procedure in establishing an optimal architecture. In this paper, the proposed RBF machine fault diagnostic system has been intensively tested with unbalanced electrical faults and mechanical faults operating at different rotating speeds. The proposed system is not only able to detect electrical and mechanical faults, but the system is also able to estimate the extent of faults.     

Current similarity based open-circuit fault diagnosis for induction motor drives with discrete wavelet transform

Three-phase voltage-source inverters fed induction motor with space-vector controlled scheme are widely used in industrial applications. The suffered failures will degrade the system performance with output torque ripple and harmonic currents. In this paper, a novel diagnosis method is proposed to detect and locate the insulated gate bipolar translator open-circuit fault. Discrete wavelet transform is used as a pre-treatment technique for three-phase output currents, the approximate coefficients are applied to obtain energy vectors. Euclidean distance between every two of the energy vectors are calculated for measuring the current similarity to diagnose fault. When IGBTs occur open-circuit fault, the values of Euclidean distance will be smaller than that under normal conditions, then faults can be detected. Faults can also be located according to extracted features. Simulation and experimental results show high efficiency and merits of the proposed method.     

局部均值分解与支持向量机相融合的传感器故障诊断

为了提高传感器故障诊断精度,提出一种基于局部均值分解和支持向量机相融合的传感器故障诊断算法。首先利用局部均值分解方法将传感器的输出信号分解成一系列由包络信号和纯调频信号相乘所得的PF分量;然后利用支持向量机进行故障识别;最后采用Matlab编程实现仿真对比实验。实验结果表明,该方法可以较好地识别传感器故障,提高了传感器故障诊断的稳定性和准确性。     

应用数据融合实现电子电路的故障诊断

在电路故障诊断中,可通过直流分析、交流分析和灵敏度分析等方法,对电路的故障进行诊断.但由于不同的诊断方法对不同的故障敏感度不同,使得每种方法都带有局限性.为此,本文提出了采用数据融合进行电路故障诊断的新方法,介绍了D-S证据理论算法在电路故障诊断中的应用,给出了具体算法和仿真实例.理论分析和仿真结果表明,将数据融合技术用于电路的故障诊断是可行的.不同的诊断方法提供的信息经多次融合、反复抽取有用信息后,大大降低了判断的盲目性,提高了电路故障诊断的准确性.     

基于NGA优化支持向量机的电力变压器故障诊断

为提高电力变压器故障诊断的准确性,提出一种支持向量机（Support Vector Machines,SVM）的故障诊断方法.该方法用添加最优保存策略的小生境策遗传算法对SVM进行参数优化,确保种群中适应度高的个体能被保留到下一代,使优化对象比较容易稳定,以得到更优良的个体,提高诊断精度.通过与遗传算法优化SVM及标准小生境遗传算法优化SVM的诊断结果相比较,根据对比结果表明：所提方法对变压器故障数据的分类辨识效果更好.     

Automatic CRP mapping using nonparametric machine learning approaches

This paper studies an uneven two-class unsupervised classification problem of satellite imagery, i.e., the mapping of U.S. Department of Agriculture's (USDA) Conservation Reserve Program (CRP) tracts. CRP is a nationwide program that encourages farmers to plant long-term, resource conserving covers to improve soil, water, and wildlife resources. With recent payments of nearly US $1.6 billion for new enrollments (2002 signup), it is imperative to obtain accurate digital CRP maps for management and evaluation purposes. CRP mapping is a complex classification problem where both CRP and non-CRP areas are composed of various cover types. Two nonparametric machine learning approaches, i.e., decision tree classifier (DTC) and support vector machine (SVMs) are implemented in this work. Specifically, considering the importance of CRP classification sensitivity, a new DTC pruning method is proposed to increase recall. We also study two SVM relaxation approaches to increase recall. Moreover, a localized and parallel framework is suggested in order to efficiently deal with the large-scale CRP mapping need. Simulation results validate the applicability of the suggested framework and proposed techniques.     

Automotive signal diagnostics using wavelets and machine learning

In this paper, we describe an intelligent signal analysis system employing the wavelet transformation in the solution of vehicle engine diagnosis problems. Vehicle engine diagnosis often involves multiple signal analysis. The developed system first partitions a leading signal into small segments representing physical events or states based on wavelet multi-resolution analysis. Second, by applying the segmentation result of the leading signal to the other signals, the detailed properties of each segment, including inter-signal relationships, are extracted to form a feature vector. Finally, a fuzzy intelligent system is used to learn diagnostic features from a training set containing feature vectors extracted from signal segments at various vehicle states. The fuzzy system applies its diagnostic knowledge to classify signals as abnormal or normal. The implementation of the system is described and experiment results are presented     

Detecting Byzantine attack in cognitive radio networks using machine learning

Wireless Networks - One primary function in a cognitive radio network (CRN) is spectrum sensing. In an infrastructure-based CRN, instead of individual nodes independently sensing the presence of...     

Soft fault diagnosis of analog circuits based on semi-supervised support vector machine

Analog Integrated Circuits and Signal Processing - Soft fault diagnosis has been validated as a very challenging problem in analog circuits. In order to improve the generalization ability and close...     

Guaranteed fault detection sequences for single transition faults in finite state machine models using concurrent fault simulation

This paper presents a new algorithm for the generation of test sequences for finite state machines. Test sequence generation is based on the transition fault model, and the generation of state-pair distinguishing sequences. We show that the use of state-pair distinguishing sequences generated from a fault-free finite state machine will remain a distinguishing sequence even in the presence of a single transition fault, thus guaranteeing complete single transition fault coverage. Analysis and experimental results show that the complexity of the test sequence generation algorithm is less than those of the previous algorithms. The utility of the transition fault model, and the generated test sequences is shown by their application to sequential logic circuits. These results show more than a factor of 10 improvement in the test generation time and some reduction in test length while maintaining 100% transition fault coverage.Now with Intel Corporation, FM5-161, 1900 Prairie City Road, Folsom, CA 95630.Now with Chrysalis Symbolic Design, 101 Billerica Ave., North Billerica, MA 01862.