基于稀疏自动编码深度神经网络的感应电动机故障诊断

针对目前感应电动机故障诊断大多采用监督学习提取故障特征的现状,提出一种将去噪编码融入稀疏自动编码器的深度神经网络,实现非监督学习的特征提取并用于感应电动机的故障诊断。稀疏自动编码器通过自动学习复杂数据的内在特征来提取简明的数据特征表达。为提高特征表达的鲁棒性,在稀疏编码器的基础上融入去噪编码,提取更有效的特征表达用来训练神经网络分类器进而完成整个深度神经网络的构建,并结合反向传播算法对深度神经网络进行整体微调,提升故障分类的准确度。整个训练过程引入"dropout"训练技巧,减少因过拟合带来的预测误差。试验结果表明,相比传统反向传播(Back propagation,BP)神经网络,提出的深度神经网络能更有效地实现感应电动机故障诊断。     

Application of Dempster–Shafer theory in fault diagnosis of induction motors using vibration and current signals

This paper presents an approach for the fault diagnosis in induction motors by using Dempster–Shafer theory. Features are extracted from motor stator current and vibration signals and with reducing data transfers. The technique makes it possible for on-line application. Neural network is trained and tested by the selected features of the measured data. The fusion of classification results from vibration and current classifiers increases the diagnostic accuracy. The efficiency of the proposed system is demonstrated by detecting motor electrical and mechanical faults originated from the induction motors. The results of the test confirm that the proposed system has potential for real-time applications.     

Detection of a static eccentricity fault in a closed loop driven induction motor by using the angular domain order tracking analysis method

In this study, a new method was presented for the detection of a static eccentricity fault in a closed loop operating induction motor driven by inverter. Contrary to the motors supplied by the line, if the speed and load, and therefore the amplitude and frequency, of the current constantly change then this also causes a continuous change in the location of fault harmonics in the frequency spectrum. Angular Domain Order Tracking analysis (AD-OT) is one of the most frequently used fault diagnosis methods in the monitoring of rotating machines and the analysis of dynamic vibration signals. In the presented experimental study, motor phase current and rotor speed were monitored at various speeds and load levels with a healthy and static eccentricity fault in the closed loop driven induction motor with vector control. The AD-OT method was applied to the motor current and the results were compared with the traditional FFT and Fourier Transform based Order Tracking (FT-OT) methods. The experimental results demonstrate that AD-OT method is more efficient than the FFT and FT-OT methods for fault diagnosis, especially while the motor is operating run-up and run-down. Also the AD-OT does not incur any additional cost for the user because in inverter driven systems, current and speed sensor coexist in the system. The main innovative parts of this study are that AD-OT method was implemented on the motor current signal for the first time.     

A Deep Learning Approach for Fault Diagnosis of Induction Motors in Manufacturing

Extracting features from original signals is a key procedure for traditional fault diagnosis of induction motors, as it directly influences the performance of fault recognition. However, high quality features need expert knowledge and human intervention. In this paper, a deep learning approach based on deep belief networks (DBN) is developed to learn features from frequency distribution of vibration signals with the purpose of characterizing working status of induction motors. It combines feature extraction procedure with classification task together to achieve automated and intelligent fault diagnosis. The DBN model is built by stacking multiple-units of restricted Boltzmann machine (RBM), and is trained using layer-by-layer pre-training algorithm. Compared with traditional diagnostic approaches where feature extraction is needed, the presented approach has the ability of learning hierarchical representations, which are suitable for fault classification, directly from frequency distribution of the measurement data. The structure of the DBN model is investigated as the scale and depth of the DBN architecture directly affect its classification performance. Experimental study conducted on a machine fault simulator verifies the effectiveness of the deep learning approach for fault diagnosis of induction motors. This research proposes an intelligent diagnosis method for induction motor which utilizes deep learning model to automatically learn features from sensor data and realize working status recognition.     

A validated model for the prediction of rotor bar failure in squirrel-cage motors using instantaneous angular speed

Instantaneous angular speed (IAS)-based condition monitoring is an area in which significant progress has been achieved over the recent years. This condition monitoring technique is less known compared to the existing conventional methods. This paper presents model-predicted simulation and experimental results of broken rotor bar faults in a three-phase induction motor using IAS variations. The simulation was performed under normal, and a broken rotor bar fault. The present paper evaluates through simulating and measuring the IAS of an induction motor at broken rotor bar faults in both time and frequency domains. Experimental results show a good agreement with the model-predicted simulation results. Three vital key features were extracted from the angular speed variations. One feature is the modulating contour of pole pass frequency periods in time domain. The other two features are in frequency domain. The primary feature is the presence of the pole pass frequency component at the low-frequency region of the IAS spectrum. The secondary feature which are the multiple of pole pass frequency sideband components around the rotor speed frequency component. Experimental results confirm the validity of the simulation results for the proposed method. The IAS has demonstrated more sensitivity than current signature analysis in detecting the fault. This research also shows the power of angular speed features as a useful tool to detect broken rotor bar deteriorations using any economical transducer such as low-resolution rotary shaft encoders; which may well be already installed for process control purposes.     

An Intelligent Harmonic Synthesis Technique for Air-Gap Eccentricity Fault Diagnosis in Induction Motors

Induction motors (IMs) are commonly used in various industrial applications. To improve energy consumption efficiency, a reliable IM health condition monitoring system is very useful to detect IM fault at its earliest stage to prevent operation degradation, and malfunction of IMs. An intelligent harmonic synthesis technique is proposed in this work to conduct incipient air-gap eccentricity fault detection in IMs. The fault harmonic series are synthesized to enhance fault features. Fault related local spectra are processed to derive fault indicators for IM air-gap eccentricity diagnosis. The effectiveness of the proposed harmonic synthesis technique is examined experimentally by IMs with static air-gap eccentricity and dynamic air-gap eccentricity states under different load conditions. Test results show that the developed harmonic synthesis technique can extract fault features effectively for initial IM air-gap eccentricity fault detection.     

Diagnostics of stator faults of the single-phase induction motor using thermal images,MoASoS and selected classifiers

Single-phase induction motors are used in the industry commonly. Induction motors are not expensive, so it is a reason to use them. Diagnostics of faults is very important. It prevents financial loss and unplanned downtimes causes by faults. In this paper the authors described fault diagnostic techniques of the single-phase induction motor. Presented techniques were based on the analysis of thermal images of electric motor. The authors measured and analysed 3 states of the single-phase induction motor. In this paper an original method of the feature extraction of thermal images called MoASoS (Method of Area Selection of States) was presented. The proposed method - MoASoS and an image histogram were used to form feature vectors. Classification of the obtained vectors was performed by NN (Nearest Neighbour classifier) and Gaussian Mixture Models (GMM). The described fault diagnostic techniques are useful for reliability of the single-phase induction motors and other rotating electrical machines such as: three-phase induction motors, synchronous motors, DC motors.     

Detection of stator winding faults in induction motors using three-phase current monitoring

The objective of this paper is to propose a new method for the detection of inter-turn short circuits in the stator windings of induction motors. In the previous reported methods, the supply voltage unbalance was the major difficulty, and this was solved mostly based on the sequence component impedance or current which are difficult to implement. Some other methods essentially are included in the offline methods. The proposed method is based on the motor current signature analysis and utilizes three phase current spectra to overcome the mentioned problem. Simulation results indicate that under healthy conditions, the rotor slot harmonics have the same magnitude in three phase currents, while under even 1 turn (0.3%) short circuit condition they differ from each other. Although the magnitude of these harmonics depends on the level of unbalanced voltage, they have the same magnitude in three phases in these conditions. Experiments performed under various load, fault, and supply voltage conditions validate the simulation results and demonstrate the effectiveness of the proposed technique. It is shown that the detection of resistive slight short circuits, without sensitivity to supply voltage unbalance is possible.     

变频供电异步电动机转子断条故障诊断

异步电机转子断条故障发生时,定子电流（变频器输出侧电流）中会出现对称频率（1±2s）f1（f1为定子电流频率）的故障特征附加电流信号。以此为依据,定子电流特征频谱分析（MCSA）发展为经典转子断条故障在线检测方法。在工程实际过程中,变频供电异步电动机容易采集到的信号是开关柜二次侧供电电流（变频器输入侧电流）．因此要实现变频异步电动机转子断条故障诊断,必须清楚供电电流中是否也含有断条故障特征信息。首次对变频异步电动机供电电流进行分析．得出供电电流中也包括转子断条故障特征信息的结论,以此为基础。利用连续细化傅立叶和自适应滤波相结合的方法,实现了变频异步电动机转子断条故障诊断。     

Automatic progressive damage detection of rotor bar in induction motor using vibration analysis and multiple classifiers

There is an increased interest in developing reliable condition monitoring and fault diagnosis systems of machines like induction motors; such interest is not only in the final phase of the failure but also at early stages. In this paper, several levels of damage of rotor bars under different load conditions are identified by means of vibration signals. The importance of this work relies on a simple but effective automatic detection algorithm of the damage before a break occurs. The feature extraction is based on discrete wavelet analysis and autocorrelation process. Then, the automatic classification of the fault degree is carried out by a binary classification tree. In each node, comparing the learned levels of the breaking off correctly identifies the fault degree. The best results of classification are obtained employing computational intelligence techniques like support vector machines, multilayer perceptron, and the k-NN algorithm, with a proper selection of their optimal parameters.     

Fault diagnosis in six-phase induction motor using a current trajectory mass center

This paper proposes a new methodology for diagnosing the occurrence of stator winding faults in the six-phase induction motor. The proposed approach uses the xy current trajectory mass center of the motor stator currents. The Park transform is applied to the acquired induction motor stator currents. This transformation allows obtaining specific patterns that are used to identify stator winding faults. For healthy motors, a single point in the xy-plane is obtained. However, for a motor with a stator winding fault a circle is obtained, whose radius is related with the severity of the fault. To identify these patterns an algorithm, entitled current trajectory mass center, was developed. A theoretical analysis of the six-phase motor in αβ and xy current coordinates, for healthy and stator fault operation modes, is also presented. In order to show the applicability of the proposed technique several simulation and experimental results are presented.     

MODEL-BASED FAULT DIAGNOSIS OF INDUCTION MOTORS USING NON-STATIONARY SIGNAL SEGMENTATION

Effective detection and diagnosis of incipient faults is desirable for on-line condition assessment, product quality assurance and improved operational efficiency of induction motors running off the power supply mains. In this paper, an empirical model-based fault diagnosis system is developed for induction motors using recurrent dynamic neural networks and multiresolution signal processing methods. In addition to nameplate information required for the initial set-up, the proposed diagnosis system uses measured motor terminal currents and voltages, and motor speed. The effectiveness of the diagnosis system is demonstrated through staged motor faults of electrical and mechanical origin. The developed system is scalable to different power ratings and it has been successfully demonstrated with data from 2.2, 373 and 597 kW induction motors. Incremental tuning is used to adapt the diagnosis system during commissioning on an new motor, significantly reducing the system development time.     

Symmetrical component filter based on line condition monitoring instrumentation system for mine winder motor

As the capital investment in underground coal mining is huge enough by the standards of any conventional industry hence coal production process has to be very efficient to make commercially viable. In a situation of intensive and massive investment, the economics of production would primarily depend on machine utilization indicated by machine availability. Thus machine available time i.e. the time that a machine is available to do productive work, has to be maximized, for best returns on capital invested and utilization of manpower.In this present research work an online condition monitoring instrumentation system has been developed for condition monitoring of mine winder motor. The instrumentation system has been developed based on current monitoring technique. The symmetrical current component present in the unbalanced motor current is sensed with the help of current transformer, current to voltage converter, all pass filters and adders. Any electrical fault in mine winder motor will produce unbalancing in the motor circuit and will cause for the development of symmetrical current component. The type of electrical fault can be determined by sensing the symmetrical current component. One important advantage of this condition monitoring technique is that the instrument can be made hand held and the same hand held instrument may be used for the fault diagnosis of other motors also.A novel condition monitoring instrumentation system based on symmetrical component filter has been developed for on-line condition monitoring of mine winder motor. The instrumentation system would be able to diagnose various incipient faults of mine winder motor and will increase the safety as well as availability of mine winder.The result obtained from symmetrical current component filter based motor diagnostic technique has been verified with the result obtained by axial leakage flux based motor diagnostic technique for similar simulated motor fault condition to pinpoint the exact faulty of condition of the model mine winder motor.     

Hybrid algorithmic approach oriented to incipient rotor fault diagnosis on induction motors

This paper investigates the current monitoring for effective fault diagnosis in induction motor (IM) by using random forest (RF) algorithms. A rotor bar breakage of IM does not derive in a catastrophic fault but its timely detection can avoid catastrophic consequences in the stator or prevent malfunctioning of those applications in which this sort of fault is the primary concern. Current-based fault signatures depend enormously on the IM power source and in the load connected to the motor. Hence, homogeneous sets of current signals were acquired through multiple experiments at particular loading torques and IM feedings from an experimental test bench in which incipient rotor severities were considered. Understanding the importance of each fault signature in relation to its diagnosis performance is an interesting matter. To this end, we propose a hybrid approach based on Simulated Annealing algorithm to conduct a global search over the computed feature set for feature selection purposes, which reduce the computational requirements of the diagnosis tool. Then, a novel Oblique RF classifier is used to build multivariate trees, which explicitly learn optimal split directions at internal nodes through penalized Ridge regression. This algorithm has been compared with other state-of-the-art classifiers through careful evaluation of performance measures not encountered in this field.     

APPLICATION OF WAVELET PACKET ANALYSIS IN DRILL WEAR MONITORING

In this work, an attempt has been made to develop a drill wear monitoring system which is independent to cutting conditions of the drilling process. A cost effective Hall-effect current sensor, which does not interfere with the process, has been used for acquiring motor current signature during drilling under different cutting conditions. An advanced signal processing technique, the wavelet packet transform has been used on the acquired current signature to extract features for indirect representation to the amount of drill wear. Experimental sensitivity analysis reveals that in comparison to time domain features, wavelet packet features are more sensitive to flank wear and less sensitive to the cutting conditions. A multilayer neural network model has then been developed to correlate the extracted wavelet packet features with drill flank wear. Experimental results show that the proposed drill wear monitoring system can successfully predict the flank wear with acceptable accuracy.     

APPLICATION OF WAVELET PACKET ANALYSIS IN DRILL WEAR MONITORING

In this work, an attempt has been made to develop a drill wear monitoring system which is independent to cutting conditions of the drilling process. A cost effective Hall-effect current sensor, which does not interfere with the process, has been used for acquiring motor current signature during drilling under different cutting conditions. An advanced signal processing technique, the wavelet packet transform has been used on the acquired current signature to extract features for indirect representation to the amount of drill wear. Experimental sensitivity analysis reveals that in comparison to time domain features, wavelet packet features are more sensitive to flank wear and less sensitive to the cutting conditions. A multilayer neural network model has then been developed to correlate the extracted wavelet packet features with drill flank wear. Experimental results show that the proposed drill wear monitoring system can successfully predict the flank wear with acceptable accuracy.     

Novel approach using Hilbert Transform for multiple broken rotor bars fault location detection for three phase induction motor

This paper presents a new approach to detect the location of multiple broken rotor bars (MBRBs) in induction motor (IM) drive, running under no load and full load conditions using direct in and variable frequency drives. This technique is based on earlier work of location detection of one broken rotor bar. The techniques are tested for various fault severity levels so the detection of the exact location of the fault at early stage helps to reach sufficient time maintenance. In this paper, the authors used Hilbert Transform to extract the fault signature from the stator current envelope which is the low frequency component. Then statistical analysis is applied which produce a formula that is used to get the exact location of the fault in IM rotor.     

An on-line condition monitoring system for induction motors via instantaneous power analysis

Condition monitoring is an important factor in assuring well-being of motors. Existing approaches of condition monitoring are dependent on expensive special sensors. This paper reviews various forms of existing condition monitoring methods and highlights the need for an economical intelligent fault diagnosis system. In this study, the methodology taken in developing a condition monitoring system for the motor bearing faults identification, utilizing the commonly available motor stator current and voltage is demonstrated. This unique diagnostic condition monitoring system provides continuous real time tracking of the various bearing defects and determines the severity which can be adopted for fast decision making. The study on different bearing faults under no-load and full-load conditions is carried out experimentally and analyzed, and the results on the real hardware implementation confirm the effectiveness of the proposed approach.     

基于旋转不变信号参数估计技术与模式搜索算法的异步电动机转子故障检测新方法

提出一种基于旋转不变信号参数估计技术(Estimation of signal parameters via rotational invariance technique,ESPRIT)与模式搜索算法(Pattern search algorithm,PSA)的异步电动机转子故障检测新方法。模拟形成转子故障情况下的定子电流信号并以之检验ESPRIT性能。结果表明:即使对于短时信号,ESPRIT仍具备高频率分辨力,可以准确估计定子电流各个分量的频率;但对其幅值、初相角的估计欠缺准确性、稳定性。随后,采用PSA确定各个频率分量的幅值、初相角。对一台异步电动机完成了转子故障检测试验,结果表明:基于ESPRIT与PSA的异步电动机转子故障检测方法是切实可行的,并且因仅需短时信号即可达到高频率分辨力而适用于负荷波动情况。     

感应电机轴承故障检测方法研究

分析了感应电机轴承发生故障时的振动信号的特性,利用带通滤波器和希尔伯特变换,对感应电机轴承振动信号进行处理,然后采用高分辨率谱估计算法--MUSIC（Multiple Signal Classification）算法对包络信号作谱分析,再从包络信号的MUSIC谱中提取故障特征频率分量.研究结果表明,该方法频率分辨率更高,故障检测更为准确.将该方法应用于电机轴承故障诊断,可准确提取轴承故障特征分量.