Permanent magnet synchronous machines have gained popularity in wind turbines due to their merits of high efficiency, power density, and reliability. The wind turbines normally work in a wide range ...of operations, and harsh environments, so unexpected faults may occur and result in productivity losses. The common faults in the permanent magnet machines occur in the bearing and stator winding, being mainly detected in steady-state operating conditions under constant loads and speeds. However, variable loads and speeds are typical operations in wind turbines and powertrain applications. Therefore, it is important to detect bearing and stator winding faults in variable speed and load conditions. This paper proposes an algorithm to diagnose multiple faults in variable speed and load conditions. The algorithm is based on tracking the frequency orders associated with faults from the normalised order spectrum. The normalised order spectrum is generated by resampling the measured vibration signal via estimated motor speeds. The fault features are then generated from the tracking orders in addition to the estimated torque and speed features. Finally, support vector machine algorithm is used to classify the faults. The proposed method is validated using experimental data, and the validated results confirm its usefulness for practical applications.
This article proposes an active learning scheme to detect multiple faults in permanent magnet synchronous motors in dynamic operations without using historical labelled faulty training data. The ...proposed method combines the self-supervised anomaly detector based on a local outlier factor (LOF and a deep Q-network (DQN) supervised reinforcement learner to classify interturn short-circuit, local demagnetisation, and mixed faults. The first fault, which is detected by LOF and verified by an expert during maintenance, is used as training data for the DQN classifier. From that point onward, the LOF anomaly detector and DQN fault classifiers are working in tandem in the identification of new faults, which require expert intervention when either of them identifies a fault. The robustness of the scheme against dynamic operations, mixed fault, and imbalanced training datasets is validated via a comparative study using stray flux data from an in-house test setup.
This paper aims to classify local demagnetisation and inter-turn short-circuit (ITSC) on position sensorless permanent magnet synchronous motors (PMSM) in transient states based on external stray ...flux and learning classifier. Within the framework, four supervised machine learning tools were tested: ensemble decision tree (EDT), k-nearest neighbours (KNN), support vector machine (SVM), and feedforward neural network (FNN). All algorithms are trained on datasets from one operational profile but tested on other different operation profiles. Their input features or spectrograms are computed from resampled time-series data based on the estimated position of the rotor from one stray flux sensor through an optimisation problem. This eliminates the need for the position sensors, allowing for the fault classification of sensorless PMSM drives using only two external stray flux sensors alone. Both SVM and FNN algorithms could identify a single fault of the magnet defect with an accuracy higher than 95% in transient states. For mixed faults, the FNN-based algorithm could identify ITSC in parallel-strands stator winding and local partial demagnetisation with an accuracy of 87.1%.
Permanent magnet synchronous motors become popular in wind turbines and industrial applications. In critical machines, it is necessary to use robust condition monitoring and fault diagnosis ...algorithms to prevent faults or shutdowns. The data-driven approach with machine learning algorithms is widely used in industrial and research communities as this method does not require a mathematical model of the system, which is difficult to obtain in practical cases. Most of the successful machine learning methods are based on supervised learning approach, requiring labelled training data. The supervised learning approach cannot use the unlabelled data, while only a few labelled data is in place in the industry. This work uses a deep autoencoder based unsupervised learning method to identify the features of the fault classification algorithm in a self-supervised way, which overcome the shortage of labelled data. The proposed algorithm uses the benefits of available unlabelled data, but it needs only a few labelled data. The fault classification algorithm is based on artificial neural network SoftMax layer and Bayes classifier. The robustness of the algorithm is improved by fusing the current and vibration information. Experimental results are used to validate the robustness of proposed algorithms under noise conditions, and the results show that the algorithm could classify faults robustly.
In this paper. the method of direct torque control in the presence of a sliding-mode speed controller is proposed for a small wind turbine being used in water heating applications. This concept and ...control system design can be expanded to grid connected or off-grid applications. Direct torque control of electrical machines has shown several advantages including very fast dynamics torque control over field-oriented control. Moreover. the torque and flux controllers in the direct torque control algorithms are based on hvsteretic controllers which are nonlinear. In the presence of a sliding-mode speed control. a nonlinear control system can be constructed which is matched for AC DC conversion of the converter that gives fast responses with low overshoots. The main control objectives of the proposed small wind turbine can be maximum power point tracking and soft-stall power control. This small wind turbine consists of permanent magnet synchronous generator and external wind speed. and rotor speed measurements are not required for the system. However. a sensor is needed to detect the rated wind speed overpass events to activate proper speed references for the wind turbine. Based on the low-cost design requirement of small wind turbines. an available wind speed sensor can be modified. or a new sensor can be designed to get the required measurement. The simulation results will be provided to illustrate the excellent performance of the closed-loop control system in entire wind speed range (4-25 m s).
Machine learning based fault diagnosis schemes have been intensively proposed to deal with faults diagnosis of rotating machineries such as gearboxes, bearings, and electric motors. However, most of ...the machine learning algorithms used in fault diagnosis are pattern recognition tools, which can classify given data into two or more classes. The underlined physical phenomena in fault diagnosis are not directly interpretable in machine learning schemes, thus it is usually called black/gray box models. In this study, convolutional neural networks (CNN) machine learning algorithm is proposed to classify gearbox faults, and the learning features of the CNN filters are visualized to understand the physical fault diagnosis phenomena. Within the framework, a detailed explanation is given based on domain knowledge of gearbox fault diagnosis, and the physical phenomena are explained by fault characteristic frequencies, allowing for observing the relationship between the characteristic frequencies and the CNN feature learning filters. The proposed algorithm is validated using an in-house experimental setup.
Electric powertrains are widely used in automotive and renewable energy industries. Reliable diagnosis for defects in the critical components such as bearings, gears and stator windings, is important ...to prevent failures and enhance the system reliability and power availability. Most of existing fault diagnosis methods are based on specific characteristic frequencies to single faults at constant speed operations. Once multiple faults occur in the system, such a method may not detect the faults effectively and may give false alarms. Furthermore, variable speed operations render a challenge of analysing nonstationary signals. In this work, a deep learning-based fault diagnosis method is proposed to detect common faults in the electric powertrains. The proposed method is based on pattern recognition using convolutional neural network to detect effectively not only single faults at constant speed but also multiple faults in variable speed operations. The effectiveness of the proposed method is validated via an in-house experimental setup.
Online condition monitoring and fault diagnosis systems are necessary to prevent unexpected downtimes in critical electric powertrains. The machine learning algorithms provide a better way to ...diagnose faults in complex cases, such as mixed faults and/or in variable speed conditions. Most of studies focus on training phases of the machine learning algorithms, but the development of the trained machine learning algorithms for an online diagnosis system is not detailed. In this study, a complete procedure of training and implementation of an online fault diagnosis system is presented and discussed. Aspects of the development of an online fault diagnosis based on machine learning algorithms are introduced. A developed fault diagnosis system based on the presented procedure is implemented on an in-house test setup and the reliably detected results suggest that such a system can be widely used to predict multiple faults in the power drivetrains under variable speeds online.
Online bearing fault detection is an important method for monitoring the health status of bearings in critical machines. This work proposes a classification algorithm, which can be extended towards ...an online bearing fault detection. The objective is to detect and classify the bearing faults in early stages. The overall design aspects of the online bearing fault detection and classification system are discussed. The proposed method is validated using experimental data, and a high accuracy of the fault classification was observed. Therefore, the proposed method can be applied for an online early fault detection and classification system.
Bearings are one of the most critical elements in rotating machinery systems. Bearing faults are the main reason for failures in electrical motors and generators. Therefore, early bearing fault ...detection is very important to prevent critical system failures in the industry. In this paper, the support vector machine algorithm is used for early detection and classification of bearing faults. Both time and frequency domain features are used for training the support vector machine learning algorithm. The trained classier can be employed for real-time bearing fault detection and classification. By using the proposed method, the bearing faults can be detected at early stages, and the machine operators have time to take preventive action before a large-scale failure. The usefulness of the algorithm is validated by using a run-to-failure experimental test data.