The limited fault information caused by small fault data samples is a major problem in wind turbine (WT) fault detection. This paper proposes a small-sample WT fault detection method with the synthetic fault data using generative adversarial nets (GANs). First, based on the prior knowledge, a rough fault data generation process is developed to transform the normal data to the rough fault data. Second, a data refiner is developed by GANs to make the rough fault data more similar with the real fault data. Moreover, to make the generated data better suited to the WT conditions, GANs are improved in both the generative model and the discriminative model. Third, AI-based WT fault detection models can be well trained by sufficient generated data. Finally, three groups of generated data evaluation experiments and four groups of WT fault detection comparative experiments are conducted. The results indicate that the proposed method is effective.

Fault detection plays an essential role in the power generation of wind turbines (WTs). However, most of the present fault detection methods are designed to detect a certain type of fault.When these methods are used to detect multiple fault types, their accuracies tend to decline. In this paper, an effective multi-fault detection method based on improved triplet loss is proposed. First, a hard sample mining method is proposed based on WT data, so that the samples which are more suitable for triplet training are selected. Second, triplet training is improved by applying a new mapping layer and optimizing the loss function. Third, based on the improved triplet training, a deep learning model is trained to map the original samples into feature vectors in an embedding space, so that the distance between two samples in the same class is shortened whereas that of different classes is lengthened. Then, a multi-classification model is trained to distinguish the feature vectors, so that the faults of different types can be detected. Finally, four groups of experiments are conducted using real WT SCADA data collected from a wind farm in northern China. The experiment results show that the proposed method is effective.

Defect size quantification plays a vital role in pipeline magnetic flux leakage detection system. However, most existing methods suffer from poor applicability and low precision due to the complex industrial process. Moreover, they often overlook the valuable knowledge that may be beneficial for defect size quantification. To solve this problem, a defect size quantification method based on multilevel knowledge-guided neural network is proposed, which effectively integrates prior knowledge and specific data. First, at the feature level, a recursive residual subnet is proposed to inject the mechanism features into the network, so that the network performance can be enhanced. Second, an experience-aided subnet is proposed to incorporate expert experience at the decision level, which supervises the network training with labels, so that the network stability can be improved. Third, at the modeling level, a cascade expression subnet based on two-point representation is first proposed in the defect size quantification area, where both the value and distribution of labels are considered to boost the network precision. These three parts are jointly trained and promote each other. Finally, extensive experiments are conducted with defects from a pipeline network in northern China and the experimental results highlight the superiority of our method.

Condition monitoring (CM) has been playing an important role in the operation and maintenance of industrial equipment. However, the changeable environments where some equipment, such as wind turbines (WTs) are installed, have put a negative effect on CM. To deal with this problem, this paper proposes a CM method of WTs based on assembled multidimensional membership functions (MFs) using fuzzy inference system (FIS). First, multidimensional MFs, which are formed by fusing environment factor into conventional MFs, are proposed to reduce the negative effect from changeable environments. Second, the input data are properly divided into segments and the segments are classified into four types to distinguish the effects of different data on CM. Different weights are assigned to each segment and the corresponding membership of each segment is calculated separately. Then, these memberships are assembled for fuzzy inference. Finally, based on the assembled multidimensional MFs, a new CM architecture is established. Four groups of experiments were carried out to evaluate the proposed method with the data collected from a wind farm in northern China. The experiments results show that the proposed method can not only detect anomalies at an early stage but also effectively decrease the false alarms and missing detections.