This paper proposes a novel intelligent fault diagnosis method to automatically identify different health conditions of wind turbine gearbox. Different from traditional approaches where feature extraction and classification are separately designed and performed, this study aims to automatically learns effective fault features directly from raw vibration signals and meanwhile classifies the type of faults in a single learning framework, thus leading to an end-to-end learning based fault diagnosis system for wind turbine gearbox without additional signal processing techniques and diagnostic expertise. Considering multi-scale characteristics inherent in vibration signals of a gearbox, in this paper, a new multiscale convolutional neural network (MSCNN) framework is proposed to perform multi-scale feature extraction and classification simultaneously. A key advantage of the proposed MSCNN is the ability to learn complementary and rich fault pattern features at different time scales by incorporating multiscale learning into the CNN architecture, which greatly improves feature learning to enable better diagnosis performance. The proposed MSCNN framework is evaluated through experiments on a wind turbine test rig. The results demonstrate that the proposed MSCNN framework outperforms state-of-the-art methods and exhibits superior robustness against noise.

Data-driven approaches have gained increasing interests in the fault detection of wind turbines (WTs) due to the difficulty in system modeling and the availability of sensor data. However, the nonlinearity of WTs, uncertainty of disturbances and measurement noise, and temporal dependence in time-series data still pose grand challenges to effective fault detection. To this end, this paper proposes a new fault detector based on a recently developed unsupervised learning method, denoising autoencoder (DAE), which offers the learning of robust nonlinear representations from data against noise and input fluctuation. A DAE is used to build a robust multivariate reconstruction model on raw time-series data from multiple sensors, and then, the reconstruction error of the DAE trained with normal data is analyzed for fault detection. In addition, we apply the sliding-window technique to consider temporal information inherent in time-series data by including the current and past information within a small time window. A key advantage of the proposed approach is the ability to capture the nonlinear correlations among multiple sensor variables and the temporal dependence of each sensor variable simultaneously, which significantly enhanced the fault detection performance. Simulated data from a generic WT benchmark and field supervisory control and data acquisition data from a real wind farm are used to evaluate the proposed approach. The results of two case studies demonstrate the effectiveness and advantages of our proposed approach.

Currently, vibration analysis has been widely considered as an effective way to fulfill the fault diagnosis task of gearboxes in wind turbines (WTs). However, vibration signals are usually with abundant noise and characterized as nonlinearity and nonstationarity. Therefore, it is quite challenging to extract robust and useful fault features from complex vibration signals to achieve an accurate and reliable diagnosis. This paper proposes a novel feature representation learning approach, named stacked multilevel-denoising autoencoders (SMLDAEs), with the aim to learn robust and discriminative fault feature representations through a deep network architecture for diagnosis accuracy improvement. In our proposed approach, we design an MLD training scheme, which uses multiple noise levels to train AEs. It enables to learn more general and detailed fault feature patterns simultaneously at different scales from the complex frequency spectra of the raw vibration data, and therefore helps enhance the feature learning and fault diagnosis capability. Furthermore, SMLDAE-based fault diagnosis is performed with an unsupervised representation learning procedure followed by a supervised fine-tuning process with label information for classification. Our approach is evaluated by using the field vibration data collected from a self-designed WT gearbox test rig. The results show that our proposed approach learned more robust and discriminative fault feature representations and achieved the best diagnosis accuracy compared with the traditional approaches.

Identification of autism spectrum disorder (ASD) in children is challenging due to the complexity and heterogeneity of ASD. Currently, most existing methods mainly rely on a single modality with limited information and often cannot achieve satisfactory performance. To address this issue, this paper investigates from internal neurophysiological and external behavior perspectives simultaneously and proposes a new multimodal diagnosis framework for identifying ASD in children with fusion of electroencephalogram (EEG) and eye-tracking (ET) data. Specifically, we designed a two-step multimodal feature learning and fusion model based on a typical deep learning algorithm, stacked denoising autoencoder (SDAE). In the first step, two SDAE models are designed for feature learning for EEG and ET modality, respectively. Then, a third SDAE model in the second step is designed to perform multimodal fusion with learned EEG and ET features in a concatenated way. Our designed multimodal identification model can automatically capture correlations and complementarity from behavior modality and neurophysiological modality in a latent feature space, and generate informative feature representations with better discriminability and generalization for enhanced identification performance. We collected a multimodal dataset containing 40 ASD children and 50 typically developing (TD) children to evaluate our proposed method. Experimental results showed that our proposed method achieved superior performance compared with two unimodal methods and a simple feature-level fusion method, which has promising potential to provide an objective and accurate diagnosis to assist clinicians.

The supervisory control and data acquisition (SCADA) systems are widely installed on wind turbines (WTs) in major wind farms, which produce a large amount of sensory data that can be used for fault diagnosis of WTs. However, these SCADA data are naturally multivariate time series, which represent complex temporal correlations within each sensor variable and spatial correlations between different sensor variables. To effectively capture spatio-temporal correlations in SCADA data, we propose a new spatio-temporal multiscale neural network (STMNN). The proposed STMNN model contains two parallel feature extraction modules: first, a multiscale deep echo state network module to extract temporal multiscale features; second, a multiscale residual network module to extract spatial multiscale features. Additionally, as the WTs are in a normal working state most of the time, there are a large amount of normal SCADA data and few failure SCADA data. To address the data imbalance problem of SCADA data and enhance the fault diagnosis performance, instead of cross-entropy loss, the STMNN model adopts focal loss as loss function. Our proposed STMNN method can provide an end-to-end fault diagnosis solution with imbalanced SCADA data, and is evaluated through experiments on an SCADA dataset from a real wind farm. The experimental results and comparative analysis have proved the effectiveness of our proposed STMNN model in practical applications.