Intelligent fault diagnosis of bearings has been a heated research topic in the prognosis and health management of rotary machinery systems, due to the increasing amount of available data collected by sensors. This ha...
详细信息
Intelligent fault diagnosis of bearings has been a heated research topic in the prognosis and health management of rotary machinery systems, due to the increasing amount of available data collected by sensors. This has given rise to more and more business desire to apply data-driven methods for health monitoring of machines. In recent years, various deep learning algorithms have been adapted to this field, including multi-layer perceptrons, autoencoders, convolutional neural networks, and so on. Among these methods, autoencoder is of particular interest for us because of its simple structure and its ability to learn useful features from data in an unsupervised fashion. Previous studies have exploited the use of autoencoders, such as denoising autoencoder, sparsity aotoencoder, and so on, either with one layer or with several layers stacked together, and they have achieved success to certain extent. In this paper, a bearing fault diagnosis method based on fully-connected winner-take-all autoencoder is proposed. The model explicitly imposes lifetime sparsity on the encoded features by keeping only k% largest activations of each neuron across all samples in a mini-batch. A soft voting method is implemented to aggregate prediction results of signal segments sliced by a sliding window to increase accuracy and stability. A simulated data set is generated by adding white Gaussian noise to original signals to test the diagnosis performance under noisy environment. To evaluate the performance of the proposed method, we compare our methods with some state-of-the-art bearing fault diagnosis methods. The experiments result show that, with a simple two-layer network, the proposed method is not only capable of diagnosing with high precision under normal conditions, but also has better robustness to noise than some deeper and more complex models.
Over the past few decades, researchers have proposed various hyperspectral unmixing (HU) methods. Among these methods, deep learning (DL) has emerged as a promising approach for HU, providing new opportunities for adv...
详细信息
Over the past few decades, researchers have proposed various hyperspectral unmixing (HU) methods. Among these methods, deep learning (DL) has emerged as a promising approach for HU, providing new opportunities for advancement. However, accurately quantifying the presence of spectral variability factors within a mixture remains a challenging task. Therefore, numerous literatures have concerned the HU with spectral variability, in which the variation spectra are generated through the network. However, there is a lack of connection between the network and spectral variability, so they fail to provide physically meaningful interpretability of spectral variability. To this end, we use the physics-driven model to represent spectral variability and introduce it to the two-stream autoencoder unmixing network, resulting in improved endmember and abundance estimations. Specifically, the endmember extraction (EE) network learns spectral variability parameters associated with the dispersion model (DM) to generate the variations of spectra, which enhances the physical interpretability of endmember variability. In addition, the abundance estimation autoencoder network, tied to the EE network by shared weights, estimates abundances using the reconstructed hyperspectral image. Compared with the state-of-the-art HU approaches on three real hyperspectral image datasets, our method outperforms these techniques with improved unmixing accuracy, especially on endmember estimation.
A hard landing is a typical accident that occurs during the landing of an aircraft. Because hard landings can cause stress buildup in the aircraft structure that can lead to fatal accidents if not properly identified,...
详细信息
A hard landing is a typical accident that occurs during the landing of an aircraft. Because hard landings can cause stress buildup in the aircraft structure that can lead to fatal accidents if not properly identified, it is necessary to clearly determine whether a hard landing has occurred. However, erroneous judgments may occur because of the limitations of the existing methods of identifying hard landings. Although many studies have been conducted to reduce misjudgments, most existing approaches require the selection of proper key (flight) parameters or predefined thresholds, which require a great deal of experience and high-level professional knowledge. Therefore, in this study, a new model for identifying hard landings without explicit selection of key parameters or manual determination of thresholds is proposed by introducing an outlier detection technique. An autoencoder, an artificial neural network model, is applied to the detection of outliers from landing data obtained through high-fidelity landing simulation. The training of the autoencoder and performance analysis is conducted to demonstrate the validity of the proposed method. Normal landing data are used as the training dataset for the autoencoder, and the percentage of abnormal landing data are gradually increased to the training dataset to check the robustness of the proposed method. The performance analysis results showed that the proposed method applying the autoencoder can be successfully used to identify hard landing situations such as late flares, even if a small amount of abnormal data are included in the training dataset, as is the case for actual landing data.
We propose a transceiver design method based on an autoencoder (AE) network for multi-color visible light communication (VLC) systems. Taking into account the chromaticity constraint described by MacAdam ellipses and ...
详细信息
We propose a transceiver design method based on an autoencoder (AE) network for multi-color visible light communication (VLC) systems. Taking into account the chromaticity constraint described by MacAdam ellipses and the peak-value constraint of transmitted signals, the proposed AE network utilizes a peak-value constraint layer and an integrated loss function which different from previous AE designs. The new structure of AE network can be suitable for VLC systems with different numbers of colors. Additionally, noisy channel state information (CSI) is employed during the training of the AE in order to achieve a better performance for the system with imperfect CSI. After training, a transceiver design with the target of minimizing block error rate (BLER) can be obtained, which simultaneously meets the requirements of lighting. The results of numerical simulation experiments demonstrate that our proposed transceiver design outperforms conventional color shift keying (CSK) constellation design in imperfect CSI channel.
The paper presents the comparative analysis of the computer systems for face recognition. autoencoder, the typical representative of deep learning is compared with the classical PCA transformation. Both, autoencoder a...
详细信息
ISBN:
(纸本)9781538610404
The paper presents the comparative analysis of the computer systems for face recognition. autoencoder, the typical representative of deep learning is compared with the classical PCA transformation. Both, autoencoder and PCA serve as the tools for feature generation and selection. However, the important difference is the nonlinearity and multilayer structure applied in autoencoder. Final task of recognition is done by the support vector machine or softmax circuit. The numerical results performed on the multiclass base of faces have shown superiority of autoencoding principle, especially when the number of recognized classes is very high.
Process signals show the characteristics of large scale, high dimension, and strong correlation in modern industrial processes, which brings a big challenge for process fault detection and diagnosis. Due to the powerf...
详细信息
Process signals show the characteristics of large scale, high dimension, and strong correlation in modern industrial processes, which brings a big challenge for process fault detection and diagnosis. Due to the powerful feature learning ability, deep learning has been widely used in image and visual processing. This article proposes a new deep neural network (DNN), convolutional long short-term memory autoencoder (CLSTM-AE) for feature learning from process signals. The convolutional LSTM (ConvLSTM) is proposed to describe the distribution of the process data and learn effective features on time series data for fault detection. A selective residual block is embedded in the deep network to improve the training accuracy and perform feature selection from process signals. Two statistics, the T-2 and the squared prediction error (SPE), are generated in the feature space and residual space of CLSTM-AE, respectively. Finally, the feasibility and advantages of CLSTM-AE are shown on a simulated process, the Tennessee-Eastman process (TEP), and the continuous stirred tank reactor (CSTR). CLSTM-AE has good fault detection performance in these cases, which shows that it is capable of learning effective features from complex process signals. The hybrid learning technique with convolutional LSTM and autoencoder provides a new way for feature learning and fault detection for complex industrial processes.
Mutual bootstrapping is a commonly used technique for many natural language processing tasks, including semantic lexicon induction. Among many bootstrapping methods, the Basilisk algorithm achieved successful applicat...
详细信息
ISBN:
(纸本)9783030298944;9783030298937
Mutual bootstrapping is a commonly used technique for many natural language processing tasks, including semantic lexicon induction. Among many bootstrapping methods, the Basilisk algorithm achieved successful applications through two key iterative steps: scoring context patterns and candidate instances. In this work, we improve Basilisk by modifying its two scoring functions. By incorporating autoencoder to the scoring functions of patterns and candidates, we can reduce the bias problems and obtain more balanced results. The experimental results demonstrate that our proposed methods for guiding bootstrapping of a semantic lexicon with autoencoder can boost overall performance.
This paper investigates the application of autoencoder (AE) in supporting the training process of federated learning by reducing communication overhead and latency. We propose a scheduling algorithm to determine when ...
详细信息
ISBN:
(纸本)9798350300673
This paper investigates the application of autoencoder (AE) in supporting the training process of federated learning by reducing communication overhead and latency. We propose a scheduling algorithm to determine when and how to use autoencoder during training. Our simulation shows that federated learning with an autoencoder significantly reduces communication overhead without compromising testing accuracy. Moreover, the testing accuracy curve shows a more consistent increase over training rounds in federated learning with an autoencoder than in federated learning without an autoencoder. Additionally, the latency of federated learning with an autoencoder is lower than that of federated learning without an autoencoder.
Internal user threats such as information leakage or system destruction can cause significant damage to the organization, however it is very difficult to prevent or detect this attack in advance. In this paper, we pro...
详细信息
Internal user threats such as information leakage or system destruction can cause significant damage to the organization, however it is very difficult to prevent or detect this attack in advance. In this paper, we propose an anomaly-based insider threat detection method with local features and global statistics over the assumption that a user shows different patterns from regular behaviors during harmful actions. We experimentally show that our detection mechanism can achieve superior performance compared to the state of the art approaches for CMU CERT dataset.
High peak-to-average-power ratio (PAPR) and the LED nonlinearity have significant impacts on the performance of indoor Visible Light Communication (VLC) orthogonal frequency division multiplexing (OFDM) systems. In th...
详细信息
High peak-to-average-power ratio (PAPR) and the LED nonlinearity have significant impacts on the performance of indoor Visible Light Communication (VLC) orthogonal frequency division multiplexing (OFDM) systems. In this paper, we aim to improve the system performance by utilizing an end-to-end learning network. A novel PAPR reduction scheme is applied based on weighted autoencoder and amplitude clipping methods to address the high PAPR and LED nonlinearity problems. The constellation mapping and de-mapping of the transmitted symbols and phase factor of each subcarrier are adaptively acquired and optimized through the deep learning technique. How hyperparameters of network, network architecture and channel types affect the performance of bit error ratio (BER) and PAPR are firstly quantified in the asymmetrically clipped optical OFDM based VLC systems. Simulation results show that the hybrid autoencoder method achieves a distinct PAPR reduction of about 12 dB and is more robustness to LED nonlinearities leading to better BER performance compared to the standard methods.
暂无评论