检索结果-内蒙古大学图书馆

Anomaly detection and critical SCADA parameters identification for wind turbines based on LSTM-AE neural network

RENEWABLE ENERGY 2021年 172卷 829-840页

作者： Chen, Hansi Liu, Hang Chu, Xuening Liu, Qingxiu Xue, Deyi Shanghai Jiao Tong Univ Sch Mech Engn Shanghai Peoples R China Zhengzhou Univ Aeronaut Sch Management Engn Zhengzhou Henan Peoples R China Univ Calgary Dept Mech Engn Calgary AB Canada Univ Calgary Dept Mfg Engn Calgary AB Canada

Continuous monitoring of wind turbine health conditions using anomaly detection methods can improve the reliability and reduce maintenance costs during operation of wind turbine. Anomaly detection aims at identifying the root causes leading to unexpected changes of product performance. Most existing methods make less use of temporal order of the data and are poor at extracting features from these data. To address these problems, a method based on long short-term memory (LSTM) and auto-encoder (AE) neural network is introduced to assess sequential condition monitoring data of the wind turbine. First, a performance assessment model is constructed using LSTM neural units and AE networks to calculate the performance indices for evaluation of the degree of anomalies in wind turbine performance. Then, an adaptive threshold estimation method based on support vector regression model is developed to identify the abnormal data instances. The mutual information theory is subsequently explored to analyze the relationships between various monitoring parameters and performance abnormal instances to identify critical condition monitoring parameters. The effectiveness of the proposed method has been verified by a case study using real-world wind turbine condition monitoring (CM) data. (c) 2021 Elsevier Ltd. All rights reserved.

关键词： Wind turbine Anomaly detection Long short-term memory auto-encoder Mutual information theory

来源：评论

学校读者我要写书评

暂无评论

A working-condition-robust method for two-stage health measurement of rolling bearings based on energy entropy distribution and dynamic adversarial transfer network

引用

MEASUREMENT SCIENCE AND TECHNOLOGY 2023年第7期34卷 075020-075020页

作者： Jiang, Wei Xue, Xiaoming Zhang, Nan Xu, Yanhe Liu, Jie Shan, Yahui Huaiyin Inst Technol Jiangsu Key Lab Adv Mfg Technol Huaian 223003 Peoples R China Huazhong Univ Sci & Technol Sch Civil & Hydraul Engn Wuhan 430074 Peoples R China Wuhan Second Ship Design & Res Inst Wuhan Peoples R China

Accurate and robust health measurement for rolling bearings under variable working conditions has great significance in guaranteeing the safe and stable operation of rotating machinery. In this paper, a two-stage and working-condition-robust health measurement method is proposed, systematically blending energy entropy theory, a deep-learning approach and transfer-learning technology. In the first stage, a state boundary of energy entropy is systematically deduced based on an adaptive variational mode decomposition (VMD) improved fruit fly optimization algorithm (IFOA) and the principle of statistical analysis to detect abnormal states in bearings, where the IFOA is developed to search for the optimal parameters of the VMD with high efficiency. In the second stage, if a fault exists, a hybrid robust auto-encoder adopting a multi-layer and deep structure is constructed to strengthen the feature extraction capacity and automatically capture valuable and robust fault features from original samples. Considering the insufficiently labeled samples and significant data distribution discrepancy, a novel dynamic adversarial transfer network (DATN) is designed to extract the transferable and domain-invariant features between source and target datasets and achieve accurate fault identification. Specifically, a dynamic adversarial coefficient based on Wasserstein distance is provided in the DATN to quantitatively evaluate the relative importance of marginal and conditional distributions. Extensive experiments on two rolling bearing datasets validate the superior performance of the proposed method compared with other state-of-the-art identification models and transfer-learning approaches.

关键词： health measurement variable working conditions energy entropy auto-encoder adversarial transfer learning

来源：评论

学校读者我要写书评

暂无评论

Marine chlorophyll-a prediction based on deep auto-encoded temporal convolutional network model

引用

OCEAN MODELLING 2023年 186卷

作者： Ying, Chen Xiao, Li Zhao, Xueliang Song, Wenyang Xu, Chongxuan Yanshan Univ Sch Elect Engn Hebei Prov Key Lab Test Measurement Technol & Inst Qinhuangdao 066004 Hebei Peoples R China Geol Environm Monitoring Engn Technol Innovat Ctr Minist Nat Resources Ctr Hydrogeol & Environm Geol China Geol Survey Baoding 071051 Hebei Peoples R China

Chlorophyll-a concentration is one of the important indicators for assessing marine ecology. Accurate prediction of marine chlorophyll-a concentrations is a prerequisite for early warning of marine hazards such as red tide. There are many factors affecting chlorophyll-a concentration. Marine water quality parameter data as a time series is non-linear. There is a strong coupling relationship between various water quality parameters. This makes the prediction difficult. Therefore, there is a need to conduct research on marine chlorophyll-a concentration prediction methods. In this study, a depth-automatically encoded temporal convolutional network (DAE-TCN) is proposed to predict marine chlorophyll-a concentrations. Specifically, the DAE-TCN algorithm performs spatial feature extraction and data dimensionality reduction on the data through encoding to alleviate the difficulty of temporal feature extraction by temporal convolution, improve the efficiency of data information utilization, and obtain accurate prediction results. In this paper, we use the data from the coast of Beihai, Guangxi, as the experimental data. The DAE-TCN algorithm is compared with three baseline algorithms, and the results indicate that the algorithm shows advantages in stability and prediction accuracy and can be well applied to marine chlorophyll-a concentration prediction. This study has scientific significance for marine environmental protection.

关键词： Chlorophyll-a concentration auto-encoder Temporal convolutional network Temporal features Spatial features

来源：评论

学校读者我要写书评

暂无评论

Synthetic multi-view clustering with missing relationships and instances

引用

IMAGE AND VISION COMPUTING 2023年第1期134卷

作者： Zhang, Jing Wei, Guiyan Sun, Fang Liaoning Normal Univ Sch Comp & Informat Technol Dalian Liaoning Peoples R China

Multi-view learning not only excavates the supplement information thoroughly, but also avoids noisy features and redundancies to improve down-stream tasks, e.g. Cluster, by constructing the fusing model. However, in practices, it is difficult to obtain complete multi-view data, which is summarized as both of Partially Sample -Missing Problem (PSP) originating from any view and Partially View-unaligned Problem (PVP). SURE (robust multi-view clustering with incomplete information) builds up the novel loss function to remove the noisy sam-ples in the embedding space and presents the aligning method to solve both of PVP and PSP. However, generated instances and relationships contain relatively little discriminate information and semantic consistency, more-over, it needs a large size of complete samples to pre-train. Aiming to above problems, we propose the novel framework (termed Synthetic Multi-view Data Clustering, SMDC). For PSP problem, we induce feature -distillation in each feature-map level to generate features containing more semantic information helping with the guidance of complete feature-maps, furthermore, exploits self-distillation to strengthen the discriminability information of deeper levels supervised from the contrastive constraint in the embedding space. Moreover, for PVP problem, we leverage attention mechanism to adaptively fuse multi-view features and auto-align instances located in different views. The proposed model is verified in images and video clustering tasks, respectively. Comparing with popular methods achieve satisfactory performances. The code of model is published in: https://***/LNNU-computer-research-526/SMDC.(c) 2023 Published by Elsevier B.V.

关键词： Synthetic multi-view data Multi-view clustering auto-encoder Knowledge distillation

来源：评论

学校读者我要写书评

暂无评论

Triphasic DeepBRCA-A Deep Learning-Based Framework for Identification of Biomarkers for Breast Cancer Stratification

引用

IEEE ACCESS 2021年 9卷 103347-103364页

作者： Rajpal, Sheetal Agarwal, Manoj Kumar, Virendra Gupta, Anamika Kumar, Naveen Univ Delhi Dept Comp Sci New Delhi 110007 India Univ Delhi Dept Comp Sci Hans Raj Coll New Delhi 110007 India All India Inst Med Sci Dept Nucl Magnet Resonance Imaging New Delhi 110029 India Univ Delhi Dept Comp Sci Shaheed Sukhdev Coll Business Studies New Delhi 110089 India

Breast cancer being major death-leading cancer demands utmost attention. Recently, the next-generation sequencing techniques capable of capturing gene expression data have been used successfully for the detection of breast cancer. The proposed work identifies a small set of biomarker genes for molecular stratification of breast cancer subtypes. In this work, we have proposed Triphasic DeepBRCA - a novel deep learning framework, for breast cancer subtype detection and biomarker discovery. In the first phase, an autoencoder is used for extracting a compact representation of the gene expression data which is provided as an input to a supervised feed-forward neural network for classification of breast cancer subtypes in the second phase. In the third phase, the proposed Biomarker Gene Discovery Algorithm (BGDA) leverages the neural network classifier of the second phase to estimate the relevance of various genes. Next, Wilcoxon rank-sum test with False Discovery Rate (FDR) Correction is applied to identify the most differentiating genes. Using the TCGA BRCA RNASeq data, the proposed framework enabled us to discover a set of 54 most-variant genes. Using 10-fold cross-validation, we obtained a mean accuracy of 0.899 +/- 0.04 at 95% confidence interval. We also validated our results on METABRIC dataset. Gene Set Analysis revealed statistically enriched pathways. Heatmap of the expression levels and t-SNE visualization reveals that these genes have an aggregated capability to distinguish amongst the different breast cancer subtypes. Further, the prognostic evaluation using 54 biomarkers revealed that over 30 genes out of 54 are significantly linked to the prognostic outcome.

关键词： Breast cancer Biomarkers Gene expression Neural networks Deep learning Cancer Tools auto-encoder biomarker genes breast cancer subtype classification deep learning Innvestigate tool TCGA

来源：评论

学校读者我要写书评

暂无评论

Constrained representation learning for recurrent policy optimisation under uncertainty

引用

ADAPTIVE BEHAVIOR 2021年第3期29卷 253-265页

作者： Viet-Hung Dang Ngo Anh Vien Chung, TaeChoong DuyTan Univ Inst Res & Dev Da Nang Vietnam Queens Univ Belfast Sch EEECS Belfast Antrim North Ireland Kyung Hee Univ Dept Comp Engn Artificial Intelligent Lab Seoul South Korea

Learning to make decisions in partially observable environments is a notorious problem that requires a complex representation of controllers. In most work, the controllers are designed as a non-linear mapping from a sequence of temporal observations to actions. These problems can, in principle, be formulated as a partially observable Markov decision process whose policy can be parameterised through the use of recurrent neural networks. In this paper, we will propose an alternative framework that (a) uses the Long-Short-Term-Memory (LSTM) encoder-Decoder framework to learn an internal state representation for historical observations and then (b) integrates it into existing recurrent policy models to improve the task performance. The LSTM encoder encodes a history of observations as input into a representation of internal states. The LSTM Decoder can perform two alternative decoding tasks: predicting the same input observation sequence or predicting future observation sequences. The first proposed decoder acts like an auto-encoder that will guide and constrain the learning of a useful internal state for the policy optimisation task. The second proposed decoder decodes the learnt internal state by the encoder to predict future observation sequences. This idea makes the network act like a non-linear predictive state representation model. Both these decoding parts, which introduce constraints to policy representation, will help guide both the policy optimisation problem and latent state representation learning. The integration of representation learning and policy optimisation aims to help learn more complex policies and improve the performance of policy learning tasks.

关键词： Policy optimisation auto-encoder representation learning partially observable Markov decision process

来源：评论

学校读者我要写书评

暂无评论

Attenuation correction using deep learning for brain perfusion SPECT images

引用

ANNALS OF NUCLEAR MEDICINE 2021年第5期35卷 589-599页

作者： Sakaguchi, Kenta Kaida, Hayato Yoshida, Shuhei Ishii, Kazunari Kindai Univ Hosp Radiol Ctr 377-2 Ohnohigashi Osakasayama Osaka 5898511 Japan Kindai Univ Inst Adv Clin Med Div Positron Emiss Tomog Osaka 5898511 Japan Kindai Univ Fac Med Dept Radiol Osaka 5898511 Japan

Objective Non-uniform attenuation correction using computed tomography (CT) improves the image quality and quantification of single-photon emission computed tomography (SPECT). However, it is not widely used because it requires a SPECT/CT scanner. This study constructs a convolutional neural network (CNN) to generate attenuation-corrected SPECT images directly from non-attenuation-corrected SPECT images. Methods We constructed an auto-encoder (AE) using a CNN to correct the attenuation in brain perfusion SPECT images. SPECT image datasets of 270 (44,528 slices including augmentation), 60 (5002 slices), and 30 (2558 slices) cases were used for training, validation, and testing, respectively. The acquired projection data were reconstructed in three patterns: uniform attenuation correction using Chang's method (Chang-AC), non-uniform attenuation correction using CT (CT-AC), and no attenuation correction (No-AC). The AE learned an end-to-end mapping between the No-AC and CT-AC images. The No-AC images in the test dataset were loaded into the trained AE, which generated images simulating the CT-AC images as output. The generated SPECT images were employed as attenuation-corrected images using the AE (AE-AC). The accuracy of the AE-AC images was evaluated in terms of the peak signal-to-noise ratio (PSNR) and the structural similarity metric (SSIM). The intensities of the AE-AC and CT-AC images were compared by voxel-by-voxel and region-by-region analysis. Results The PSNRs of the AE-AC and Chang-AC images, compared using CT-AC images, were 62.2, and 57.9, and their SSIM values were 0.9995 and 0.9985, respectively. The AE-AC and CT-AC images were visually and statistically in good agreement. Conclusions The proposed AE-AC method yields highly accurate attenuation-corrected brain perfusion SPECT images.

关键词： Deep learning auto-encoder Brain perfusion SPECT Attenuation correction

来源：评论

学校读者我要写书评

暂无评论

Efficient degradation prediction of PEMFCs using ELM-AE based on fuzzy extension broad learning system

引用

APPLIED ENERGY 2023年 331卷

作者： Deng, Zhihua Chan, Siew Hwa Chen, Qihong Liu, Hao Zhang, Liyan Zhou, Keliang Tong, Sirui Fu, Zhichao Wuhan Univ Technol Sch Automation Wuhan 430070 Peoples R China Nanyang Technol Univ Sch Mech & Aerosp Engn Singapore 639798 Singapore Nanyang Technol Univ Energy Res Inst NTU ERIN Singapore 637141 Singapore Zhejiang Univ Coll Control Sci & Engn State Key Lab Ind Control Technol Hangzhou 310027 Peoples R China

Degradation prediction is an important part of proton exchange membrane fuel cells (PEMFCs) prognostic research, and it is essential for prolonging the durability of PEMFCs. An accuracy and efficient degradation prediction of PEMFCs method is proposed, which combines extreme learning machine based on auto-encoder (ELM-AE) and fuzzy extension broad learning system (FEBLS). The ELM-AE inherits the fast learning ability of ELM while retaining the dimensionality reduction and feature learning ability of AE. Specifically, the proposed FEBLS consists of three parts, the Takagi-Sugeno-Kang (TSK) fuzzy subsystems, enhancement layer, and extension layer. First, a set of TSK fuzzy subsystems are utilized to replace the mapping nodes of the existing broad learning system (BLS), where the latent representation of the power time series of PEMFCs are processed. Then, the outputs of all fuzzy subsystems are sent to the next layer for nonlinear transformation to obtain enhancement nodes. After that, the enhancement nodes are further fed into the latest layer for further nonlinear transformation to obtain the extension nodes. Finally, the accuracy and efficiency of the proposed ELM-AE-FEBLS method are verified by two different aging datasets of PEMFCs. Compared with ELM, ELM based on particle swarm optimization, and BLS methods, the comparison analysis shows that ELM-AE-FEBLS of multi-input variables with power time series has more competitive in accuracy. Therefore, the proposed method can accurately predict the remaining useful life of fuel cells.

关键词： Degradation Proton exchange membrane fuel cells Extreme learning machine auto-encoder Broad learning system Takagi-Sugeno-Kang fuzzy subsystems

来源：评论

学校读者我要写书评

暂无评论

Industrial users load pattern extraction method based on multidimensional electrical consumption feature construction

引用

FRONTIERS IN ENERGY RESEARCH 2023年 11卷

作者： Wang, Li Zhao, Jian Xia, Xiangwu Liu, Jun Lu, Yang Zhao, Lei Shanghai Univ Elect Power Coll Elect Engn Shanghai Peoples R China Shanghai Co Ltd China Coal Technol & Engn Grp Shanghai Peoples R China State Grid Zhejiang Elect Power Co Ltd Huzhou Power Supply Co Huzhou Peoples R China

The rapid development of renewable energy generation aggravates the imbalance between supply and demand in power grid, and exploring the potential of demand side resource can effectively improve such problems. Industrial users (IU) is an important demand response resource of power grid, and mining the load patterns of IU is the basis of studying the demand response ability of IU, which plays an important role in the safe operation and lean management of power grid. Lately, the popularity of advanced metering infrastructures provides data support for studying the load patterns of IU. However, the high dimensionality and the complex non-linear relationship of IU's load data bring difficulties to the task of clustering. To solve the above problems, this paper proposes a load pattern extraction method based on multidimensional electrical consumption feature construction. Firstly, industrial load characteristic set of IU is created with five load characteristic indexes weighted by improved entropy weight method. In addition, convolutional autoencoder is established to extract the temporal feature of industrial load data which is combined with industrial load characteristic set to build a multidimensional feature set (MFS) for IU and finish multidimensional electrical consumption feature construction (MECFC). Then, MFS is used as the input of Self-Organization Map network to select the initial clustering centers of K-means algorithm, overcoming the problem of local optimal solution, and complete the IU daily load clustering. The experiment shows that the algorithm based on MECFC solves the local optimal problem and have better performance in stability and clustering effect than traditional methods.

关键词： industrial user load clustering feature construction auto-encoder selforganizing map(SOM) k-means

来源：评论

学校读者我要写书评

暂无评论

Synthesized feature learning model on news aggregator for chatbot

引用

SOFT COMPUTING 2023年

作者： Verma, Prince Jyoti, Kiran Sahni, Varsha Ibrahim, S. Jafar Ali Chakravarthy, N. S. Kalyan IKG Punjab Tech Univ Jalandhar Punjab India Guru Nanak Dev Engn Coll Dept Comp Sci & Engn Ludhiana Punjab India Lovely Profess Univ Sch Comp Sci & Engn Jalandhar Punjab India Vellore Inst Technol Sch Comp Sci & Engn Dept Internet Things Vellore 632014 Tamil Nadu India QIS Coll Engn & Technol Ctr Data Sci Sch Comp Sci & Engn Ongole 523272 Andhra Pradesh India

The introduction of Deep learning over Shallow Learning has powered machines in the current era. The Conversational Software Agent (often referred to as a Chatbot) employs deep learning techniques to behave like a human. The paper has enlisted the various existing chatbots. The performance evaluation of research has shown that their stakeholders experience atrocious service. The initiation of significant and utilitarian feedback encounters an arduous and oppressive assignment as enactment for chatbots confides with buildout templates and predefined handwritten rules. Accordingly, subsisted chatbots are famous dearth in building responses and therefore contradict first-rate communication. So, a comprehensive understanding of such models may overcome this deficiency and replace the void with DNN. Stacked auto-encoder, sparse auto-encoders, predictive sparse, and denoising auto-encoders are some of the DNNs utilised for this till now. However, these DNNs cannot address the huge amounts of heterogeneous records. Tensor auto-encoder, which overcomes this disadvantage, is a time-consuming model. The paper will introduce a hybrid DL design that will overpower the drawbacks of the existing models and can add a benchmark to its performance.

关键词： Chatbot Deep learning (DL) Tensor flow Deep neural networks Feed-forward networks auto-encoder MLP DNN ANN CNN Feedback network RNN

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：