相关参考文献
注意：仅列出部分参考文献，下载原文获取全部文献信息。
Article
Engineering, Electrical & Electronic
Xiuyong Yu et al.
Summary: In this paper, a semi-supervised learning model based on generative adversarial networks (GAN) is proposed to reliably identify faulty feeders. By revising the loss functions and playing a maximum and minimum game, the proposed method significantly improves performance when only a small percentage of training samples are labeled.
INTERNATIONAL JOURNAL OF ELECTRICAL POWER & ENERGY SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Kaiyang Chu et al.
Summary: Fault source diagnosis methodology is crucial for quality control and assurance in multi-source and multi-stage manufacturing processes, particularly in small sample manufacturing systems. This study proposes a Bayesian network-based methodology by analyzing existing research on fault source diagnosis methods. The use of gray correlation theory and mechanism analysis in the construction of the Bayesian network model reduces the reliance on sample data size for structure learning in complex product manufacturing with small samples. Additionally, two fault source diagnosis methods based on manufacturing principle analysis and reverse Bayesian network are introduced, along with a strategy for their combined use in real manufacturing scenarios.
ADVANCED ENGINEERING INFORMATICS
(2023)
Article
Thermodynamics
Jiachi Yao et al.
Summary: Accurate estimation of lithium-ion battery capacity is crucial. However, obtaining cycling data during charging/discharging processes can be challenging. To address this issue, a deep transfer learning method is proposed that uses partial segments of charging/discharging data for battery capacity estimation.
Review
Computer Science, Artificial Intelligence
Tao Zhou et al.
Summary: This paper summarizes the basic principles and training process of the Generative Adversarial Network (GAN) models, and also explores the application and advantages of GAN models in the field of medical image fusion. In addition, it discusses the challenges faced by GAN and provides future prospects.
INFORMATION FUSION
(2023)
Article
Computer Science, Artificial Intelligence
Mattia Segu et al.
Summary: Domain generalization aims to train machine learning models that can perform robustly across different domains. This study proposes a new approach that explicitly trains domain-dependent representations and maps domains in a shared latent space, achieving better generalization performance.
PATTERN RECOGNITION
(2023)
Article
Engineering, Mechanical
Qi Zhong et al.
Summary: A semi-supervised learning method based on multi-sensor information fusion is proposed to diagnose faults of the hydraulic directional valve in operation. The method trains a classification model with a small amount of data with fault labels and generates pseudo labels for unmarked data. The article introduces a multi-sensor fusion algorithm to obtain high-confidence pseudo labels and an adaptive threshold model to generate thresholds for selecting pseudo labels. The experimental results demonstrate high accuracy in hydraulic valve fault diagnosis, providing a new idea for developing intelligent hydraulic valves.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2023)
Article
Computer Science, Artificial Intelligence
Kanghyeok Ko et al.
Summary: Image-to-image translation with GANs is an active research area. StarGAN stands out by achieving multiple domain translation with a single generator, but it has limitations in learning large-scale domain mappings and expressing small feature changes. To overcome these limitations, an improved version called SuperstarGAN is proposed. It incorporates the idea of Controllable GAN and uses data augmentation techniques to handle overfitting. Evaluated with face image dataset, SuperstarGAN achieves better performance in terms of FID and LPIPS compared to StarGAN, and it can also control the degree of expression of target domain features in generated images.
Article
Computer Science, Artificial Intelligence
Dongfang Zhao et al.
Summary: In this paper, a single vibration signal-driven variable speed intelligent fault diagnosis scheme is proposed, which uses extreme multi-scale entropy as the alternative characterization parameter for speed information and develops a deep branch attention network to integrate vibration and speed information more flexibly. Experimental results show that the proposed method can effectively integrate vibration and speed information and achieve outstanding results even without a speed signal.
ADVANCED ENGINEERING INFORMATICS
(2023)
Article
Computer Science, Artificial Intelligence
Jipu Li et al.
Summary: This study proposes a novel framework called DCTLN-DWA for addressing the catastrophic forgetting and domain adaptation problems in fault diagnosis of rotating machinery. The framework uses the DWA module to retain previous diagnostic knowledge and learn new knowledge from new samples, and applies adversarial training strategy to eliminate data distribution discrepancy. Experimental results show that the proposed framework effectively handles industrial streaming data under different working conditions and can be a promising tool for solving actual industrial problems.
ADVANCED ENGINEERING INFORMATICS
(2023)
Article
Computer Science, Artificial Intelligence
Diwang Ruan et al.
Summary: As a representative deep learning network, Convolutional Neural Network (CNN) has been extensively used in bearing fault diagnosis and many good results have been reported. However, the current parameter specifications of CNN are not optimal and efficient. This paper elaborates the physical characteristics of bearing acceleration signals and utilizes them to guide the CNN design, achieving better performance compared to the baseline CNN.
ADVANCED ENGINEERING INFORMATICS
(2023)
Article
Automation & Control Systems
Pengfei Liang et al.
Summary: This paper proposes a novel deep residual deformable subdomain adaptation framework for cross-domain fault diagnosis of wind turbine bearing under time-varying speeds. The approach enhances the feature representation by using a deformable convolution module in the traditional residual network and improves the diagnostic model by removing the adversarial training mechanism and using a local maximum mean discrepancy to align the distributions of the same fault type in different domains. Experimental results show that this method can effectively diagnose faults in the target domain and significantly improve diagnosis accuracy and efficiency compared to existing methods.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2023)
Article
Engineering, Multidisciplinary
Yong Zhu et al.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2023)
Article
Engineering, Mechanical
Jing Liu et al.
Summary: In this study, a dynamic model for a DPGS including all components was established, considering the comprehensive excitations and vibrations of planetary bearing roller, planet, carrier, and planetary bearing cage. The correctness of the proposed model was verified through simulations and experiments, and the advantage of the proposed model over the model without the cage was demonstrated.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2023)
Article
Astronomy & Astrophysics
Zihao Cai et al.
Summary: Due to atmospheric turbulence, solar images received by solar telescopes are often severely degraded. To restore and improve these images, an improved Cycle Generative Adversarial Network (CycleGAN) is proposed. By incorporating identity loss and perceptual loss into the CycleGAN's loss function, the network's ability to capture image pixels and high-frequency information is significantly enhanced. The influence of turbulence intensity and image noise on the restoration effect of solar images is also analyzed, showing that the improved CycleGAN produces images with lower error and higher clarity. The proposed method provides a new processing approach for optical system telescopes and other image processing fields, without the need for paired training datasets.
Article
Computer Science, Artificial Intelligence
Pengfei Liang et al.
Summary: In this article, a novel approach named IVTN-SA is proposed for fault diagnosis of rolling bearings in variable working conditions. The approach integrates subdomain adaptation and an improved vision transformer network to overcome the challenges. Experimental results demonstrate that the proposed method achieves excellent performance in diagnosis accuracy and model complexity compared to existing methods.
ADVANCED ENGINEERING INFORMATICS
(2023)
Article
Computer Science, Artificial Intelligence
Chong Chen et al.
Summary: Fault diagnosis is crucial in operating and maintaining industrial assets. A fault diagnosis knowledge graph can assist engineers in conducting maintenance tasks. However, manually labeling the corpus from multiple sources is time-consuming, and the presence of noisy sentences hampers the performance of relation extraction modeling.
ADVANCED ENGINEERING INFORMATICS
(2023)
Article
Engineering, Multidisciplinary
Hongchun Sun et al.
Summary: This paper proposes an Efficient MultiScale Convolutional Neural Network (EMSCNN) with anti-noise based on visualization methods of interpretability for rolling bearing fault diagnosis. By using an improved visualization method and analyzing the anti-noise ability of modules, the diagnostic accuracy is improved.
Article
Computer Science, Artificial Intelligence
Juan Xu et al.
Summary: A Zero-shot Learning Compound Fault Diagnosis Model for bearings was proposed to address the challenge of diagnosing unknown compound faults without sufficient labeled samples. By embedding the semantic features of faults into the visual space of fault data, the model achieved a high accuracy of identifying compound faults even without any compound fault samples during training.
EXPERT SYSTEMS WITH APPLICATIONS
(2022)
Article
Engineering, Mechanical
Tianfu Li et al.
Summary: Deep learning methods have advanced the field of Prognostics and Health Management, but handling irregular data in non-Euclidean space remains a challenge. Research has proposed a practical guideline for utilizing graph neural networks for intelligent fault diagnostics and prognostics, and established a framework based on GNN for this purpose.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2022)
Article
Engineering, Mechanical
Kai Zhou et al.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2022)
Article
Engineering, Electrical & Electronic
Lijie Zhang et al.
Summary: This paper proposes a new high-precision prediction method based on improved convolutional neural network (ICNN), residual attention mechanism with soft thresholding, and gated recurrent unit (GRU) to overcome the difficulties of extracting useful features and predicting the remaining useful life (RUL) of mechanical equipment from industrial signals. Experimental results demonstrate that the proposed method outperforms other methods in prediction accuracy and time cost.
IEEE SENSORS JOURNAL
(2022)
Article
Automation & Control Systems
Zhijian Wang et al.
Summary: Due to data distribution discrepancy, fault diagnosis models trained in one scene may fail in classifying data from other scenes. To address this issue, we propose a new model called SATLN, which combines subdomain adaptation and domain adaptation techniques to reduce both marginal and conditional distribution biases, improving classification performance and generalization.
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
(2022)
Article
Materials Science, Multidisciplinary
Jiahui Tang et al.
Summary: Deep learning is widely used in fault diagnosis for machinery core components' health monitoring. To address the limitations of deep networks, a bearing fault diagnosis method based on mixed pooling deep belief network is proposed, which exhibits superior feature extraction ability and diagnosis efficiency.
RESULTS IN PHYSICS
(2022)
Article
Computer Science, Artificial Intelligence
Kun Xu et al.
Summary: Bearing fault diagnosis is crucial for the safe and stable operation of rotating machinery equipment. This paper presents a novel approach of false-real data synthesis to address the zero-shot problem in new conditions. By combining a bearing dynamic model with a generated adversarial network, the proposed method demonstrates excellent diagnostic performance.
ADVANCED ENGINEERING INFORMATICS
(2022)
Article
Computer Science, Artificial Intelligence
Ruixin Wang et al.
Summary: This paper proposes a deep feature enhanced reinforcement learning method for fault diagnosis of rolling bearing. The method improves the neural network with the Elu activation function and utilizes an attention model to enhance the feature and acquire global information accurately.
ADVANCED ENGINEERING INFORMATICS
(2022)
Article
Engineering, Mechanical
Xiaoming Yuan et al.
Summary: This study proposes a calculation model of dynamic bulk modulus that considers temperature factors. The effects of system pressure, air fraction, and temperature on bulk modulus are investigated using the controlled variable method. The theoretical model is validated by comparing it with experimental values.
CHINESE JOURNAL OF MECHANICAL ENGINEERING
(2022)
Article
Automation & Control Systems
Jiedi Sun et al.
Summary: This paper proposes a multi-hierarchical compression method to reduce the complexity of deep neural networks in industrial environments with constrained resources, while achieving intelligent fault diagnosis without a decrease in recognition accuracy.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2022)
Article
Engineering, Multidisciplinary
Ya-Jing Huang et al.
Summary: The paper introduces a Multi-scale Convolutional Neural Network with Channel Attention for bearing fault diagnosis, which achieves the highest diagnosis accuracy under noisy environments and varying working speeds compared to other fault diagnosis models.
Article
Engineering, Industrial
Jiabei He et al.
Summary: This paper proposes a data-driven model based on advanced machine-learning techniques for the quick and accurate prediction of battery capacity. The model utilizes wavelet transform for denoising initial features, random forest for feature selection, and approximate entropy theory for constructing multi-time scale sliding window data. It also introduces a novel QGA-ASM-LSTM model, which optimizes its parameters using a quantum genetic algorithm. Experimental results show that the proposed model outperforms state-of-the-art models in terms of prediction accuracy and training time.
RELIABILITY ENGINEERING & SYSTEM SAFETY
(2022)
Article
Computer Science, Artificial Intelligence
Shengnan Tang et al.
Summary: The research constructs an improved CNN model with adapting learning rate for fault diagnosis of a hydraulic piston pump using deep learning. By converting three raw signals into time-frequency images via continuous wavelet transform, the model can effectively identify different fault types. Analysis with confusion matrix shows the classification accuracy of each fault type.
ADVANCED ENGINEERING INFORMATICS
(2021)
Article
Automation & Control Systems
Long Wen et al.
Summary: This article introduces a novel learning rate scheduler based on reinforcement learning (RL) for convolutional neural networks (RL-CNN) in fault classification, which can efficiently and automatically adjust the learning rate, outperforming traditional methods and showing potential in fault classification.
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
(2021)
Article
Computer Science, Artificial Intelligence
Shen Liu et al.
Summary: In this study, an unsupervised representation learning method called BIMGAN was proposed for fault diagnosis of rotating machinery under small sample size conditions. By maximizing mutual information estimation and feature matching strategy, the method learns the mapping relationship between samples and their feature representations, achieving average accuracies of 99.73% and 98.36% in case studies.
KNOWLEDGE-BASED SYSTEMS
(2021)
Article
Engineering, Industrial
Xin Li et al.
Summary: The proposed fault diagnosis method combining feature-fusion covariance matrix and multi-Riemannian kernel ridge regression can effectively improve the fault diagnosis performance of gearboxes. Experimental results demonstrate that this approach consistently exhibits excellent diagnostic performance on multi-sensor datasets.
RELIABILITY ENGINEERING & SYSTEM SAFETY
(2021)
Article
Engineering, Industrial
Yeongmin Yoo et al.
Summary: A new predictive model DGAN is proposed, which can construct a predictive model through machine learning using only actual test data, improve the prediction accuracy of an actual test model, and present design variables that affect the response data.
RELIABILITY ENGINEERING & SYSTEM SAFETY
(2021)
Article
Engineering, Industrial
Yiwei Wang et al.
Summary: This paper proposes a new TCN model with soft threshold and attention mechanism for machinery prognostics, which directly uses multi-channel sensor data for RUL prediction. The adaptive learning of threshold values improves the prediction accuracy by assigning customized values to feature maps.
JOURNAL OF MANUFACTURING SYSTEMS
(2021)
Article
Computer Science, Artificial Intelligence
Daoguang Yang et al.
Summary: This paper introduces a novel deep learning framework for highly accurate rotating machinery fault diagnosis using a residual wide-kernel deep convolutional auto-encoder, which can learn important features from limited raw vibration signals. The model ensures depth without gradient vanishing and overfitting issues, and demonstrates denoising and feature extraction ability with very few training samples.
Article
Automation & Control Systems
Hongfeng Tao et al.
JOURNAL OF THE FRANKLIN INSTITUTE-ENGINEERING AND APPLIED MATHEMATICS
(2020)
Article
Automation & Control Systems
Yue Zhuo et al.
JOURNAL OF PROCESS CONTROL
(2020)
Article
Automation & Control Systems
Jun Dai et al.
IEEE-ASME TRANSACTIONS ON MECHATRONICS
(2020)
Article
Automation & Control Systems
Siyu Shao et al.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2019)
Article
Engineering, Electrical & Electronic
Yinghui Zhang et al.
SIGNAL PROCESSING-IMAGE COMMUNICATION
(2019)
Article
Computer Science, Artificial Intelligence
Han Liu et al.
