An evolutional deep learning method based on multi-feature fusion for fault diagnosis in sucker rod pumping system

Li, Juanni; Shao, Jun Peng; Wang, Wei; Xie, Wei

doi:10.1016/j.aej.2022.11.028

Cited by 10 publications

(7 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Zhang et al [ 24 ] established the operating condition recognition model utilizing a sparse multi-graph regularized extreme-learning machine based on the fast discrete curvelet transform feature extraction method. Li et al [ 1 ] proposed an improved deep learning method based on the fusion of Fourier descriptor features and graphic features of dynamometer cards. To solve the problem that the recognition effect and generalization performance of the deep learning working condition recognition model were prone to decline in few-shot training samples, He et al [ 25 ] proposed to compress the original working condition data by using the 4-dimensional time-frequency feature extraction method, and then adjust the parameters of the convolutional shrinkage neural network by meta-learning.…”

Section: Research Background Of the Proposed Methodsmentioning

confidence: 99%

“…Nevertheless, the current operating condition recognition research suffers from the following constraints. First, most methods solely rely on one information source [ [1] , [2] , [3] , [4] , [5] , [6] ], e.g., electrical parameters or dynamometer cards. However, employing one information source easily triggers a complex electron machinery liquid integrating system, producing false alarms.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Multisource working condition recognition via nonlinear kernel learning and p-Laplacian manifold learning

Zhou,

Niu,

Yang

et al. 2024

Heliyon

View full text Add to dashboard Cite

Section: Research Background Of the Proposed Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Multisource working condition recognition via nonlinear kernel learning and p-Laplacian manifold learning

Zhou,

Niu,

Yang

et al. 2024

Heliyon

View full text Add to dashboard Cite

“…Some recent works aim to further improve diagnosis performance. Li et al 35 proposed a multi‐input convolutional neural network that fuses graphical and Fourier descriptor features from indicator diagrams. However, complex downhole environments lead to sophisticated indicator diagrams with small sample sizes.…”

Section: Related Workmentioning

confidence: 99%

Multiscale attention for few‐shot image classification

Zhou,

Dong,

Song

et al. 2024

Computational Intelligence

View full text Add to dashboard Cite

In recent years, the application of traditional deep learning methods in the agricultural field using remote sensing techniques, such as crop area and growth monitoring, crop classification, and agricultural disaster monitoring, has been greatly facilitated by advancements in deep learning. The accuracy of image classification plays a crucial role in these applications. Although traditional deep learning methods have achieved significant success in remote sensing image classification, they often involve convolutional neural networks with a large number of parameters that require extensive optimization using numerous remote sensing images for training purposes. To address these challenges, we propose a novel approach called multiscale attention network (MAN) for sample‐based remote sensing image classification. This method consists primarily of feature extractors and attention modules to effectively utilize different scale features through multiscale feature training during the training phase. We evaluate our proposed method on three datasets comprising agricultural remote sensing images and observe superior performance compared to existing approaches. Furthermore, we validate its generalizability by testing it on an oil well indicator diagram specifically designed for classification tasks.

show abstract

“…While the current neural network diagnosis method is gradually developed, Ying H [6] showed that it does not need to rely on manual experience and is capable of fault diagnosis by comprehensive analysis of pumping machine working condition data with only a small consumption of resources. Li J [7] used a multi-feature fusion model to diagnose eight types of pumping machine faults with an average accuracy of more than 93%, but the method is too demanding for oil eld He Y [8] used an improved CNN-LSTM model to achieve more than 95% accuracy in ideal state detection, but this approach is not ideal when the samples of the schematic power diagram are small and the structure is too complex. Zhou W [9] developed an intelligent oil eld fault diagnosis system using UKF-RBF model, which has good results in multi-fault classi cation.…”

Section: Introductionmentioning

confidence: 99%

A Fourier descriptor and PSCS-RBF fusion method for pumping machine fault diagnosis

Raja

Li³

et al. 2023

Preprint

View full text Add to dashboard Cite

For the current oilfield pumping machine fault diagnosis, there are time-consuming and inefficient problems, and at the same time high requirements for hardware resources and no universality, this study proposes a pumping machine fault diagnosis method based on improved Fourier descriptor combined with dynamic adaptive cuckoo search (PS cuckoo search, PSCS) to optimize RBF neural network. Firstly, the starting point position of the contour of the power diagram is determined by the minimum inertia axis, and the Fourier transform is performed to achieve the best matching between contours, and the effect of starting point irrelevance. Then feature extraction is performed by combining shape invariant moments as the input layer information of RBF neural network. Then dynamic discovery probability and adaptive step size are introduced to make the cuckoo search easier to retain the better solution, and the step size can be automatically adjusted according to the convergence rate of the objective function to maintain a balanced state of efficiency and accuracy in different search stages. Finally, the RBF neural network is optimized by the improved cuckoo search to obtain the optimal relevant parameters such as the width and weights of RBF, and the PSCS-RBF fault diagnosis model is established. The model is applied to the diagnosis of different fault types of pumping machines and is compared and analyzed with a variety of current mainstream models. The average detection accuracy of the PSCS-RBF fault diagnosis method proposed in the article reaches 96.3%, and the measured results have high accuracy and short time, which verifies the practical value and advantages of the method.

show abstract

An evolutional deep learning method based on multi-feature fusion for fault diagnosis in sucker rod pumping system

Cited by 10 publications

References 26 publications

Multisource working condition recognition via nonlinear kernel learning and p-Laplacian manifold learning

Multisource working condition recognition via nonlinear kernel learning and p-Laplacian manifold learning

Multiscale attention for few‐shot image classification

A Fourier descriptor and PSCS-RBF fusion method for pumping machine fault diagnosis

Contact Info

Product

Resources

About