Incremental learning BiLSTM based on dynamic proportional adjustment mechanism and experience replay for quantitative detection of blade crack propagation

Shen, Jun‐Xian; Ma, Tianchi; Song, Di; Xu, Feiyun

doi:10.1177/14759217231170723

Cited by 3 publications

(1 citation statement)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Convolutional layers and max-pooling layers are used in all the convolution modules. The neuron activation function uses a ReLU to solve the network's gradient dispersion [20][21][22].…”

Section: Construction Of Feature Adaptation Extractormentioning

confidence: 99%

A cross-domain intelligent fault diagnosis method based on multi-source domain feature adaptation and selection

Jia,

Huang,

Cheng

et al. 2024

Meas. Sci. Technol.

View full text Add to dashboard Cite

Although fault diagnosis methods integrating transfer learning are research hotspots, their ability to handle industrial fault diagnosis problems with large domain differences still needs to be improved. A multi-source domain feature adaptation and selection (MDFAS) method is presented to address the issues of domain mismatch and domain negative transfer. The method integrates the top-level network parameter transfer strategy with the 2D Convolutional Neural Network (2DCNN) backbone network to acquire the target domain feature extractor quickly. Multiple feature adaptive extractors (FAEs) are constructed using a multi-branch structure to align the source and target domain's feature distributions, respectively. The inter-domain distance computed by multi-kernel maximum mean discrepancy (MK-MMD) is embedded in the FAEs loss function to improve the inter-domain matching degree. Based on the information gain of the adaptively integrated features, the ensemble adaptive selection is performed on the extracted feature matrices to exclude the negative transfer feature. Finally, the effective feature matrix is input into the diagnosis classifier for classification. Cross-domain fault diagnosis experiments are developed based on the data set gathered from several types of rotating machinery operated under varied working conditions. The experimental results show that the proposed method outperforms the existing intelligent fault diagnosis methods in terms of fault detection accuracy, generalization, and stability.

show abstract

“…Convolutional layers and max-pooling layers are used in all the convolution modules. The neuron activation function uses a ReLU to solve the network's gradient dispersion [20][21][22].…”

Section: Construction Of Feature Adaptation Extractormentioning

confidence: 99%

A cross-domain intelligent fault diagnosis method based on multi-source domain feature adaptation and selection

Jia,

Huang,

Cheng

et al. 2024

Meas. Sci. Technol.

View full text Add to dashboard Cite

show abstract

An embedded physical information network for blade crack detection considering dynamic multi-level credibility

Shen,

Ma,

Song

et al. 2025

Mechanical Systems and Signal Processing

View full text Add to dashboard Cite

Behavior Expectation‐Based Anomaly Detection in Bridge Deflection Using AOA‐BiLSTM‐TPA: Considering Temperature and Traffic‐Induced Temporal Patterns

Qu,

Xia,

Sun

et al. 2024

Structural Control and Health Monitoring

View full text Add to dashboard Cite

In the realm of structural health monitoring (SHM), understanding the expected behavior of a structure is vital for the timely identification of anomalous activities. Existing methods often model only the physical quantities of monitoring data, neglecting the corresponding temporal information. To address this, this paper presents an innovative deep learning framework that synergistically combines a BiLSTM model, fortified by a temporal pattern attention (TPA) mechanism, with time-encoded temperature and traffic-induced deflection-temporal patterns. The arithmetic optimization algorithm (AOA) is employed for optimal hyperparameter tuning, and incremental learning was implemented to enable real-time updates of the model. Based on the proposed framework, an anomaly detection method was subsequently developed. This method is bidirectional: it uses quantile loss to provide expected ranges for structural behavior, identifying isolated anomalies, while the windowed normalized mutual information (WNMI) based on multivariate kernel density estimation (MKDE) helps detect trend variability caused by decreases in structural stiffness. This framework and the anomaly detection method were validated using data from an operational cable-stayed bridge. The results demonstrate that the method effectively predicts structural behavior and detects anomalies, highlighting the critical role of temporal information in SHM.

show abstract

Incremental learning BiLSTM based on dynamic proportional adjustment mechanism and experience replay for quantitative detection of blade crack propagation

Cited by 3 publications

References 23 publications

A cross-domain intelligent fault diagnosis method based on multi-source domain feature adaptation and selection

A cross-domain intelligent fault diagnosis method based on multi-source domain feature adaptation and selection

An embedded physical information network for blade crack detection considering dynamic multi-level credibility

Behavior Expectation‐Based Anomaly Detection in Bridge Deflection Using AOA‐BiLSTM‐TPA: Considering Temperature and Traffic‐Induced Temporal Patterns

Contact Info

Product

Resources

About