Yuan Li
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Yuan Li has a total of 20 co-authors. Here are the top co-authors by number of shared publications:
- Guangya Zhu
College of Electrical Engineering, Sichuan University, Chengdu, China
3 shared publications - Lu Lu
College of Electronics and Information Engineering, Sichuan University, Chengdu, China
3 shared publications - Kai Zhou
College of Electrical Engineering, Sichuan University, Chengdu, China
2 shared publications - Pengfei Meng
High Voltage Laboratory, College of Electrical Engineering, Sichuan University, Chengdu, China
2 shared publications - Yao Fu
Sichuan University
2 shared publications Bin Luo
School of Mechanical Engineering, Northwestern Polytechnical University, Xian, China
1 shared publication
Latest Publications
- 16 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 75 (00189456) , 1-16
Multihead Self-Attention Incremental Network for Rotating Machinery Fault Diagnosis Across Nonstationary Operating Conditions
The advancement of artificial intelligence has enabled significant progress in data-driven fault diagnosis for rotating machinery. However, in real-world industrial environments, such machinery often operates under time-varying and non-stationary conditions, leading to continuous changes in data distributions. These variations pose a major challenge to conventional deep learning models, which are prone to catas-trophic forgetting when sequentially learning new diagnostic tasks. To address this, this paper proposes a novel Multi-head Self-attention Incremental Network (MSIN) for fault diagnosis across non-stationary operating conditions. MSIN employs a stack of multi-head self-attention encoder layers to effectively extract fault-relevant features from frequency-domain signals. To alleviate catastrophic forgetting, the model integrates a selective exemplar replay strategy and introduces a joint cross-head knowledge distillation loss, enabling effective knowledge retention across tasks. The proposed method is evaluated on two benchmark datasets under both stationary and non-stationary conditions. Experimental results demonstrate that MSIN achieves superior diagnostic accuracy and knowledge retention compared to existing Incremental Learning methods, offering a promising solution for continuous fault diagnosis in dynamic industrial settings. The source code of the MSIN model is available at: https://github.com/bankuaimianbao/MSIN-Non-stationary-Code.
- 16 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 75 (00189456) , 1-16
Multihead Self-Attention Incremental Network for Rotating Machinery Fault Diagnosis Across Nonstationary Operating Conditions
The advancement of artificial intelligence has enabled significant progress in data-driven fault diagnosis for rotating machinery. However, in real-world industrial environments, such machinery often operates under time-varying and non-stationary conditions, leading to continuous changes in data distributions. These variations pose a major challenge to conventional deep learning models, which are prone to catas-trophic forgetting when sequentially learning new diagnostic tasks. To address this, this paper proposes a novel Multi-head Self-attention Incremental Network (MSIN) for fault diagnosis across non-stationary operating conditions. MSIN employs a stack of multi-head self-attention encoder layers to effectively extract fault-relevant features from frequency-domain signals. To alleviate catastrophic forgetting, the model integrates a selective exemplar replay strategy and introduces a joint cross-head knowledge distillation loss, enabling effective knowledge retention across tasks. The proposed method is evaluated on two benchmark datasets under both stationary and non-stationary conditions. Experimental results demonstrate that MSIN achieves superior diagnostic accuracy and knowledge retention compared to existing Incremental Learning methods, offering a promising solution for continuous fault diagnosis in dynamic industrial settings. The source code of the MSIN model is available at: https://github.com/bankuaimianbao/MSIN-Non-stationary-Code.
- 28 November 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 73 (00189456) , 1-9
A Partial Discharge Signal Separation Method Applicable for Various Sensors Based on Time–Frequency Feature Extraction of t-SNE
Currently employed features for separating partial discharge (PD) pulse signals are proposed based on signals captured by a particular sensor. If these features are used to separate PD signals captured by other sensors, the distribution of these features may overlap, making it difficult to distinguish different PD signals. To address the challenge of poor applicability of traditional features to sensors with different amplitude frequency characteristics in separating different PD signals, this paper proposes a pulse time-frequency (T-F) feature extraction method based on t-distributed Stochastic Neighbor Embedding (t-SNE). The proposed method uses the wavelet T-F spectrum of each PD pulse and reduces the dimensionality of the T-F spectrum by the t-SNE algorithm. The obtained features effectively discriminate between T-F spectra of different discharge types. Furthermore, to ensure satisfactory separation outcomes, the paper proposes an optimisation method for setting the perplexity parameter of the t-SNE algorithm based on root mean square error (RMSE). Finally, the proposed feature extraction method’s effectiveness is validated using test data from different sensors and various tested objects. Additionally, this method shows better visual separation results compared to traditional T-F features and other dimensionality reduction methods.
- 28 November 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 73 (00189456) , 1-9
A Partial Discharge Signal Separation Method Applicable for Various Sensors Based on Time–Frequency Feature Extraction of t-SNE
Currently employed features for separating partial discharge (PD) pulse signals are proposed based on signals captured by a particular sensor. If these features are used to separate PD signals captured by other sensors, the distribution of these features may overlap, making it difficult to distinguish different PD signals. To address the challenge of poor applicability of traditional features to sensors with different amplitude frequency characteristics in separating different PD signals, this paper proposes a pulse time-frequency (T-F) feature extraction method based on t-distributed Stochastic Neighbor Embedding (t-SNE). The proposed method uses the wavelet T-F spectrum of each PD pulse and reduces the dimensionality of the T-F spectrum by the t-SNE algorithm. The obtained features effectively discriminate between T-F spectra of different discharge types. Furthermore, to ensure satisfactory separation outcomes, the paper proposes an optimisation method for setting the perplexity parameter of the t-SNE algorithm based on root mean square error (RMSE). Finally, the proposed feature extraction method’s effectiveness is validated using test data from different sensors and various tested objects. Additionally, this method shows better visual separation results compared to traditional T-F features and other dimensionality reduction methods.
- 25 October 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 71 (00189456) , 1-9
Partial Discharge Localization on Power Cables Based on a Novel Signal Relay System
Time domain reflectometry (TDR) is an effective method used to detect partial discharge (PD) and localize PD sources on power cables. However, for long-length power cables, the attenuation and dispersion of PD signals are significant due to the propagated distance. The attenuation causes a serious reduction of the pulse amplitude, which makes the defects located far away from the PD detector difficult to find. The dispersion causes appreciable asymmetry in the PD pulse wave form. Thus, the localization accuracy is reduced. To solve the problem of long-length cable fault localization, a novel signal relay system is proposed in this paper. Once a weak PD pulse is captured in the signal relay system, another standard Gaussian pulse signal is initiated by a function generator. The Gaussian pulse signal is injected into the cable and travels along the cable. During this procedure, the weak PD pulse is likely to be amplified, and the fault localization accuracy is notably enhanced. To validate the effectiveness of the proposed method, an experimental platform is built in a laboratory, and PD detection of an artificial defective cable is performed. The experimental results show that the signal relay system can effectively “enlarge” the PD pulse on a long-length cable. For long-length power cables, PD detection can achieve a higher localization accuracy based on the proposed method.
- 25 October 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 71 (00189456) , 1-9
Partial Discharge Localization on Power Cables Based on a Novel Signal Relay System
Time domain reflectometry (TDR) is an effective method used to detect partial discharge (PD) and localize PD sources on power cables. However, for long-length power cables, the attenuation and dispersion of PD signals are significant due to the propagated distance. The attenuation causes a serious reduction of the pulse amplitude, which makes the defects located far away from the PD detector difficult to find. The dispersion causes appreciable asymmetry in the PD pulse wave form. Thus, the localization accuracy is reduced. To solve the problem of long-length cable fault localization, a novel signal relay system is proposed in this paper. Once a weak PD pulse is captured in the signal relay system, another standard Gaussian pulse signal is initiated by a function generator. The Gaussian pulse signal is injected into the cable and travels along the cable. During this procedure, the weak PD pulse is likely to be amplified, and the fault localization accuracy is notably enhanced. To validate the effectiveness of the proposed method, an experimental platform is built in a laboratory, and PD detection of an artificial defective cable is performed. The experimental results show that the signal relay system can effectively “enlarge” the PD pulse on a long-length cable. For long-length power cables, PD detection can achieve a higher localization accuracy based on the proposed method.
- 23 March 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 71 (00189456) , 1-9
A Novel Dampness Diagnosis Method for Distribution Power Cables Based on Time-Frequency Domain Conversion
Dampness invasion of power cable joints may lead to cable insulation degradation gradually. Once the cable joint is severely damp, it rapidly leads to breakdown failure for the power cable insulation system and finally results in catastrophic cable accidents. To avoid power outages, it is of great significance to diagnose cable joints with severe dampness defects. Existing methods have certain limitations. In this article, a novel time-frequency domain conversion method for determining cable joints with severe dampness defects is proposed. For this method, the cable joint is first located by applying the conventional frequency domain reflectometry (FDR) technology. Then, the time domain waveform at the cable joint is obtained based on the proposed time-frequency domain conversion algorithm. Finally, the cable joint with severe dampness defect can be distinguished by judging the polarities of the time domain spectrum at the cable joint. To verify the effectiveness of the proposed method, simulations and experiments are performed with comparisons with the conventional FDR method. Simulation and field detected results prove that the proposed method can effectively distinguish cable joints with severe dampness defects.
- 23 March 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 71 (00189456) , 1-9
A Novel Dampness Diagnosis Method for Distribution Power Cables Based on Time-Frequency Domain Conversion
Dampness invasion of power cable joints may lead to cable insulation degradation gradually. Once the cable joint is severely damp, it rapidly leads to breakdown failure for the power cable insulation system and finally results in catastrophic cable accidents. To avoid power outages, it is of great significance to diagnose cable joints with severe dampness defects. Existing methods have certain limitations. In this article, a novel time-frequency domain conversion method for determining cable joints with severe dampness defects is proposed. For this method, the cable joint is first located by applying the conventional frequency domain reflectometry (FDR) technology. Then, the time domain waveform at the cable joint is obtained based on the proposed time-frequency domain conversion algorithm. Finally, the cable joint with severe dampness defect can be distinguished by judging the polarities of the time domain spectrum at the cable joint. To verify the effectiveness of the proposed method, simulations and experiments are performed with comparisons with the conventional FDR method. Simulation and field detected results prove that the proposed method can effectively distinguish cable joints with severe dampness defects.