Guozhi Li
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Guozhi Li has a total of 5 co-authors. Here are the top co-authors by number of shared publications:
- Lifeng Zhang
Yanzhao Electric Power Laboratory of North China Electric Power University
3 shared publications - Chuanbao Wu
Baoding Key Laboratory of State Detection and Optimization Regulation for Integrated Energy System, North China Electric Power University, Baoding, China
1 shared publication Yufu Liu
Baoding Key Laboratory of State Detection and Optimization Regulation for Integrated Energy System, North China Electric Power University, Baoding 071003, People’s Republic of China
1 shared publication- Yufu Liu
Baoding Key Laboratory of State Detection and Optimization Regulation for Integrated Energy System, North China Electric Power University, Baoding, China
1 shared publication - Zhihao Tang
Baoding Key Laboratory of State Detection and Optimization Regulation for Integrated Energy System, North China Electric Power University, Baoding, China
1 shared publication
Latest Publications
- 1 September 2025
- Published by AIP Publishing in Physics of Fluids
- Vol. 37 (9)
Flow characteristic analysis and flow pattern identification of gas–liquid two-phase flow based on energy symmetrized dot pattern and Conv2former
This study proposes a gas–liquid two-phase flow pattern recognition method based on an energy symmetric dot pattern (SDP) and the Conv2former network. A conductivity sensor array is installed in a vertical pipe to collect raw signal data under different flow regimes. To reduce redundancy and computational cost, dimensionality reduction is performed using a mean-based approach. The reduced signals are then transformed into energy representations using the Choi–Williams distribution, and corresponding SDP images are generated based on the energy sequences. These images visually reveal the structural differences among various flow patterns and are used to construct the dataset, which is divided into training and testing sets before being input into the Conv2former network for training and evaluation. Experimental results demonstrate that the proposed SDP image representation effectively enhances the discriminative features between different flow patterns. Leveraging its capability to model both local details and global dependencies, the Conv2former achieves excellent performance across four typical flow regimes, with an average recognition accuracy of 95.83%. Compared with conventional convolutional neural networks, such as ResNet-18, GoogLeNet, and SqueezeNet, Conv2former exhibits superior adaptability and robustness, particularly in classifying complex, transitional, and high-frequency disturbed flow patterns. These findings validate the effectiveness and potential applicability of the proposed method in intelligent multiphase flow pattern recognition.
- 1 September 2025
- Published by AIP Publishing in Physics of Fluids
- Vol. 37 (9)
Flow characteristic analysis and flow pattern identification of gas–liquid two-phase flow based on energy symmetrized dot pattern and Conv2former
This study proposes a gas–liquid two-phase flow pattern recognition method based on an energy symmetric dot pattern (SDP) and the Conv2former network. A conductivity sensor array is installed in a vertical pipe to collect raw signal data under different flow regimes. To reduce redundancy and computational cost, dimensionality reduction is performed using a mean-based approach. The reduced signals are then transformed into energy representations using the Choi–Williams distribution, and corresponding SDP images are generated based on the energy sequences. These images visually reveal the structural differences among various flow patterns and are used to construct the dataset, which is divided into training and testing sets before being input into the Conv2former network for training and evaluation. Experimental results demonstrate that the proposed SDP image representation effectively enhances the discriminative features between different flow patterns. Leveraging its capability to model both local details and global dependencies, the Conv2former achieves excellent performance across four typical flow regimes, with an average recognition accuracy of 95.83%. Compared with conventional convolutional neural networks, such as ResNet-18, GoogLeNet, and SqueezeNet, Conv2former exhibits superior adaptability and robustness, particularly in classifying complex, transitional, and high-frequency disturbed flow patterns. These findings validate the effectiveness and potential applicability of the proposed method in intelligent multiphase flow pattern recognition.
- 26 June 2025
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 36 (7) , 075303
Identification method of gas–liquid two-phase flow pattern based on dual-domain multi-scale feature fusion framework
Accurate identification of the flow pattern of gas–liquid two-phase flow is of great significance to ensure the transportation safety of industrial pipelines. We propose a flow pattern identification method based on dual-domain and multi-scale feature fusion. Firstly, the gas–liquid two-phase flow pattern information in the vertical pipeline is collected by array resistance sensors, and the measured data is processed by dimensionality reduction. Firstly, the gas–liquid two-phase flow pattern information in the vertical pipeline is acquired using array resistance sensors, and the measured data undergoes dimensionality reduction for preprocessing. Subsequently, the reduced-dimensional data is segmented through a sliding window approach. The segmented data is analyzed using Choi-Williams Distribution and Discrete Fourier Transform to extract the energy signal and power spectrum signal, respectively. These signals are then encoded into two-dimensional grayscale images by pseudo-image encoding, providing a basis for investigating the flow pattern transition mechanisms in two-phase flow. Finally, the two grayscale images are fed into the dual-domain multi-scale feature fusion network for flow pattern recognition. The results show that this method can reveal the evolution behavior of gas–liquid two-phase flow from both the frequency domain and the time–frequency domain. The average recognition accuracy reaches 97.78%, achieving accurate identification of flow patterns.
- 26 June 2025
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 36 (7) , 075303
Identification method of gas–liquid two-phase flow pattern based on dual-domain multi-scale feature fusion framework
Accurate identification of the flow pattern of gas–liquid two-phase flow is of great significance to ensure the transportation safety of industrial pipelines. We propose a flow pattern identification method based on dual-domain and multi-scale feature fusion. Firstly, the gas–liquid two-phase flow pattern information in the vertical pipeline is collected by array resistance sensors, and the measured data is processed by dimensionality reduction. Firstly, the gas–liquid two-phase flow pattern information in the vertical pipeline is acquired using array resistance sensors, and the measured data undergoes dimensionality reduction for preprocessing. Subsequently, the reduced-dimensional data is segmented through a sliding window approach. The segmented data is analyzed using Choi-Williams Distribution and Discrete Fourier Transform to extract the energy signal and power spectrum signal, respectively. These signals are then encoded into two-dimensional grayscale images by pseudo-image encoding, providing a basis for investigating the flow pattern transition mechanisms in two-phase flow. Finally, the two grayscale images are fed into the dual-domain multi-scale feature fusion network for flow pattern recognition. The results show that this method can reveal the evolution behavior of gas–liquid two-phase flow from both the frequency domain and the time–frequency domain. The average recognition accuracy reaches 97.78%, achieving accurate identification of flow patterns.
- 4 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 25 (8) , 13480-13489
A Framework for Flow Pattern Analysis and Identification Based on Dual-Domain Feature Extraction and Deep Learning
A flow pattern analysis framework based on the combination of dual-domain feature extraction and deep learning is proposed. Measurement data of flow patterns are collected using a resistance sensor array. For flow pattern analysis, A novel approach combining Choi-Williams distribution (CWD) and limited penetrable visibility graph (LPVG) is proposed to construct complex networks. The average degree index and global efficiency index are calculated, and the topological structure of the complex network is analyzed to reveal the nonlinear dynamic behavior of flow patterns. Regarding flow pattern identification, a pseudo-image encoding (PIE) method is employed to encode the energy time series from time-frequency analysis and the degree sequence from complex networks into two types of two-dimensional grayscale images to complete the feature extraction in the energy domain and the network domain. A novel deep learning classification model, the dual-input feature fusion network (DIFFN), is proposed to use two types of grayscale images as network inputs to complete flow pattern identification. The results indicate that our framework allows effectively characterizing the nonlinear dynamic behaviors during the evolution of different gas-liquid flow patterns. Meanwhile, the identification accuracy of the three flow patterns can reach 95%.
- 4 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 25 (8) , 13480-13489
A Framework for Flow Pattern Analysis and Identification Based on Dual-Domain Feature Extraction and Deep Learning
A flow pattern analysis framework based on the combination of dual-domain feature extraction and deep learning is proposed. Measurement data of flow patterns are collected using a resistance sensor array. For flow pattern analysis, A novel approach combining Choi-Williams distribution (CWD) and limited penetrable visibility graph (LPVG) is proposed to construct complex networks. The average degree index and global efficiency index are calculated, and the topological structure of the complex network is analyzed to reveal the nonlinear dynamic behavior of flow patterns. Regarding flow pattern identification, a pseudo-image encoding (PIE) method is employed to encode the energy time series from time-frequency analysis and the degree sequence from complex networks into two types of two-dimensional grayscale images to complete the feature extraction in the energy domain and the network domain. A novel deep learning classification model, the dual-input feature fusion network (DIFFN), is proposed to use two types of grayscale images as network inputs to complete flow pattern identification. The results indicate that our framework allows effectively characterizing the nonlinear dynamic behaviors during the evolution of different gas-liquid flow patterns. Meanwhile, the identification accuracy of the three flow patterns can reach 95%.