Kai Huang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Kai Huang has a total of 21 co-authors. Here are the top co-authors by number of shared publications:
- Zeyu Jiao
Guangdong Key Laboratory of Modern Control Technology, Institute of Intelligent Manufacturing, Guangdong Academy of Sciences, Guangzhou, China
5 shared publications Qun Wang
School of Logistics, Beijing Wuzi University, Beijing, China
4 shared publicationsGuozhu Jia
School of Economics and Management, Beihang University, Beijing 100191, People’s Republic of China
3 shared publicationsYingjie Cai
Department of Electronic Engineering, The Chinese University of Hong Kong, Hong Kong Special Administrative Region of China 999077, People’s Republic of China
3 shared publications- Guozhu Jia
Beihang University, Beijing, China
2 shared publications - Yingjie Cai
Department of Electronic Engineering, The Chinese University of Hong Kong, Hong Kong, China
2 shared publications
Latest Publications
- 18 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 25 (21) , 40775-40792
Remaining Useful Life Prediction Through Adaptive Spatiotemporal Graph Neural Network Based on Relationship Mining for Complex Aviation Equipment
The evolution of the aviation industry has led to increased demands for enhanced reliability in aviation equipment, underscoring the need for reliable system operations, reduced production costs, and the prevention of unscheduled downtimes. Traditional maintenance methods, such as fault correction and time-based preventive maintenance, are becoming increasingly inadequate due to the heightened complexity and precision requirements of modern aviation equipment. To this end, an adaptive spatiotemporal graph attention network based on relation mining is proposed for predicting the remaining useful life (RUL) of complex aviation equipment. This method starts by processing raw equipment data through a temporal extractor, capturing time-dependent patterns and inherent features. It then applies a relation mining algorithm, inspired by the Decision Making Trial and Evaluation Laboratory (DEMATEL) method, to identify multi-order coupling relationships among sensor data, creating a dynamic relationship matrix that encapsulates these temporal features. This matrix, along with temporal data, is integrated into a spatiotemporal graph neural network for effective information fusion, emphasizing key features from both spatial and temporal domains. Compared with the state of the art methods, the experimental results on the Commercial Modular Aerospace Propulsion System Simulation (C-MAPSS) dataset demonstrate the superior performance, with the root mean square error (RMSE) value improvements of 1.32%, 0.77%, 6.06%, and 14.98% across four subsets. respectively. By merging traditional DEMATEL relationship mining with graph neural network technology, and embedding artificial intelligence within domain knowledge to model complex systems, this method accurately predicts RUL within complex aviation systems, demonstrating superior efficacy and performance. The proposed method offers significant potential for enhancing system reliability and safety in the aviation industry.
- 18 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 25 (21) , 40775-40792
Remaining Useful Life Prediction Through Adaptive Spatiotemporal Graph Neural Network Based on Relationship Mining for Complex Aviation Equipment
The evolution of the aviation industry has led to increased demands for enhanced reliability in aviation equipment, underscoring the need for reliable system operations, reduced production costs, and the prevention of unscheduled downtimes. Traditional maintenance methods, such as fault correction and time-based preventive maintenance, are becoming increasingly inadequate due to the heightened complexity and precision requirements of modern aviation equipment. To this end, an adaptive spatiotemporal graph attention network based on relation mining is proposed for predicting the remaining useful life (RUL) of complex aviation equipment. This method starts by processing raw equipment data through a temporal extractor, capturing time-dependent patterns and inherent features. It then applies a relation mining algorithm, inspired by the Decision Making Trial and Evaluation Laboratory (DEMATEL) method, to identify multi-order coupling relationships among sensor data, creating a dynamic relationship matrix that encapsulates these temporal features. This matrix, along with temporal data, is integrated into a spatiotemporal graph neural network for effective information fusion, emphasizing key features from both spatial and temporal domains. Compared with the state of the art methods, the experimental results on the Commercial Modular Aerospace Propulsion System Simulation (C-MAPSS) dataset demonstrate the superior performance, with the root mean square error (RMSE) value improvements of 1.32%, 0.77%, 6.06%, and 14.98% across four subsets. respectively. By merging traditional DEMATEL relationship mining with graph neural network technology, and embedding artificial intelligence within domain knowledge to model complex systems, this method accurately predicts RUL within complex aviation systems, demonstrating superior efficacy and performance. The proposed method offers significant potential for enhancing system reliability and safety in the aviation industry.
- Open Access
- 22 June 2025
- Published by JMIR Publications Inc. in JMIR Preprints
A Systematic Review of Prolonged Posture Monitoring With Accelerometer-Based Wearable Sensors (Preprint)
BACKGROUND Prolonged postures, such as sustained sitting and standing, contribute to musculoskeletal disorders, yet daily monitoring approaches vary widely, limiting comparability and the development of standardized posture monitoring strategies. OBJECTIVE This systematic review aimed to summarize current practices in prolonged posture monitoring using accelerometer-based wearable sensors and to identify gaps across four stages: data collection, data preprocessing, data analysis, and user feedback. METHODS This systematic review synthesized 79 studies (82.3% rated low risk of bias) extracted from 16645 records in 10 databases, including PubMed, IEEE Xplore, Scopus, Web of Science, EI, ACM, ProQuest, CINAHL, SPORTDiscus, and EMBASE. The risk of bias was assessed using NOS, QUADAS-2, ROBINS-I, RoB 2, and a self-developed tool for algorithmic studies, with 87.7% rated as low risk. Due to methodological heterogeneity, results were synthesized qualitatively. RESULTS Thigh- and waist-worn sensors were most common, particularly activPAL and ActiGraph. Sensor placement, population characteristics, and preprocessing methods varied widely. Most studies used manually engineered features with statistical analysis, yet few leveraged deep learning or biomechanical metrics. Only 11 studies implemented user feedback strategies, typically visual or vibratory. Personalized user feedback remains underexplored, and few studies have considered specific free-living contexts or privacy concerns. CONCLUSIONS Findings highlight the promising potential of accelerometer-based posture monitoring systems. However, a lack of standardization in sensor placement and data processing, underrepresentation of diverse populations, and limited real-world feedback mechanisms constrain scalability. Future research should develop standard protocols for sensor placement, include diverse population groups, and prioritize the use of interpretable machine learning to support robust classification. Developers should design mobile-compatible systems with personalized, real-time feedback strategies and embed privacy-preserving methods to enable ethically sound deployment in daily scenarios. CLINICALTRIAL Registered with PROSPERO (ID: CRD42025637387).
- 22 June 2025
- Published by JMIR Publications Inc. in JMIR Preprints
Open AccessA Systematic Review of Prolonged Posture Monitoring With Accelerometer-Based Wearable Sensors (Preprint)
BACKGROUND Prolonged postures, such as sustained sitting and standing, contribute to musculoskeletal disorders, yet daily monitoring approaches vary widely, limiting comparability and the development of standardized posture monitoring strategies. OBJECTIVE This systematic review aimed to summarize current practices in prolonged posture monitoring using accelerometer-based wearable sensors and to identify gaps across four stages: data collection, data preprocessing, data analysis, and user feedback. METHODS This systematic review synthesized 79 studies (82.3% rated low risk of bias) extracted from 16645 records in 10 databases, including PubMed, IEEE Xplore, Scopus, Web of Science, EI, ACM, ProQuest, CINAHL, SPORTDiscus, and EMBASE. The risk of bias was assessed using NOS, QUADAS-2, ROBINS-I, RoB 2, and a self-developed tool for algorithmic studies, with 87.7% rated as low risk. Due to methodological heterogeneity, results were synthesized qualitatively. RESULTS Thigh- and waist-worn sensors were most common, particularly activPAL and ActiGraph. Sensor placement, population characteristics, and preprocessing methods varied widely. Most studies used manually engineered features with statistical analysis, yet few leveraged deep learning or biomechanical metrics. Only 11 studies implemented user feedback strategies, typically visual or vibratory. Personalized user feedback remains underexplored, and few studies have considered specific free-living contexts or privacy concerns. CONCLUSIONS Findings highlight the promising potential of accelerometer-based posture monitoring systems. However, a lack of standardization in sensor placement and data processing, underrepresentation of diverse populations, and limited real-world feedback mechanisms constrain scalability. Future research should develop standard protocols for sensor placement, include diverse population groups, and prioritize the use of interpretable machine learning to support robust classification. Developers should design mobile-compatible systems with personalized, real-time feedback strategies and embed privacy-preserving methods to enable ethically sound deployment in daily scenarios. CLINICALTRIAL Registered with PROSPERO (ID: CRD42025637387).
- 21 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 74 (00189456) , 1-17
MSEvoNet: Multiscale Spatiotemporal Evolution Networks for Remaining Useful Life Prediction in Complex Industrial Systems
This paper introduces MSEvoNet, a multiscale spatiotemporal evolution network that analyzes complex industrial systems’ dynamic changes during service, focusing on predicting the remaining useful life (RUL). By integrating a gated dilated convolution module, an adaptive relationship perception module, and a diffusion graph convolution module in a step-by-step stacking approach, MSEvoNet effectively enhances heterogeneous data management, spatiotemporal fusion and model interpretability in complex systems, providing deeper insights into industrial system degradation. The adopted spatiotemporal graph structure portrayed the evolutionary process of complex industrial systems, depicting the intricate interactions and influences among spatiotemporal variables during degradation and the intrinsic conduits governing RUL factors, thereby constructing an eXplainable Artificial Intelligence (XAI) model for complex industrial systems. Validated with the C-MAPSS and N-CMAPSS datasets, representing complex aerospace engines’ systems, MSEvoNet demonstrated robustness and adaptability across diverse operational scenarios, significantly advancing the potential for RUL prediction. The findings underscore the importance of leveraging XAI and multi-scale spatiotemporal insights for RUL prediction, offering a pathway towards more reliable and efficient prognostics for complex industrial systems in the Industry 4.0 and 5.0 era.
- 21 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 74 (00189456) , 1-17
MSEvoNet: Multiscale Spatiotemporal Evolution Networks for Remaining Useful Life Prediction in Complex Industrial Systems
This paper introduces MSEvoNet, a multiscale spatiotemporal evolution network that analyzes complex industrial systems’ dynamic changes during service, focusing on predicting the remaining useful life (RUL). By integrating a gated dilated convolution module, an adaptive relationship perception module, and a diffusion graph convolution module in a step-by-step stacking approach, MSEvoNet effectively enhances heterogeneous data management, spatiotemporal fusion and model interpretability in complex systems, providing deeper insights into industrial system degradation. The adopted spatiotemporal graph structure portrayed the evolutionary process of complex industrial systems, depicting the intricate interactions and influences among spatiotemporal variables during degradation and the intrinsic conduits governing RUL factors, thereby constructing an eXplainable Artificial Intelligence (XAI) model for complex industrial systems. Validated with the C-MAPSS and N-CMAPSS datasets, representing complex aerospace engines’ systems, MSEvoNet demonstrated robustness and adaptability across diverse operational scenarios, significantly advancing the potential for RUL prediction. The findings underscore the importance of leveraging XAI and multi-scale spatiotemporal insights for RUL prediction, offering a pathway towards more reliable and efficient prognostics for complex industrial systems in the Industry 4.0 and 5.0 era.
- 19 September 2024
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 35 (12) , 125019
MSTAN: multi-scale spatiotemporal attention network with adaptive relationship mining for remaining useful life prediction in complex systems
In the era of smart manufacturing and advanced industrial systems, the high degree of integration and intelligence of equipment demands higher reliability and safety from systems. Existing methods often rely on historical data for Remaining Useful Life (RUL) prediction to achieve Prognostic and Health Management (PHM). However, the internal units of complex equipment exhibit significant spatial correlation and temporal diversity, making PHM for complex equipment a multidimensional challenge involving both temporal and spatial information, thereby severely limits the effectiveness of RUL prediction for complex systems. Addressing these challenges, this study introduces a multi-scale spatiotemporal attention network with adaptive relationship mining, specifically designed for the remaining useful life (RUL) prediction of such equipment. The core of the proposed method lies in the multi-scale feature perception module, which adeptly extracts varied scale features from multidimensional sensor data. Following this, an innovative adaptive relationship mining module is integrated to uncover multi-order coupling relationships between diverse sensors, enhancing the model’s predictive accuracy. Furthermore, a spatiotemporal attention module is employed to discern and emphasize crucial spatiotemporal correlations. To validate the effectiveness and superiority of the proposed method, the Commercial Modular Aero-propulsion System Simulation (C-MAPSS) dataset is employed for comprehensive performance evaluation, the IEEE 2012 PHM bearing dataset is also adopted to demonstrate the generalization and robustness of the proposed method. The results not only show a notable improvement over existing methods but also offer a more intuitive understanding through visual representations, marking a significant stride in enhancing the safety and efficiency of complex systems.
- 19 September 2024
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 35 (12) , 125019
MSTAN: multi-scale spatiotemporal attention network with adaptive relationship mining for remaining useful life prediction in complex systems
In the era of smart manufacturing and advanced industrial systems, the high degree of integration and intelligence of equipment demands higher reliability and safety from systems. Existing methods often rely on historical data for Remaining Useful Life (RUL) prediction to achieve Prognostic and Health Management (PHM). However, the internal units of complex equipment exhibit significant spatial correlation and temporal diversity, making PHM for complex equipment a multidimensional challenge involving both temporal and spatial information, thereby severely limits the effectiveness of RUL prediction for complex systems. Addressing these challenges, this study introduces a multi-scale spatiotemporal attention network with adaptive relationship mining, specifically designed for the remaining useful life (RUL) prediction of such equipment. The core of the proposed method lies in the multi-scale feature perception module, which adeptly extracts varied scale features from multidimensional sensor data. Following this, an innovative adaptive relationship mining module is integrated to uncover multi-order coupling relationships between diverse sensors, enhancing the model’s predictive accuracy. Furthermore, a spatiotemporal attention module is employed to discern and emphasize crucial spatiotemporal correlations. To validate the effectiveness and superiority of the proposed method, the Commercial Modular Aero-propulsion System Simulation (C-MAPSS) dataset is employed for comprehensive performance evaluation, the IEEE 2012 PHM bearing dataset is also adopted to demonstrate the generalization and robustness of the proposed method. The results not only show a notable improvement over existing methods but also offer a more intuitive understanding through visual representations, marking a significant stride in enhancing the safety and efficiency of complex systems.
- 29 February 2024
- Published by Taylor & Francis in Ergonomics
- Vol. 67 (10) , 1356-1370
Improved REBA: deep learning based rapid entire body risk assessment for prevention of musculoskeletal disorders
Preventing work-related musculoskeletal disorders (WMSDs) is crucial in reducing their impact on individuals and society. However, the existing mainstream 2D image-based approach is insufficient in capturing the complex 3D movements and postures involved in many occupational tasks. To address this, an improved deep learning-based rapid entire body assessment (REBA) method has been proposed. The method takes working videos as input and automatically outputs the corresponding REBA score through 3D pose reconstruction. The proposed method achieves an average precision of 94.7% on real-world data, which is comparable to that of ergonomic experts. Furthermore, the method has the potential to be applied across a wide range of industries as it has demonstrated good generalisation in multiple scenarios. The proposed method offers a promising solution for automated and accurate risk assessment of WMSDs, with implications for various industries to ensure the safety and well-being of workers.
- 29 February 2024
- Published by Taylor & Francis in Ergonomics
- Vol. 67 (10) , 1356-1370
Improved REBA: deep learning based rapid entire body risk assessment for prevention of musculoskeletal disorders
Preventing work-related musculoskeletal disorders (WMSDs) is crucial in reducing their impact on individuals and society. However, the existing mainstream 2D image-based approach is insufficient in capturing the complex 3D movements and postures involved in many occupational tasks. To address this, an improved deep learning-based rapid entire body assessment (REBA) method has been proposed. The method takes working videos as input and automatically outputs the corresponding REBA score through 3D pose reconstruction. The proposed method achieves an average precision of 94.7% on real-world data, which is comparable to that of ergonomic experts. Furthermore, the method has the potential to be applied across a wide range of industries as it has demonstrated good generalisation in multiple scenarios. The proposed method offers a promising solution for automated and accurate risk assessment of WMSDs, with implications for various industries to ensure the safety and well-being of workers.
- 23 November 2023
- Published by Elsevier in Applied Ergonomics
- Vol. 115, 104176
Spatial relationship-aware rapid entire body fuzzy assessment method for prevention of work-related musculoskeletal disorders
- 23 November 2023
- Published by Elsevier in Applied Ergonomics
- Vol. 115, 104176
Spatial relationship-aware rapid entire body fuzzy assessment method for prevention of work-related musculoskeletal disorders
- Open Access
Modified DEMATEL Method Based on Objective Data Grey Relational Analysis for Time Series
Smart data selection can quickly sieve valuable information from initial data. Doing so improves the efficiency of analyzing situations to aid in better decision-making. Past methods have mostly been based on expert experience, which may be subjective and inefficient when dealing with large, complex datasets. Recently, the system analysis method has been exploited to find the key data. However, few studies address the indirect effects and heterogeneity of time series data. In this study, a data selection method, the modified Decision-Making Trial and Evaluation Laboratory (DEMATEL) method based on the objective data grey relational analysis (GRA), is used to enhance the ability to analyze time-series data. GRA was first applied to assess the direct impact in the raw data indicators. Then, a modified DEMATEL was adopted to find the overall impact by including the indirect impact and data heterogeneity. We applied the method to analyze the Commercial Modular Aero-Propulsion System Simulation (C-MAPSS) dataset and perform the remaining useful life (RUL) prediction of aircraft engines. The results suggest that our method predicts well. Our work offers a nuanced approach of identifying key information in time series data and has potential applications.
Open AccessModified DEMATEL Method Based on Objective Data Grey Relational Analysis for Time Series
Smart data selection can quickly sieve valuable information from initial data. Doing so improves the efficiency of analyzing situations to aid in better decision-making. Past methods have mostly been based on expert experience, which may be subjective and inefficient when dealing with large, complex datasets. Recently, the system analysis method has been exploited to find the key data. However, few studies address the indirect effects and heterogeneity of time series data. In this study, a data selection method, the modified Decision-Making Trial and Evaluation Laboratory (DEMATEL) method based on the objective data grey relational analysis (GRA), is used to enhance the ability to analyze time-series data. GRA was first applied to assess the direct impact in the raw data indicators. Then, a modified DEMATEL was adopted to find the overall impact by including the indirect impact and data heterogeneity. We applied the method to analyze the Commercial Modular Aero-Propulsion System Simulation (C-MAPSS) dataset and perform the remaining useful life (RUL) prediction of aircraft engines. The results suggest that our method predicts well. Our work offers a nuanced approach of identifying key information in time series data and has potential applications.