Wujun Si
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Wujun Si has a total of 26 co-authors. Here are the top co-authors by number of shared publications:
- Yunfei Shao
Department of Industrial, Systems and Manufacturing Engineering, Wichita State University, Wichita, Kansas
3 shared publications Ali Asgari
2 shared publications- Hao Zeng
School of Media Engineering, Communication University of Zhejiang, Hangzhou, Zhejiang, China
2 shared publications Krishna Krishnan
2 shared publicationsLifeng Xi
2 shared publications- Mingyang Li
Department of Industrial and Management Systems Engineering, College of Engineering, University of South Florida, 4202 E. Fowler Avenue, Tampa, Florida 33620
2 shared publications
Latest Publications
- 22 May 2026
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 37 (21) , 216118
Unsupervised machinery health indicator construction based on a temporal adversarial neural wavelet model
Unsupervised deep learning-based (DL-based) health indicator (HI) exhibits broad application potential in machinery condition monitoring of long-lifespan and high-value equipment, as it enables degradation-discriminative feature analysis without requiring end-of-life data. However, the evolution patterns of discriminative features extracted by traditional DL workflows still lack physical consistency with actual degradation behavior, which limits the credibility and performance of the constructed HI. For this issue, this study proposes a novel HI construction model that extracts discriminative features to appropriately represent the physical degradation behavior and constructs an HI endowed with an explicit statistical expression. Firstly, a neural wavelet decomposition framework performs multi-resolution sub-band feature extraction on degradation signals using learnable wavelet bases. It is trained with a designed temporal adversarial learning scheme to suppress the confounding influence of operating time and enhance the intrinsic correlation between degradation-discriminative features and the actual physical degradation behavior. Subsequently, each sub-band is independently characterized using the alpha-stable distribution. A dual-domain HI construction strategy, which measures the probability density and the distributional characteristic, constructs independent HIs for each sub-band. It provides explicit statistical analysis of degradation-discriminative features’ evolution based on the theoretical foundation of fault characteristic representation in signal morphology. Finally, an adaptive correlation-aware fusion mechanism integrates these sub-band HIs into a unified HI. Validation results on multiple industrial datasets demonstrate the superior performance of the proposed HI.
- 22 May 2026
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 37 (21) , 216118
Unsupervised machinery health indicator construction based on a temporal adversarial neural wavelet model
Unsupervised deep learning-based (DL-based) health indicator (HI) exhibits broad application potential in machinery condition monitoring of long-lifespan and high-value equipment, as it enables degradation-discriminative feature analysis without requiring end-of-life data. However, the evolution patterns of discriminative features extracted by traditional DL workflows still lack physical consistency with actual degradation behavior, which limits the credibility and performance of the constructed HI. For this issue, this study proposes a novel HI construction model that extracts discriminative features to appropriately represent the physical degradation behavior and constructs an HI endowed with an explicit statistical expression. Firstly, a neural wavelet decomposition framework performs multi-resolution sub-band feature extraction on degradation signals using learnable wavelet bases. It is trained with a designed temporal adversarial learning scheme to suppress the confounding influence of operating time and enhance the intrinsic correlation between degradation-discriminative features and the actual physical degradation behavior. Subsequently, each sub-band is independently characterized using the alpha-stable distribution. A dual-domain HI construction strategy, which measures the probability density and the distributional characteristic, constructs independent HIs for each sub-band. It provides explicit statistical analysis of degradation-discriminative features’ evolution based on the theoretical foundation of fault characteristic representation in signal morphology. Finally, an adaptive correlation-aware fusion mechanism integrates these sub-band HIs into a unified HI. Validation results on multiple industrial datasets demonstrate the superior performance of the proposed HI.
- Open Access
Dynamic Weight Adjusting Deep Q-Networks for Real-Time Environmental Adaptation
Deep Reinforcement Learning has shown excellent performance in generating efficient solutions for complex tasks. However, its efficacy is often limited by static training modes and heavy reliance on vast data from stable environments. To address these shortcomings, this study explores integrating dynamic weight adjustments into Deep Q-Networks (DQN) to enhance their adaptability. We implement these adjustments by modifying the sampling probabilities in the experience replay to make the model focus more on pivotal transitions as indicated by real-time environmental feedback and performance metrics. We design a novel Interactive Dynamic Evaluation Method (IDEM) for DQN that successfully navigates dynamic environments by prioritizing significant transitions based on environmental feedback and learning progress. Additionally, when faced with rapid changes in environmental conditions, IDEM-DQN shows improved performance compared to baseline methods. Our results indicate that under circumstances requiring rapid adaptation, IDEM-DQN can more effectively generalize and stabilize learning. Extensive experiments across various settings confirm that IDEM-DQN outperforms standard DQN models, particularly in environments characterized by frequent and unpredictable changes.
Open AccessDynamic Weight Adjusting Deep Q-Networks for Real-Time Environmental Adaptation
Deep Reinforcement Learning has shown excellent performance in generating efficient solutions for complex tasks. However, its efficacy is often limited by static training modes and heavy reliance on vast data from stable environments. To address these shortcomings, this study explores integrating dynamic weight adjustments into Deep Q-Networks (DQN) to enhance their adaptability. We implement these adjustments by modifying the sampling probabilities in the experience replay to make the model focus more on pivotal transitions as indicated by real-time environmental feedback and performance metrics. We design a novel Interactive Dynamic Evaluation Method (IDEM) for DQN that successfully navigates dynamic environments by prioritizing significant transitions based on environmental feedback and learning progress. Additionally, when faced with rapid changes in environmental conditions, IDEM-DQN shows improved performance compared to baseline methods. Our results indicate that under circumstances requiring rapid adaptation, IDEM-DQN can more effectively generalize and stabilize learning. Extensive experiments across various settings confirm that IDEM-DQN outperforms standard DQN models, particularly in environments characterized by frequent and unpredictable changes.
- 23 April 2024
- Published by Elsevier in Reliability Engineering & System Safety
- Vol. 248
Multivariable degradation modeling and life prediction using multivariate fractional Brownian motion
- 23 April 2024
- Published by Elsevier in Reliability Engineering & System Safety
- Vol. 248
Multivariable degradation modeling and life prediction using multivariate fractional Brownian motion
- Open Access
- 29 February 2024
- Published by MDPI AG in Mathematics
- Vol. 12 (5) , 740
A Bayesian Approach for Lifetime Modeling and Prediction with Multi-Type Group-Shared Missing Covariates
In the field of reliability engineering, covariate information shared among product units within a specific group (e.g., a manufacturing batch, an operating region), such as operating conditions and design settings, exerts substantial influence on product lifetime prediction. The covariates shared within each group may be missing due to sensing limitations and data privacy issues. The missing covariates shared within the same group commonly encompass a variety of attribute types, such as discrete types, continuous types, or mixed types. Existing studies have mainly considered single-type missing covariates at the individual level, and they have failed to thoroughly investigate the influence of multi-type group-shared missing covariates. Ignoring the multi-type group-shared missing covariates may result in biased estimates and inaccurate predictions of product lifetime, subsequently leading to suboptimal maintenance decisions with increased costs. To account for the influence of the group-shared missing covariates with different structures, a new flexible lifetime model with multi-type group-shared latent heterogeneity is proposed. We further develop a Bayesian estimation algorithm with data augmentation that jointly quantifies the influence of both observed and multi-type group-shared missing covariates on lifetime prediction. A tripartite method is then developed to examine the existence, identify the correct type, and quantify the influence of group-shared missing covariates. To demonstrate the effectiveness of the proposed approach, a comprehensive simulation study is carried out. A real case study involving tensile testing of molding material units is conducted to validate the proposed approach and demonstrate its practical applicability.
- 29 February 2024
- Published by MDPI AG in Mathematics
- Vol. 12 (5) , 740
Open AccessA Bayesian Approach for Lifetime Modeling and Prediction with Multi-Type Group-Shared Missing Covariates
In the field of reliability engineering, covariate information shared among product units within a specific group (e.g., a manufacturing batch, an operating region), such as operating conditions and design settings, exerts substantial influence on product lifetime prediction. The covariates shared within each group may be missing due to sensing limitations and data privacy issues. The missing covariates shared within the same group commonly encompass a variety of attribute types, such as discrete types, continuous types, or mixed types. Existing studies have mainly considered single-type missing covariates at the individual level, and they have failed to thoroughly investigate the influence of multi-type group-shared missing covariates. Ignoring the multi-type group-shared missing covariates may result in biased estimates and inaccurate predictions of product lifetime, subsequently leading to suboptimal maintenance decisions with increased costs. To account for the influence of the group-shared missing covariates with different structures, a new flexible lifetime model with multi-type group-shared latent heterogeneity is proposed. We further develop a Bayesian estimation algorithm with data augmentation that jointly quantifies the influence of both observed and multi-type group-shared missing covariates on lifetime prediction. A tripartite method is then developed to examine the existence, identify the correct type, and quantify the influence of group-shared missing covariates. To demonstrate the effectiveness of the proposed approach, a comprehensive simulation study is carried out. A real case study involving tensile testing of molding material units is conducted to validate the proposed approach and demonstrate its practical applicability.
- 14 November 2023
- Published by Taylor & Francis in Journal of Quality Technology
- Vol. 56 (2) , 91-112
Spatial modeling and monitoring considering long-range dependence
- 14 November 2023
- Published by Taylor & Francis in Journal of Quality Technology
- Vol. 56 (2) , 91-112
Spatial modeling and monitoring considering long-range dependence
- 23 December 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Reliability
- Vol. 72 (1) , 177-189
Degradation Modeling With Long-Term Memory Considering Measurement Errors
With the advancement of measurement technology, the long-term memory (LTM) effect within the degradation data of many assets has recently been detected, which implies that the future degradation process highly correlates with both the current degradation status and historical degradation trajectory across a long time period. To capture the LTM effect, several LTM-integrated degradation models have been developed in the literature. In practice, degradation data are often contaminated with measurement errors, which are ignored in most existing LTM-integrated degradation studies. Without considering measurement errors, the asset degradation modeling and reliability analysis may be biased. In this article, we propose a novel LTM-integrated degradation model that quantitatively incorporates measurement errors. Both fixed-effect and random-effect scenarios of degradation growth are considered. A maximum-likelihood estimation approach is developed to estimate the parameters of the proposed model. Based on this model, asset reliability analysis and lifetime prediction are developed. Simulation studies are implemented to evaluate the performance of the proposed model. A real case study using the capacity degradation data of lithium-ion pouch cells is conducted to illustrate the superiority of the proposed model. Results demonstrate that conventional LTM-integrated degradation models, which ignore the measurement errors, significantly misestimate the uncertainty of asset lifetime.
- 23 December 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Reliability
- Vol. 72 (1) , 177-189
Degradation Modeling With Long-Term Memory Considering Measurement Errors
With the advancement of measurement technology, the long-term memory (LTM) effect within the degradation data of many assets has recently been detected, which implies that the future degradation process highly correlates with both the current degradation status and historical degradation trajectory across a long time period. To capture the LTM effect, several LTM-integrated degradation models have been developed in the literature. In practice, degradation data are often contaminated with measurement errors, which are ignored in most existing LTM-integrated degradation studies. Without considering measurement errors, the asset degradation modeling and reliability analysis may be biased. In this article, we propose a novel LTM-integrated degradation model that quantitatively incorporates measurement errors. Both fixed-effect and random-effect scenarios of degradation growth are considered. A maximum-likelihood estimation approach is developed to estimate the parameters of the proposed model. Based on this model, asset reliability analysis and lifetime prediction are developed. Simulation studies are implemented to evaluate the performance of the proposed model. A real case study using the capacity degradation data of lithium-ion pouch cells is conducted to illustrate the superiority of the proposed model. Results demonstrate that conventional LTM-integrated degradation models, which ignore the measurement errors, significantly misestimate the uncertainty of asset lifetime.
- 16 June 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Reliability
- Vol. 69 (4) , 1254-1266
Accelerated Degradation Testing With Long-Term Memory Effects
The accelerated degradation testing (ADT) has been widely applied as an efficient strategy to obtain the reliability (life) information of the assets in a shorter-than-normal period of time by exposing the assets to higher-than-normal stresses. Recently, with advances in the sensor technology, it has been revealed that the degradation of some assets demonstrates a long-term memory effect, which implies that the future degradation process not only depends on the current degradation state but also strongly correlates with the past degradation history across a long period of time, and the degradation increments are correlated for nonoverlapping time intervals. The existing ADT methods do not consider the long-term memories, which could lead to biased life testing results. In this article, we propose a novel ADT model by integrating the long-term degradation memory effect based on a utilization of the fractional Brownian motion. A maximum likelihood approach is developed to estimate the model parameters. A likelihood-ratio hypothesis test is designed to test the existence of long-term memories. Simulation studies are implemented to illustrate the developed methods. Physical experiments on accelerated testing of a photocatalyst are designed and conducted to demonstrate the proposed model and its advantage over benchmark approaches. The results show that the traditional ADT paradigm, which ignores the long-term memories, significantly underestimates asset lifetime uncertainties.
- 16 June 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Reliability
- Vol. 69 (4) , 1254-1266
Accelerated Degradation Testing With Long-Term Memory Effects
The accelerated degradation testing (ADT) has been widely applied as an efficient strategy to obtain the reliability (life) information of the assets in a shorter-than-normal period of time by exposing the assets to higher-than-normal stresses. Recently, with advances in the sensor technology, it has been revealed that the degradation of some assets demonstrates a long-term memory effect, which implies that the future degradation process not only depends on the current degradation state but also strongly correlates with the past degradation history across a long period of time, and the degradation increments are correlated for nonoverlapping time intervals. The existing ADT methods do not consider the long-term memories, which could lead to biased life testing results. In this article, we propose a novel ADT model by integrating the long-term degradation memory effect based on a utilization of the fractional Brownian motion. A maximum likelihood approach is developed to estimate the model parameters. A likelihood-ratio hypothesis test is designed to test the existence of long-term memories. Simulation studies are implemented to illustrate the developed methods. Physical experiments on accelerated testing of a photocatalyst are designed and conducted to demonstrate the proposed model and its advantage over benchmark approaches. The results show that the traditional ADT paradigm, which ignores the long-term memories, significantly underestimates asset lifetime uncertainties.