Article
Computer Science, Artificial Intelligence
Yuxiong Li, Xianzhen Huang, Tianhong Gao, Chengying Zhao, Shangjie Li
Summary: In this paper, a novel method for remaining useful life (RUL) prediction considering multiple degradation patterns is developed. The method estimates degradation parameters and predicts RUL through offline and online algorithms. Comparison with other methods shows that the proposed method achieves higher accuracy.
ADVANCED ENGINEERING INFORMATICS
(2023)
Article
Computer Science, Interdisciplinary Applications
Baoping Cai, Hongyan Fan, Xiaoyan Shao, Yonghong Liu, Guijie Liu, Zengkai Liu, Renjie Ji
Summary: This study proposed a RUL re-prediction method based on the Wiener process, which combines the current monitoring status and historical degradation data of the system. The dynamic Bayesian networks model is used to address the uncertainty caused by missing data.
COMPUTERS & INDUSTRIAL ENGINEERING
(2021)
Article
Engineering, Industrial
Wennian Yu, Yimin Shao, Jin Xu, Chris Mechefske
Summary: This paper introduces a generalized Wiener process-based degradation model with an adaptive drift to characterize complex degradation behaviors. A recursive Bayesian filtering algorithm is used to update the drift distribution and an expectation-maximization algorithm is utilized for online estimation of model parameters. An analytical approximation of the Remaining Useful Life distribution is derived and validated using a practical milling dataset, demonstrating superior performance compared to existing methods.
RELIABILITY ENGINEERING & SYSTEM SAFETY
(2022)
Article
Engineering, Industrial
Kai Song, Lirong Cui
Summary: Due to the complexity and multi-functionality of modern products, there are usually multiple performance characteristics that can reflect the degradation states. This paper proposes a gamma process based degradation model for analyzing bivariate dependent degradation data. The model captures the dependency between the two degradation processes using a common random effect. A real-time prediction method for the remaining useful life of the product is also proposed using Bayesian method.
RELIABILITY ENGINEERING & SYSTEM SAFETY
(2022)
Article
Engineering, Multidisciplinary
Dezhong Wu, Minping Jia, Yudong Cao, Peng Ding, Xiaoli Zhao
Summary: This paper proposes an adaptive nonlinear Wiener process model to predict the remaining useful life (RUL) of rolling bearings, using the closed skew-normal (CSN) distribution to model the non-normal and asymmetrical degradation characteristic in inter-product variation. An online recursive algorithm based on Bayes' theorem is derived to update the hidden drift distribution, and a Bayesian smoother based on CSN distribution is developed to estimate model parameters and derive an analytical expression of the RUL distribution.
Article
Engineering, Industrial
Guobo Liao, Hongpeng Yin, Min Chen, Zheng Lin
Summary: The paper presents a multi-phase degradation model based on Wiener process and uses a Bayesian approach to integrate historical and real-time data for remaining useful life prediction. By estimating hyperparameters using the expectation maximization algorithm, model parameters can be effectively updated to account for multiple random change points.
RELIABILITY ENGINEERING & SYSTEM SAFETY
(2021)
Article
Engineering, Industrial
Xiaodong Xu, Shengjin Tang, Chuanqiang Yu, Jian Xie, Xuebing Han, Minggao Ouyang
Summary: This paper proposes a novel prediction method for remaining useful life (RUL) of lithium-ion battery under time-varying temperature condition based on Arrhenius temperature model and Wiener process. The aging model is developed using maximum likelihood estimation and genetic algorithm, leading to the derivation of probability density function (PDF) of RUL. The effectiveness of the method is verified through a case study, demonstrating higher accuracy and smaller uncertainty.
RELIABILITY ENGINEERING & SYSTEM SAFETY
(2021)
Article
Automation & Control Systems
Mohammad S. Ramadan, Robert R. Bitmead
Summary: A Maximum Likelihood recursive state estimator is proposed for non-linear state-space models, which combines a particle filter and the Expectation Maximization algorithm. Algorithms for maximum likelihood state filtering, prediction, and smoothing are derived, and their convergence properties are examined, demonstrating the effectiveness of the method in nonlinear systems.
Article
Engineering, Electrical & Electronic
Mursel Yildiz
Summary: The expectation maximization algorithm for univariate problems often requires prior information, which can be problematic for highly dynamic environments. This study presents an EM approach based on Fourier series that can approximate the true probability distribution function and ensure tractability and closed form.
Article
Engineering, Electrical & Electronic
A. Hippert Ferrer, M. N. El Korso, A. Breloy, G. Ginolhac
Summary: This paper addresses robust mean and covariance matrix estimation in mixed effects models, proposing an expectation-conditional maximization algorithm to handle outliers, parallelized for computational efficiency and extended to deal with missing data. Numerical simulations evaluate the performance in robust regression estimators, probabilistic principal component analysis, and its robust version.
Article
Engineering, Industrial
Xingheng Liu, Jose Matias, Johannes Jaschke, Jorn Vatn
Summary: This paper focuses on predicting degradation growth and estimating remaining useful life based on noisy observations by introducing a Transformed Gamma process model and using an improved Gibbs sampler with Expectation-Maximization. The study discusses degradation increment prediction and parameter estimation under noisy conditions, as well as derives false/failed alarm probability and remaining useful life distribution.
RELIABILITY ENGINEERING & SYSTEM SAFETY
(2022)
Article
Engineering, Industrial
Xuefeng Kong, Jun Yang, Lei Li
Summary: A general degradation model considering random shock fluctuations and measurement uncertainty was developed to describe the degradation process, with a two-step approach and expectation-maximization algorithm used for parameter estimation. The effectiveness of the method was verified through numerical examples and practical case studies.
JOURNAL OF MANUFACTURING SYSTEMS
(2021)
Article
Automation & Control Systems
Man Zheng, Yoshito Ohta
Summary: This paper introduces a new method for Bayesian identification of positive finite impulse response (FIR) models by using a truncated Gaussian prior. The parameterization in the truncated Gaussian prior can better reflect the characteristics of the impulse response of the system to be identified. Compared to the traditional Gaussian prior, the truncated Gaussian prior outperforms in positive FIR system identification.
SYSTEMS & CONTROL LETTERS
(2021)
Article
Engineering, Electrical & Electronic
Shuxin Zhang, Zhitao Liu, Hongye Su
Summary: This article proposes a deep learning method for RUL prediction using a Bayesian mixture neural network and quasi-Gramian angular field to improve the accuracy of prediction and uncertainty estimation.
IEEE TRANSACTIONS ON TRANSPORTATION ELECTRIFICATION
(2022)
Article
Multidisciplinary Sciences
Mohamed Kayid, Abdulrahman Abouammoh, Ghadah Alomani, Dario Ferreira, Calogero Vetro
Summary: Statistical probability models are often used to analyze real-world data in many research fields. However, data from fields such as the environment, economics, and health care may not fit traditional models. This study investigates an extension of the quasi-Lindley model that is asymmetrically distributed on the positive real number line. Various algorithms are used to estimate the parameters, and the results show that all techniques provide accurate and reliable estimates. The proposed model outperforms alternative models when analyzing a reliability dataset.
Article
Engineering, Electrical & Electronic
Zhenan Pang, Hong Pei, Tianmei Li, Jianxun Zhang, Changhua Hu, Xiaosheng Si
Summary: This paper proposes an adaptive RUL estimation method for partially observable degrading products. The method includes modeling, estimation, and prognostic aspects, utilizing STF and ECM algorithms to estimate states and parameters, and obtaining the RUL distribution. The accuracy and effectiveness of the proposed approach is verified through numerical examples and case studies.
IEEE SENSORS JOURNAL
(2021)
Article
Computer Science, Hardware & Architecture
Jian-Xun Zhang, Dang-Bo Du, Xiao-Sheng Si, Yang Liu, Chang-Hua Hu
Summary: Degradation-model-based RUL estimation is crucial for effective prognostic and health management, with stochastic-process-based methods widely preferred for capturing stochastic dynamics. Existing studies often underestimate RUL using the conservative first passage time (FPT) method, especially in nonmonotonic stochastic degradation processes. By introducing the last exit time (LET) perspective, this study provides a new definition for lifetime/RUL estimation, illustrated using the Wiener-process-based model with examples and solutions. The proposed method shows potential in preventing premature maintenance and resource wastage by avoiding conservative results.
IEEE TRANSACTIONS ON RELIABILITY
(2021)
Article
Computer Science, Information Systems
Wei You, Xue Wang, Weihang Zhang, Zhenfeng Qiang
Summary: This paper proposes a multi-level kinematic constraints method to construct multiple skeleton features, which can effectively utilize valid information and enhance the performance of skeleton-based action recognition methods.
MULTIMEDIA TOOLS AND APPLICATIONS
(2022)
Article
Engineering, Electrical & Electronic
Junfeng Chen, Xue Wang, Xiaotian Zhang, Weihang Zhang
Summary: Non-intrusive load monitoring (NILM) is a promising technology that can monitor appliance operating state and energy consumption without sub-meters. This paper proposes a method using temporal and spectral load signatures for appliance recognition in NILM. Deep learning techniques and affinity propagation clustering strategy are used to extract features and mitigate the negative impact of multi-state loads. Experimental results show that the proposed method outperforms existing methods in recognition accuracy.
IEEE TRANSACTIONS ON SMART GRID
(2022)
Article
Engineering, Electrical & Electronic
Qin Hu, Xiaosheng Si, Aisong Qin, Yunrong Lv, Mei Liu
Summary: This study proposes a novel fault diagnosis method based on enhanced multi-scale sample entropies and balanced adaptation regularization based transfer learning to address the issue of inconsistent distribution between training and testing data in fault diagnosis. The method enhances feature discriminability and similarity of fault information, and uses balanced adaptation regularization based transfer learning to learn an adaptive classifier for cross-domain fault diagnosis.
IEEE SENSORS JOURNAL
(2022)
Article
Automation & Control Systems
Hong Pei, Xiao-Sheng Si, Changhua Hu, Tianmei Li, Chuan He, Zhenan Pang
Summary: Deep learning has become a promising tool for processing massive data and has gained attention in degradation modeling and remaining useful life (RUL) prediction. However, existing methods face challenges in representing prediction uncertainty and training without label data. To address these issues, this study proposes a prognostic model based on Bayesian deep learning and verifies its feasibility through a case study.
IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS
(2023)
Article
Engineering, Electrical & Electronic
Zhengxin Zhang, Tianmei Li, Jianxun Zhang, Dangbo Du, Xiaosheng Si
Summary: Degradation modeling-based remaining useful life (RUL) predicting has gained significant attention as a basis for reliability prognotics and system health management (PHM) in the field of reliability. This article presents a nonlinear degradation model based on a purely time-dependent diffusion process, focusing on the proportional relationship between the age-dependent expectation and variance of the degradation processes. The article derives explicit expressions for probability density function (pdf) and cumulative distribution function (cdf) of lifetime and RUL, incorporating item-to-item variability using the concept of first hitting time (FHT). A framework for estimating unknown parameters using condition monitoring (CM) data is proposed, and case studies are conducted to validate the proposed prognostics model using fatigue crack length data of alloy and capacity data of electrolytic capacitors. The results demonstrate the effectiveness of the proposed model in accurately predicting RUL.
IEEE SENSORS JOURNAL
(2023)
Article
Automation & Control Systems
Haiming Yao, Wenyong Yu, Xue Wang
Summary: Recent advances in industrial inspection of textured surfaces have made efficient and flexible manufacturing systems possible. This paper proposes an unsupervised feature memory rearrangement network (FMR-Net) that accurately detects various textural defects simultaneously. The network utilizes background reconstruction and artificial synthetic defects to recognize anomalies, achieving state-of-the-art inspection accuracy and showing great potential for use in smart industries.
IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING
(2022)
Article
Engineering, Electrical & Electronic
Qiang Liu, Yijie Zhang, Xiaosheng Si, Zizhu Fan
Summary: This article proposes a novel data-driven bearing degradation modeling method, called dynamic latent variable reconstruction nonlinear Wiener process (DLVR-NWP). The method extracts reduced-dimensional degradation-relevant dynamic features from time-domain and frequency-domain features, and incorporates a DLV-based nonlinear degradation detection mechanism into the RUL estimation model. Case studies on real bearings show that the proposed method significantly improves the accuracy of RUL estimation compared to traditional methods.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2023)
Article
Computer Science, Artificial Intelligence
Hong Pei, Xiaosheng Si, Tianmei Li, Zhengxin Zhang, Yaguo Lei
Summary: This article proposes an interactive prognosis framework between deep learning and a stochastic process model for the prediction of remaining useful life (RUL). The framework includes stacked contractive autoencoders for health indicator (HI) construction and an exponential-like degradation model for degradation modeling. By optimizing the objective function and using a gradient descent algorithm, the accuracy of the prediction results is improved.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Computer Science, Information Systems
Wei You, Xue Wang
Summary: A novel self-supervised learning method is proposed in this study, which introduces the view enhanced jigsaw puzzle (VEJP) pretext task and the view pooling encoder (VPE) to improve feature learning. Experimental results show that moderately difficult pretext tasks can effectively enhance feature learning.
Article
Computer Science, Artificial Intelligence
Weihang Zhang, Xue Wang, Junfeng Chen, Wei You
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2020)
Article
Computer Science, Information Systems
Tianmei Li, Xiaosheng Si, Zonghao Yang, Hong Pei, Yuzhe Ma
Article
Computer Science, Information Systems
Shengjin Tang, Xiaodong Xu, Chuanqiang Yu, Xiaoyan Sun, Hongdong Fan, Xiao-Sheng Si
Article
Computer Science, Artificial Intelligence
Weihang Zhang, Xue Wang, Wei You, Junfeng Chen, Peng Dai, Pengbo Zhang
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2020)
