Article
Engineering, Industrial
Nick Eleftheroglou, Georgios Galanopoulos, Theodoros Loutas
Summary: This study proposes a new stochastic model, SLHSMM, to address the challenge of reliable RUL prediction in cases with unexpected phenomena. By assigning higher importance to training structures with greater similarity to the testing structure, the estimated parameters effectively capture the specific characteristics of the testing structure.
RELIABILITY ENGINEERING & SYSTEM SAFETY
(2024)
Article
Automation & Control Systems
Milad Rezamand, Mojtaba Kordestani, Marcos E. Orchard, Rupp Carriveau, David S-K Ting, Mehrdad Saif
Summary: A hybrid prognostic method using SCADA and vibration signals is introduced to predict the remaining useful life of wind turbine bearings. Experimental data validation shows higher RUL accuracy compared to the Bayesian algorithm.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2021)
Article
Engineering, Mechanical
Tao Yan, Yaguo Lei, Naipeng Li, Xiaosheng Si, Liliane Pintelon, Reginald Dewil
Summary: This paper proposes a method for ensemble RUL prediction that takes into account the nonlinear relationships among individual prediction models and formulates an online joint replacement-order model using the ensemble RUL prediction results. Experimental results show that the proposed method has higher accuracy and provides more effective joint policies.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2022)
Article
Engineering, Mechanical
Ying Du, Chaoqun Duan, Tonghai Wu
Summary: Lubricating oil plays a crucial role in machine performance throughout its life cycle, and deterioration modeling helps determine when the oil can no longer function properly. Wear debris analysis can partially detect lubricating oil degradation, which can be categorized into different states. The study proposes a method to predict remaining useful life of lubricating oil based on time series modeling and hidden semi-Markov model.
PROCEEDINGS OF THE INSTITUTION OF MECHANICAL ENGINEERS PART J-JOURNAL OF ENGINEERING TRIBOLOGY
(2022)
Article
Automation & Control Systems
Leonardo Ramos Rodrigues, Takashi Yoneyama
Summary: This paper proposed a novel repair priority rule based on a Prognostics and Health Monitoring system, and numerical experiments showed that it consistently reduces inventory system cost.
Article
Computer Science, Information Systems
Muquan Lin, Song Wanqing, Dongdong Chen, Enrico Zio
Summary: In order to prevent issues caused by tool wear, this paper extracts time domain and frequency domain statistical features using a multi-sensory fusion approach and establishes a prediction model for tool wear. Through analyzing the distance correlation coefficient between 140 feature vectors and wear value, seven eigenvectors are selected as inputs for the prediction model, providing the most sensitive features to wear faults.
Article
Engineering, Multidisciplinary
Xieyi Chen, Yi Wang, Haoran Sun, Hulin Ruan, Yi Qin, Baoping Tang
Summary: Gear is crucial for mechanical equipment, and its health directly influences the overall operation of the equipment. Therefore, accurately predicting the remaining useful life (RUL) of gearboxes is of great significance. However, current deep learning-based RUL prediction methods often overlook trend characteristics and focus on the fluctuation patterns of degradation data. To address this issue, a generalized degradation tendency tracking strategy (GDTTS) is proposed to improve the prediction performance by capturing both trend and fluctuation characteristics. Experimental results on real gearbox datasets demonstrate the effectiveness of the proposed strategy.
Article
Computer Science, Information Systems
Qiankun Hu, Yongping Zhao, Yuqiang Wang, Pei Peng, Lihua Ren
Summary: This paper proposes a novel DRL-based prognostic approach for estimating the RUL of engineered systems. The approach formulates the task into an MDP model and employs an advanced DRL algorithm to learn the optimal RUL estimation policy. The effectiveness and superiority of the approach are demonstrated through a case study on turbofan engines in the C-MAPSS dataset.
Article
Computer Science, Artificial Intelligence
Tao Jing, Pai Zheng, Liqiao Xia, Tianyuan Liu
Summary: This article presents an interpretable RUL prediction method based on TF-SCN, HLS-VAE, and a regressor. Experimental results demonstrate that the proposed approach achieves high-quality RUL prediction while providing a visual latent space for evaluating RUL degradation patterns.
ADVANCED ENGINEERING INFORMATICS
(2022)
Article
Computer Science, Information Systems
Gabriel San Martin, Enrique Lopez Droguett
Summary: Deep learning is popular in prognostics and health management, but the challenge of obtaining labeled data remains. This paper introduces a temporal aware Variational Auto-Encoders method with a semi-supervised training scheme to reduce the amount of labeled data required for learning.
Article
Engineering, Multidisciplinary
David A. Najera-Flores, Zhen Hu, Mayank Chadha, Michael D. Todd
Summary: In order to predict the remaining useful life (RUL) of lithium-ion batteries, simplified physical laws and machine learning-based methods can be used to develop a capacity degradation model. While simplified physical models are easy to implement, they may result in large errors in failure prognostics. Data-driven models can provide more accurate degradation forecasting but may require a large amount of training data and may produce predictions inconsistent with physical laws. Existing methods also face challenges in predicting RUL at the early stages of battery life.
APPLIED MATHEMATICAL MODELLING
(2023)
Article
Engineering, Industrial
Luciano Sanchez, Nahuel Costa, Ines Couso
Summary: A method for designing simple models of the remaining lifetime of a system is proposed, which learns a health state model that consistently varies with the remaining useful life. The model and the criterion used to measure its fit are jointly learned, aiming to find the simplest expression within a family of stochastic orderings. This method performs comparably to recent AI-based models but requires fewer parameters, making it applicable in systems with reduced computational capacity.
RELIABILITY ENGINEERING & SYSTEM SAFETY
(2023)
Article
Computer Science, Artificial Intelligence
Xiang Li, Wei Zhang, Hui Ma, Zhong Luo, Xu Li
Summary: This paper proposes a deep learning-based RUL prediction method, which aligns the data of different entities in similar degradation levels through a cycle-consistent learning scheme to improve prediction performance. Experimental results suggest that the method offers a novel perspective on RUL estimations.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Multidisciplinary Sciences
Zhonghai Lu, Chao Guo, Mingrui Liu, Rui Shi
Summary: Estimating the Remaining Useful Lifetime (RUL) of discrete power electronics is crucial for predictive maintenance and system safety. This paper proposes a Physics-Informed Neural Network (PINN) approach to enhance RUL estimation using Recurrent Neural Network (RNN). Results from experiments on the NASA IGBT dataset show that PINN can improve the realism of neural network training and achieve better performance in estimation error and coefficient of determination.
SCIENTIFIC REPORTS
(2023)
Article
Computer Science, Interdisciplinary Applications
Moncef Soualhi, Khanh T. P. Nguyen, Kamal Medjaher, Fatiha Nejjari, Vicenc Puig, Joaquim Blesa, Joseba Quevedo, Francesc Marlasca
Summary: Data-driven prognostics and health management is crucial for the future industry, allowing accurate estimation of system RUL through machine learning algorithms. However, the high variability in end-of-life time due to different fault types and degradation rates results in uncertainties in RUL estimation.
COMPUTERS IN INDUSTRY
(2023)
Article
Statistics & Probability
Zifei Han, Keying Ye, Min Wang
Summary: This article investigates the application of the power prior and its variations in Bayesian inference. The derivation of the marginal likelihood reveals that using conventional initial priors may result in an infinite scaling factor, altering the admissible set of the power parameter. The findings suggest that special attention should be given when the suggested level of borrowing is close to 0, as the actual optimum might be below the suggested value.
AMERICAN STATISTICIAN
(2023)
Article
Statistics & Probability
Chanseok Park, Min Wang
Summary: In this paper, the g and h control charts commonly used for monitoring nonconforming cases are revisited. The correct estimators and biases of these charts are investigated, and a method for constructing charts with unbalanced samples is proposed. Two real-data applications are provided for illustration purposes.
COMMUNICATIONS IN STATISTICS-THEORY AND METHODS
(2023)
Article
Management
Linhan Ouyang, Mei Han, Yizhong Ma, Min Wang, Chanseok Park
Summary: This paper proposes a new metamodeling method for simulation models with non-Gaussian responses. The method uses robust estimators from robust statistics, and through numerical comparisons, it is shown to be efficient.
JOURNAL OF THE OPERATIONAL RESEARCH SOCIETY
(2023)
Article
Computer Science, Interdisciplinary Applications
Shen Zhang, Keying Ye, Min Wang
Summary: In this paper, a simple and easily-implemented Bayesian hypothesis test is proposed for determining the presence of an association between two variables. Through simulation studies and real-data application, the effectiveness of the proposed method is demonstrated, making it suitable for teaching students' Bayesian thinking in data analysis.
JOURNAL OF STATISTICAL COMPUTATION AND SIMULATION
(2023)
Article
Computer Science, Interdisciplinary Applications
Chanseok Park, Min Wang, Linhan Ouyang
Summary: This paper proposes novel robust g and h control charts based on the generalized Kullback-Leibler divergence. By avoiding the pitfalls of existing methods using the asymptotically fully efficient statistical minimum distance method, the proposed method outperforms several existing ones in terms of average run length and relative efficiency, especially when the data contain outliers.
COMPUTERS & INDUSTRIAL ENGINEERING
(2023)
Article
Cardiac & Cardiovascular Systems
Mostafa Rezaeitaleshmahalleh, Kevin W. Sunderland, Zonghan Lyu, Tonie Johnson, Kristin King, David A. Liedl, Janet M. Hofer, Min Wang, Xiaoming Zhang, Wiktoria Kuczmik, Todd E. Rasmussen, Robert D. McBane, Jingfeng Jiang
Summary: Fast-growing abdominal aortic aneurysms (AAA) have a high rupture risk and poor outcomes if not promptly identified and treated. A combination of patient health information, computational hemodynamics, geometric analysis, and artificial intelligence was used to improve the differentiation of small AAAs' growth status. Geometric and hemodynamic analyses were conducted using 3D computed tomography angiography (CTA) data from 70 patients with known growth status. Ten metrics showed statistically significant differences between fast and slow-growing groups. A support vector machine (SVM) classifier achieved an AUROC of 0.86 and total accuracy of 77.50% for classifying AAAs' growth status. This analytics has the potential to guide resource allocation for the management of patients with AAAs.
JOURNAL OF CARDIOVASCULAR TRANSLATIONAL RESEARCH
(2023)
Article
Radiology, Nuclear Medicine & Medical Imaging
N. Mu, M. Rezaeitaleshmahalleh, Z. Lyu, M. Wang, J. Tang, C. M. Strother, J. J. Gemmete, A. S. Pandey, J. Jiang
Summary: Although applying machine learning algorithms to rupture status assessment of intracranial aneurysms has shown promising results, the lack of transparency in some ML methods has hindered their clinical applicability. This study compared six commonly used ML algorithms and demonstrated their consistent performance and explainable predictions. The findings suggest that ML classifiers can contribute to improved understanding and trust in the clinical translation of IA rupture assessment.
BIOMEDICAL PHYSICS & ENGINEERING EXPRESS
(2023)
Article
Statistics & Probability
Fang Chen, Qiuchen Hai, Min Wang
Summary: The classical Hotelling's T-2 test and Bayesian hypothesis tests are not applicable for comparing high-dimensional population means when the model dimension exceeds the sample size. This paper proposes a new Bayesian testing procedure based on a split-and-merge technique. The high-dimensional data is split into lower-dimensional random spaces using subspace clustering, and then the results are merged using the geometric mean to obtain a novel test statistic. Simulation studies and real-data applications demonstrate the performance of the proposed method.
COMPUTATIONAL STATISTICS
(2023)
Article
Computer Science, Interdisciplinary Applications
Chanseok Park, Linhan Ouyang, Min Wang
Summary: The Shewhart-type control chart based on the geometric distribution is affected by the high skewness of the distribution and relies on the normality assumption, resulting in unsatisfactory performance. To correct the skewness, the probability-limit control chart based on the geometric distribution can be used, but it leads to a biased average run length or false alarm rate. This paper suggests using the control chart based on the negative binomial distribution, which not only reduces the bias caused by the discreteness of the geometric distribution, but also outperforms the geometric control chart in terms of smaller bias and higher alarm rate.
COMPUTERS & INDUSTRIAL ENGINEERING
(2023)
Article
Chemistry, Analytical
Kai Yang, Jiajia Liu, Min Wang, Hua Wang, Qingtai Xiao
Summary: A visualization experiment was conducted to study the gas-liquid two-phase flow characteristics in rectangular channels. Different flow patterns were captured and their fluctuation characteristics were analyzed using an electrical conductivity sensor. The results showed that feature vectors constructed from quantized characteristic parameters can successfully reflect the time-frequency characteristics of conductivity measurements. The intelligent two-phase flow-pattern identification method achieved a recognition rate of 93.33% based on support vector machine, providing technical support for online recognition of gas-liquid two-phase flow patterns in rectangular channels.
Article
Statistics & Probability
Zhuanzhuan Ma, Min Wang, Chanseok Park
Summary: In this paper, we propose several alternative estimators for the log-logistic distribution. These estimators not only have closed-form expressions, but also exhibit robustness to data contamination. We assess the performance of each estimator through Monte Carlo simulations and a real-data application, demonstrating their favorable performance compared to the maximum likelihood estimator in the presence of outliers.
JOURNAL OF STATISTICAL THEORY AND PRACTICE
(2023)
Article
Statistics & Probability
Chanseok Park, Xuehong Gao, Min Wang
Summary: This paper consists of two parts. The first part proposes an explicit robust estimation method for linear regression coefficients based on power-weighted repeated medians technique. It investigates the trade-offs between efficiency and robustness and analyzes the bounds of the proposed method's finite-sample breakdown point. The second part shows that the proposed method can be applied to obtain robust estimators for Weibull and Birnbaum-Saunders distributions commonly used in reliability and survival analysis, using linearization of the cumulative distribution function. Numerical studies demonstrate the method's superior performance in the presence of data contamination.
JOURNAL OF APPLIED STATISTICS
(2023)
Article
Engineering, Multidisciplinary
Refah Alotaibi, Ehab M. M. Almetwally, Min Wang, Hoda Rezk
Summary: In this paper, an optimization design for a step-stress accelerated life test (ALT) with two stress variables assuming the lifespan follows the inverse Weibull (IW) distribution is provided. Progressive type-I censoring and accelerated life testing are used to reduce the testing time and cost. A cumulative exposure (CE) model is adopted to examine the impact of varying stress levels, assuming a log-linear relationship between the scale parameter of IW distribution and stress. Maximum likelihood estimators and Bayes estimators of the unknown model parameters are obtained. An optimal test plan is designed through minimizing the asymptotic variance (AV) of the percentile life under normal operating conditions. Simulation studies and real-life data analysis are conducted to illustrate the performance and optimality of the proposed model.
QUALITY AND RELIABILITY ENGINEERING INTERNATIONAL
(2023)
Article
Statistics & Probability
Zhuanzhuan Ma, Zifei Han, Min Wang
Summary: In this paper, a Bayesian hierarchical model with global-local shrinkage priors is proposed for high-dimensional quantile regression analysis. The model includes the horseshoe prior and normal-gamma prior. The proposed methods are evaluated through Monte Carlo simulations and real-data applications, and their performance in terms of parameter estimation and variable selection is demonstrated.
COMMUNICATIONS IN STATISTICS-SIMULATION AND COMPUTATION
(2023)
Article
Thermodynamics
Kai Yang, Yelin Wang, Min Wang, Jianxin Pan, Hua Wang, Qingtai Xiao
Summary: Gas-liquid two-phase flow in rectangular channels is a complex process with varying mixing patterns. In this study, a hybrid model combining time series decomposition and neural network algorithm was developed to identify the steady flow pattern. The model improved the recognition accuracy by 2.54-6.93% compared to existing models, and provided a unified method for calculating industrial heat transfer.
APPLIED THERMAL ENGINEERING
(2024)