Article
Engineering, Mechanical
Zuo Zhu, Siu-Kui Au, Binbin Li, Yan-Long Xie
Summary: Operational modal analysis (OMA) is increasingly used to identify modal properties of constructed structures economically. A Bayesian approach is proposed in this paper to compute the most probable value (MPV) of modal parameters from multiple setups and modes. The study shows the proposed Bayesian approach consistently yields reasonable results when data quality is low in some setups, outperforming existing algorithms.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2021)
Article
Automation & Control Systems
Peng Ding, Minping Jia, Yifei Ding, Yudong Cao, Jichao Zhuang, Xiaoli Zhao
Summary: Few-shot learning based machinery prognostics are feasible for intelligent operation and maintenance with scarce monitoring data. To improve the reliability of predictive maintenance, a novel Bayesian approximation enhanced probabilistic meta-learning (BA-PML) algorithm is proposed to estimate uncertainty in few-shot prognostics. This algorithm consists of a designed base probabilistic predictor and an episodic training strategy.
IEEE-ASME TRANSACTIONS ON MECHATRONICS
(2023)
Article
Automation & Control Systems
Jeremias Knoblauch, Jack Jewson, Theodoros Damoulas
Summary: The paper advocates for an optimization-centric view of Bayesian inference, introducing the Rule of Three (ROT) as a generalized method for Bayesian posteriors. It also explores the applications of Generalized Variational Inference (GVI) posteriors and their potential to improve robustness and posterior marginals in Bayesian Neural Networks and Deep Gaussian Processes.
JOURNAL OF MACHINE LEARNING RESEARCH
(2022)
Article
Computer Science, Interdisciplinary Applications
Douglas J. Brinkerhoff
Summary: In this study, we use stochastic variational inference to characterize the joint Bayesian posterior distribution over spatially-varying basal traction and ice rheology of an ice sheet model. By controlling the prior assumptions on parameter smoothness and length scale and casting the problem in terms of eigenfunctions of a kernel, we achieve tractable inference. The method recovers known parameters, accounts for parameter indeterminacy, and shows computational scalability for catchment-sized problems.
JOURNAL OF COMPUTATIONAL PHYSICS
(2022)
Article
Engineering, Civil
Pinghe Ni, Qiang Han, Xiuli Du, Xiaowei Cheng, Hongyuan Zhou
Summary: This paper presents a data-driven approach for post-earthquake reliability assessments of civil structures. It updates the probability density functions of random variables using measured vibration data, and generates the posterior probability density functions of structural parameters using two approximate Bayesian computation techniques. The updated probability density functions are then used for reliability assessments, and numerical studies verify the accuracy and efficiency of the proposed techniques.
ENGINEERING STRUCTURES
(2022)
Article
Automation & Control Systems
Xiaolong Chen, Yi Chai, Qie Liu, Pengfei Huang, Linchuan Fan
Summary: In this paper, a novel Bayesian sparse multiple kernel-based identification method (BSMKM) for multiple-input single-output (MISO) Hammerstein system is proposed. The method represents the nonlinear part and the linear part using basis-function model and finite impulse response model respectively and estimates all unknown model parameters through hierarchical prior distribution and full Bayesian method based on variational Bayesian inference.
Article
Engineering, Multidisciplinary
Vahid Keshavarzzadeh, Robert M. Kirby, Akil Narayan
Summary: Inverse problems are common in engineering simulations, and Bayesian inference is a predominant approach to infer unknown parameters. This paper presents a variational inference method that incorporates observation data and the gradient information of the forward map to invert unknown latent parameters. The method utilizes a trained neural network to generate samples for statistical calculations. The effectiveness of the method is demonstrated through examples, and future research directions are discussed.
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING
(2022)
Article
Evolutionary Biology
Mathieu Fourment, Christiaan J. Swanepoel, Jared G. Galloway, Xiang Ji, Karthik Gangavarapu, Marc A. Suchard, Frederick A. Matsen
Summary: Gradients of probabilistic model likelihoods are crucial for computational statistics and machine learning. General-purpose machine-learning libraries like TensorFlow and PyTorch offer automatic differentiation for arbitrary models. However, for phylogenetic cases, these libraries may be slower compared to specialized code. This paper compares six gradient implementations and finds that automatic differentiation is slower than carefully implemented methods. A mixed approach combining phylogenetic libraries and machine learning libraries is recommended for optimal speed and model flexibility.
GENOME BIOLOGY AND EVOLUTION
(2023)
Article
Economics
Filipe Rodrigues
Summary: This study proposes an amortized variational inference approach that utilizes stochastic backpropagation, automatic differentiation, and GPU-accelerated computation to enable Bayesian inference in mixed multinomial logit models on large datasets. Furthermore, it demonstrates how normalizing flows can enhance the flexibility of variational posterior approximations. Simulation and real data analysis show that this approach achieves significant computational speedups without compromising estimation accuracy on large datasets.
TRANSPORTATION RESEARCH PART B-METHODOLOGICAL
(2022)
Article
Engineering, Mechanical
Felipe Igea, Alice Cicirello
Summary: Multi-modal distributions of physics-based model parameters are common in engineering, but traditional sampling techniques struggle to accurately capture them with limited simulations. This can lead to numerical errors in assessing structures under uncertainty. To overcome this, a cyclical annealing schedule is proposed for the Variational Bayes Monte Carlo (VBMC) method, improving exploration and finding high probability areas in multi-modal distributions. Comparisons with other algorithms show that the proposed cyclical VBMC outperforms in terms of accuracy and required model runs.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2023)
Article
Engineering, Geological
Wenmin Yao, Changdong Li, Changbin Yan, Hongbin Zhan
Summary: The study proposes a hybrid framework for slope reliability based on Bayesian sequential updating technology, integrating prior knowledge, multiple estimation methods, and model uncertainties to estimate slope reliability with limited geotechnical data. Through experiments with three slope examples, the framework is shown to provide reliable and accurate estimations of slope reliability.
Article
Engineering, Mechanical
Wyatt Bridgman, Reese E. Jones, Mohammad Khalil
Summary: This paper proposes a method for improving variational inference by constructing an initial Gaussian mixture model approximation, and it is demonstrated through synthetic tests and inversion problems in structural dynamics.
PROBABILISTIC ENGINEERING MECHANICS
(2023)
Article
Physics, Multidisciplinary
Semih Akbayrak, Ivan Bocharov, Bert de Vries
Summary: VMP provides an automated and efficient algorithmic framework for approximating Bayesian inference. Executing VMP in complex models relies on the ability to compute the expectations of hidden variable statistics. The Extended VMP approach extends the applicability of VMP through importance sampling and Laplace approximation.
Article
Computer Science, Artificial Intelligence
Bjarne Grimstad, Mathilde Hotvedt, Anders T. Sandnes, Odd Kolbjornsen, Lars S. Imsland
Summary: Recent works have shown promising results in using machine learning for modeling flow rates in oil and gas wells. This paper introduces a probabilistic virtual flow meter based on Bayesian neural networks, which helps to describe uncertainty in the model and measurements using variational inference. The research findings suggest the need for alternative strategies to enhance the robustness of data-driven virtual flow meters.
APPLIED SOFT COMPUTING
(2021)
Article
Engineering, Electrical & Electronic
Wonkeun Youn, Nak Yong Ko, Stephen Gadsden, Hyun Myung
Summary: This article proposes a novel adaptive Kalman filter algorithm, which improves filtering performance by estimating the unknown measurement loss probability.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2021)
Article
Engineering, Industrial
Xiaochen Xian, Chen Zhang, Scott Bonk, Kaibo Liu
Summary: This paper proposes a rank-based monitoring and sampling algorithm based on data augmentation to quickly detect mean shifts in a process when only a limited portion of observations are available online. The method automatically augments information for unobservable variables based on online observations and intelligently allocates monitoring resources to the most suspicious data streams. By accurately inferring the status of all variables based on a small number of observable variables and constructing a global monitoring statistic, the proposed method enables quick detection of out-of-control status even with limited shifted variables observed in real time.
JOURNAL OF QUALITY TECHNOLOGY
(2021)
Article
Statistics & Probability
Xiaochen Xian, Honghan Ye, Xin Wang, Kaibo Liu
Summary: The article proposes a multivariate Poisson log-normal model tailored to traffic demand prediction, which automatically clusters routes based on demand correlations to enhance prediction accuracy. Markov chain Monte Carlo sampling is used to overcome computational challenges and facilitate model estimation and prediction.
Article
Engineering, Industrial
Di Wang, Kaibo Liu, Xi Zhang
Summary: Thermal fields play a critical role in engineering systems and industries, and an accurate prediction of thermal field distribution is essential for system surveillance, maintenance, and improvement. In this study, we propose a spatiotemporal prediction approach based on transfer learning techniques to address the challenges of data sparsity and missing data in thermal field prediction using sensor networks. Our approach considers the dynamics of 3D thermal field and utilizes information from homogeneous fields to achieve accurate thermal field distribution. A case study on grain storage validates the effectiveness of our proposed approach.
JOURNAL OF QUALITY TECHNOLOGY
(2022)
Article
Engineering, Industrial
Minhee Kim, Jing-Ru C. Cheng, Kaibo Liu
Summary: A novel sensor selection framework is proposed in this study to enhance remaining useful life prediction by adaptively deciding which sensors to use. The framework is generic, trained in a unified manner, improves interpretability, and introduces regularization techniques for stability in the training process.
JOURNAL OF QUALITY TECHNOLOGY
(2021)
Article
Automation & Control Systems
Honghan Ye, Kaibo Liu
Summary: This article proposes a generic online nonparametric monitoring and sampling scheme for quickly detecting mean shifts in high-dimensional heterogeneous processes. By integrating the Thompson sampling algorithm with a quantile-based nonparametric cumulative sum procedure, local statistics of all data streams are constructed based on partially observed data. A global monitoring scheme is developed to detect a wide range of possible mean shifts. The proposed method balances between exploration and exploitation, and has shown superiority in performance through simulations and a case study.
IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING
(2022)
Article
Automation & Control Systems
Minhee Kim, Changyue Song, Kaibo Liu
Summary: This study proposes a generic framework to handle the heterogeneity across units by effectively leveraging intrinsic covariate information, leading to better degradation modeling and prognostics. Through simulation studies and a case study on Alzheimer's disease data set, the proposed method demonstrates its effectiveness and advantages over existing benchmark approaches.
IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING
(2022)
Article
Automation & Control Systems
Di Wang, Kaibo Liu, Xi Zhang
Summary: This article proposes a novel indirect deep learning method to construct a health index (HI) by combining multiple sensor signals for better characterizing the degradation process, seamlessly integrating a deep neural network (DNN) and a long short term memory (LSTM) model. Domain knowledge is considered to enhance interpretability, and an indirect gradient descent (IGD) algorithm is developed for parameter estimation. Simulation studies and a case study on aircraft gas turbine engine degradation are presented to validate the method's performance.
IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING
(2022)
Article
Automation & Control Systems
Changyue Song, Ziqian Zheng, Kaibo Liu
Summary: Sensors are commonly used to monitor the degradation process of engineering systems. Existing studies have limitations, prompting the proposal of a systematic method for degradation modeling and prognosis that is versatile and easily understandable, regardless of sensor spacing or asynchronous signals.
IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING
(2022)
Article
Nuclear Science & Technology
Elisa Ou, Minhee Kim, Po-Ling Loh, Todd Allen, Robert Agasie, Kaibo Liu
Summary: With the increasing number of data-driven models in nuclear applications, this study proposes an automatic pipeline for extracting information from handwritten tabular documents collected from nuclear power plants. The results demonstrate the high accuracy and practicality of the proposed method.
NUCLEAR ENGINEERING AND DESIGN
(2022)
Article
Engineering, Industrial
Honghan Ye, Ziqian Zheng, Jing-Ru C. Cheng, Brock Hable, Kaibo Liu
Summary: Recent advancements in sensor technology have enabled the monitoring of high-dimensional data streams in manufacturing systems for quality improvement. However, current monitoring schemes assume uniform sampling intervals for all data streams, which is not always the case in practice. This paper proposes a nonparametric framework to monitor asynchronous and heterogeneous data streams with different sampling intervals and arbitrary distributions. The proposed method includes a quantile-based approach for local monitoring and a compensation strategy for unsampled measurements, as well as a global monitoring scheme using top -r local statistics.
INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH
(2023)
Article
Automation & Control Systems
Di Wang, Kaibo Liu
Summary: Accurate prognostics are essential for preventing unexpected failures in industrial and service systems. This paper proposes an integrated deep learning-based method that combines a DNN fusion model and an LSTM degradation model to characterize the relationship between health index (HI) and multiple sensor signals, as well as the underlying degradation status of units. The proposed method achieves promising performance in a case study on aircraft gas turbine engines.
IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING
(2023)
Article
Engineering, Industrial
Honghan Ye, Xiaochen Xian, Jing-Ru C. Cheng, Brock Hable, Robert W. Shannon, Mojtaba Kadkhodaie Elyaderani, Kaibo Liu
Summary: This article proposes a nonparametric monitoring and sampling algorithm for quickly detecting abnormalities in heterogeneous data streams. The algorithm collectively estimates the underlying status of all data streams based on partially observed data and uses an intelligent sampling strategy to dynamically observe informative data streams for quick anomaly detection.
Article
Engineering, Industrial
Di Wang, Fangyu Li, Kaibo Liu
Summary: This article presents a multivariate spatio-temporal modeling and monitoring method for network systems using multiple types of sensor signals. It aims to ensure information security and support system automation.
Article
Automation & Control Systems
Ziqian Zheng, Wei Zhao, Brock Hable, Yutao Gong, Xuan Wang, Robert W. Shannon, Kaibo Liu
Summary: This paper proposes a transfer learning-based independent component analysis (ICA) method to address the issue of degraded component extraction accuracy with limited available data. By transferring component distribution from a source domain, accurate component extraction results can be achieved in the target domain. Numerical simulations and a case study demonstrate the effectiveness of the proposed method in transferring knowledge and reducing negative transfer.
IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING
(2022)
Article
Automation & Control Systems
Honghan Ye, Xi Wang, Kaibo Liu
Summary: This article addresses the critical problem of balancing production scheduling and maintenance in a flow shop production line. By dynamically updating the PM interval to coordinate maintenance activities and job scheduling, overall cost savings are achieved.
IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING
(2021)
Article
Engineering, Industrial
Xubo Yue, Raed Al Kontar, Ana Maria Estrada Gomez
Summary: This article presents a federated data analytics approach for linear regression models, utilizing hierarchical modeling and information sharing to handle data distributed across different devices. It provides uncertainty quantification, variable selection, hypothesis testing, and fast adaptation to new data.