Adaptive Two-Stage Model for Bearing Remaining Useful Life Prediction Using Gaussian Process Regression With Matched Kernels

Created by W.Langdon from gp-bibliography.bib Revision:1.8187

@Article{Zheng:Reliability,

• author = "Xinyu Zheng and Wei Fan and Chao Chen and Zhike Peng",
• journal = "IEEE Transactions on Reliability",
• title = "Adaptive Two-Stage Model for Bearing Remaining Useful Life Prediction Using Gaussian Process Regression With Matched Kernels",
• year = "2024",
• volume = "73",
• number = "4",
• pages = "1958--1966",
• month = dec,
• keywords = "genetic algorithms, genetic programming, Degradation, Adaptation models, Kernel, Predictive models, Hidden Markov models, Feature extraction, Computational modelling, Adaptive two-stage degradation, change point, Gaussian process, remaining useful life (RUL)",
• ISSN = "1558-1721",
• DOI = "doi:10.1109/TR.2024.3359212",
• abstract = "Stemming from complex mechanisms and working conditions of bearings, single degradation models often fail to adequately describe complex degradation process and provide reliable prediction of remaining useful life (RUL). To address this challenge, an adaptive two-stage degradation framework based on Gaussian process regression (GPR) is proposed. This framework dynamically selects the appropriate degradation model based on the observed characteristics of the actual degradation data, resulting in improved prediction accuracy and adaptability. In the process of constructing the degradation model, the suitable detection method is adaptively determined based on change point locations, facilitating real-time monitoring of degradation pattern shifts. Within the two-stage GPR model, matched kernel functions are chosen based on degradation rate and trend changes, and a degradation indicator is constructed using the genetic programming algorithm. This indicator serves as target set for the GPR model to accurately estimate RUL. The reliability of the proposed method is validated through comparison with other alternative models on prognostics and health management (PHM) challenge bearing datasets, confirming its effectiveness, robustness, and superiority.",
• notes = "Also known as \cite{10424014}",

}

Genetic Programming entries for Xinyu Zheng Wei Fan Chao Chen Zhike Peng

Citations