An Optimal Stacking Ensemble for Remaining Useful Life Estimation of Systems Under Multi-Operating Conditions

Abstract

Remaining useful life (RUL) estimation is expected to provide appropriate maintenance for components or systems in industry to improve the reliability of the systems. Most data-based methods are limited to a single model, which is susceptible to various factors like environmental variability and diversity of operating conditions. In this paper, we propose an optimal stacking ensemble method combining different learning algorithms as meta-learners to mitigate the impact of multi-operating conditions. The selection of meta-learners follows a multi-objective evolutionary algorithm named non-dominated sorting genetic algorithms-II to balance the two conflicting objectives in terms of accuracy and diversity. Then the eventually evolved meta-learners are integrated by the meta-classifier for RUL estimation. In addition, a long-short-term feature extraction strategy is proposed to capture more degradation information from lifecycle data dynamically. Extensive experiments are performed on aero-engine dataset and battery dataset provided by NASA, which achieves the higher prognostic accuracy compared with the single models and existing methods. INDEX TERMS Remaining useful life (RUL), reliability, optimal stacking ensemble, meta-learners.

A Prediction Framework for Lifestyle-Related Disease Prediction Using Healthcare Data

“…On the other hand, it is challenging to employ a single model to generate more accurate forecasts and attain higher levels of performance due to the noise from attributes and classes. In machine learning, an ensemble is a sort of model that is built by merging the predictions of various individual models [33]. Typically, ensembles increase performance by reducing the mistakes created by each individual model that contributes to the ensemble.…”

Remaining Useful Lifetime Estimation of Bearings Operating under Time-Varying Conditions

“…Fu et al (2021) and Javanmardi & Hüllermeier (2023) suggest normalizing data according to OC. F. Li et al (2020) integrate several algorithms into one model and select an optimal algorithm set for different OC to minimize their impact. Numerous studies address this problem by employing transfer learning or domain adaptation to handle the distribution shift between the training (source) and testing (target) domains (Mao et al, 2019;Fan et al, 2020;da Costa et al, 2020;Ding, Jia, Miao, & Huang, 2021;Ding, Jia, & Cao, 2021;Zhang et al, 2021).…”

Combining Linear Classifiers Using Probability-Based Potential Functions

“…The network intrusion detection approach used EoC was proposed in [3] whereas paper [4] presents usefulness of EoC in detection cross-site scripting attack for web security. EoC have been also applied in many industrial fields like: the optimal stacking ensemble for remaining useful life estimation was proposed in [5], classification of cutting tools [6] or in the in-line detection of surface defects on glass substrates of thin-film transistor liquid crystal displays [7]. In addition, EoCs are used in other applications such as: the marine sediments classification [8], the land cover type classification [9] or in medical diagnostics [10].…”