Bearing-grouped, stage-resolved uncertainty assessment for remaining useful life prediction of rolling bearings

Abstract Accurate prediction of the remaining useful life of rolling bearings is important for reducing downtime costs and ensuring system safety. However, existing data-driven methods still face severe challenges, such as large sensor noise interference, non-monotonic degradation indicators, and the lack of reliable confidence intervals for point prediction results. This paper proposes a bearing RUL uncertainty quantification prediction framework that integrates robust health indicators construction, bidirectional long short-term memory network and inductive conformal prediction. First, the method smooths and accumulates the vibration signals to construct a strictly monotonic health indicator that conforms to the irreversible physical characteristics to suppress random fluctuations. Then, it uses the BiLSTM network to capture bidirectional temporal evolution features and achieve high-precision mapping from degradation features to RUL. Finally, it introduces the conformal prediction theory, which quantifies the prediction risk without assuming the error distribution, and uses an independent calibration set to construct a prediction interval with strict statistical coverage guarantee for the predicted values. The experimental results on the PHM IEEE 2012 and XJTU-SY datasets show that in terms of the root mean square error and mean absolute error metrics, the proposed method significantly outperforms traditional models.Furthermore, the coverage probability of the prediction interval remains above 95% to ensure high prediction accuracy and reliable uncertainty quantification capability. The study shows that it can provide a theoretically sound and technically supported basis for state-based maintenance decisions in complex industrial scenarios with high credibility, by combining physical monotonicity constraint and statistical calibration.

Authors

Institutions

Publication Details

Journal
Measurement Science and Technology
Published
2026-11-16
DOI
https://doi.org/10.1088/1361-6501/aec284
Primary Topic
Machine Fault Diagnosis Techniques
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Bearing-grouped, stage-resolved uncertainty assessment for remaining useful life prediction of rolling bearings

Min Huang, Huanwei Xu, Wencai Cheng, Wenjing Yu et al.
5 links
Measurement Science and Technology
Machine Fault Diagnosis Techniques
article

Bearing-grouped, stage-resolved uncertainty assessment for remaining useful life prediction of rolling bearings

Min Huang, Huanwei Xu, Wencai Cheng, Wenjing Yu, Yaoyang Liu
article en

Abstract

Abstract Accurate prediction of the remaining useful life of rolling bearings is important for reducing downtime costs and ensuring system safety. However, existing data-driven methods still face severe challenges, such as large sensor noise interference, non-monotonic degradation indicators, and the lack of reliable confidence intervals for point prediction results. This paper proposes a bearing RUL uncertainty quantification prediction framework that integrates robust health indicators construction, bidirectional long short-term memory network and inductive conformal prediction. First, the method smooths and accumulates the vibration signals to construct a strictly monotonic health indicator that conforms to the irreversible physical characteristics to suppress random fluctuations. Then, it uses the BiLSTM network to capture bidirectional temporal evolution features and achieve high-precision mapping from degradation features to RUL. Finally, it introduces the conformal prediction theory, which quantifies the prediction risk without assuming the error distribution, and uses an independent calibration set to construct a prediction interval with strict statistical coverage guarantee for the predicted values. The experimental results on the PHM IEEE 2012 and XJTU-SY datasets show that in terms of the root mean square error and mean absolute error metrics, the proposed method significantly outperforms traditional models.Furthermore, the coverage probability of the prediction interval remains above 95% to ensure high prediction accuracy and reliable uncertainty quantification capability. The study shows that it can provide a theoretically sound and technically supported basis for state-based maintenance decisions in complex industrial scenarios with high credibility, by combining physical monotonicity constraint and statistical calibration.

Measurement Science and Technology
University of Electronic Science and Technology of China (CN)
National Natural Science Foundation of China, Natural Science Foundation of Sichuan Province
Industry, innovation and infrastructure
Machine Fault Diagnosis Techniques
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon