Recursive Stage Discovery for Early Warning in Multi-Sensor Degradation Monitoring

Turbofan engines are monitored continuously by onboard sensors, yet the short interval before failure accounts for only a small fraction of the recorded data. We propose an ordered stage index for early warning and for describing an engine’s degradation state in sensor feature space. The method starts with a K-means partition, refines its nodes recursively using remaining useful life (RUL) labels from the training engines, and ranks its terminal nodes by their median training RUL. We evaluated the index on four subsets of the NASA C-MAPSS turbofan simulation benchmark using five engine-level folds, with preprocessing and model fitting performed on training engines only. On held-out run-to-failure trajectories, Spearman correlations between stage rank and RUL at first arrival ranged from −0.86 to −0.81. The index issued a warning for every held-out run-to-failure engine at operating thresholds of 50 cycles and above. A binned random forest produced stronger rank correlations. Each stage in the proposed index identifies a region of sensor feature space whose split path and empirical training RUL distribution can be reported.

Authors

Institutions

Publication Details

Journal
Sensors
Published
2026-11-19
DOI
https://doi.org/10.3390/s26227305
Primary Topic
Machine Fault Diagnosis Techniques
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Recursive Stage Discovery for Early Warning in Multi-Sensor Degradation Monitoring

Imtiaz Ahmed, Hamdy Soliman
Sensors
Machine Fault Diagnosis Techniques
article

Recursive Stage Discovery for Early Warning in Multi-Sensor Degradation Monitoring

Imtiaz Ahmed, Hamdy Soliman
article en

Abstract

Turbofan engines are monitored continuously by onboard sensors, yet the short interval before failure accounts for only a small fraction of the recorded data. We propose an ordered stage index for early warning and for describing an engine’s degradation state in sensor feature space. The method starts with a K-means partition, refines its nodes recursively using remaining useful life (RUL) labels from the training engines, and ranks its terminal nodes by their median training RUL. We evaluated the index on four subsets of the NASA C-MAPSS turbofan simulation benchmark using five engine-level folds, with preprocessing and model fitting performed on training engines only. On held-out run-to-failure trajectories, Spearman correlations between stage rank and RUL at first arrival ranged from −0.86 to −0.81. The index issued a warning for every held-out run-to-failure engine at operating thresholds of 50 cycles and above. A binned random forest produced stronger rank correlations. Each stage in the proposed index identifies a region of sensor feature space whose split path and empirical training RUL distribution can be reported.

SensorsVol. 26(22)
New Mexico Institute of Mining and Technology (US)
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