1
Robust Health Condition Prediction of Helicopter Turboshaft
Engines Using Ensemble Machine Learning Models
Zihan Wu, Junzhe Wang, and Meng Li
NOV, Houston, Texas, 77042, United States of America
zihan.wu@nov.com
eric.wang@nov.com
meng.li@nov.com
ABSTRACT
This paper presents a novel ensemble approach that combines
multiple machine-learning algorithms to deliver robust
predictions of helicopter turboshaft engine health status
(nominal or faulty) using operational data. Engine health is
evaluated through the torque margin, defined as the
percentage difference between the measured and target
torque values. A Gaussian process model is used to estimate
the torque margin as a probability distribution function
(PDF), and these predictions are incorporated as features into
various machine-learning models. These models are then
employed to perform binary classification, determining the
engine's health state. To enhance performance, a reference set
is defined for each unseen data point, allowing a comparison
of the relative performances of the models, with the best
performer selected for the final prediction. Our ensemble
method achieves high accuracy in health classification while
providing precise torque margin estimates. The results
demonstrate that ensemble models offer superior
generalization and reliability compared to individual
machine-learning algorithms, especially when applied to
complex, multivariate datasets like those from helicopter
turboshaft engines.
1. INTRODUCTION
Helicopter turboshaft engines are complex mechanical
systems whose health is critical to the safety and performance
of aviation operations. Effective monitoring and prediction of
engine health can prevent costly failures and ensure
operational readiness (Elasha et al., 2021). Traditional
maintenance practices rely heavily on scheduled inspections,
which can lead to unnecessary downtime or missed detection
of early-stage faults (Achouch et al., 2022, Wu et al., 2023).
To address these limitations, there is a growing interest in
data-driven approaches (Daouayry et al., 2018,) that leverage
operational data to assess engine health in real time and
enable predictive maintenance strategies. A comprehensive
review of data-driven prognostic methods was provided in
(Schwabacher et al., 2005). These approaches typically
involve the fusion of sensor data, feature extraction, and
statistical pattern recognition. For predicting the health
condition, techniques such as interpolation (Wang et al.,
2008), extrapolation (Coble et al., 2008), or machine learning
(Wu et al., 2022) are often employed, among others.
Machine learning (ML) techniques have shown great
potential in condition monitoring (Surucu et al., 2023) and
fault detection (Nelson et al., 2023, Wang et al., 2024, Zheng
et al., 2024) across various industries. In particular, ensemble
methods—where multiple machine learning models are
combined to improve prediction accuracy and robustness—
have proven effective in handling complex, multivariate
datasets. By leveraging diverse models, ensemble methods
can capture different patterns and relationships within the
data, leading to improved generalization on unseen assets
(Mian et al., 2024).
In this work, we propose an ensemble machine-learning
framework for predicting the health of helicopter turboshaft
engines using operational measurements such as outside air
temperature, compressor speed, and torque margin. Our
approach focuses on two key tasks: (1) binary classification
of engine health state (nominal or faulty) and (2) probabilistic
regression to estimate the torque margin. This dual-task
framework provides not only a fault diagnosis but also a
confidence metric for the torque margin, enhancing the
interpretability of the predictions.
The contributions of this paper are twofold: first, we
introduce an ensemble learning model for engine health
classification and torque margin estimation; second, we
validate our approach on a dataset of seven engines,
demonstrating its generalization capability across unseen
assets. Our results show that the proposed ensemble method
outperforms individual machine learning models, offering a
Zihan Wu et al. This is an open-access article distributed under the terms
of the Creative Commons Attribution 3.0 United States License, which
permits unrestricted use, distribution, and reproduction in any medium,
provided the original author and source are credited.
