A Graph Neural Network Approach to System-Level Health Index
and Remaining Useful Life Estimation
Ark Ifeanyi
1
1
University of Tennessee, Knoxville, TN, 37996, USA
aifeanyi@vols.utk.edu
ABSTRACT
Current methods for predicting health index and remaining
useful life (RUL) in complex systems struggle to account for
performance dependencies between components, leading to
inaccurate system-level estimates. This research proposes a
novel approach utilizing graph neural networks (GNNs) to
improve system-level health index and RUL estimation. GNNs
excel at capturing complex interdependencies within a sys-
tem, making them ideal for this task. The proposed method-
ology is designed for systems with synchronously sampled
process data. To illustrate the application of the proposed
approach, we will use the Condensate Extraction Subsystem
(CES) of a nuclear power plant (NPP) as a case study. Sensor
data like temperature, pressure, and flow rates will be used
to train GNNs to predict the overall health and RUL of the
CES over time. To evaluate the effectiveness of GNNs, a
custom NPP simulator will be used to model the CES un-
der various realistic fault modes across a variety of compo-
nents. The GNN’s performance will be verified and its ro-
bustness will be tested under diverse scenarios. This research
aims to demonstrate the effectiveness and resilience of GNNs
for system-level prognostics. By providing valuable insights
for maintenance decision-making, this approach can enhance
operational reliability and safety in complex engineering sys-
tems. The proposed framework has the potential to be applied
across various industries, leading to advancements in predic-
tive maintenance practices.
1. PROBLEM STATEMENT
In recent years, significant attention has been drawn towards
prognostics and health management (PHM) techniques for
predicting the remaining useful life (RUL) of complex sys-
tems. This paper aims to address the limitations of exist-
ing methods in integrating RUL information from individual
components to derive accurate system-level RUL estimations.
Ark Ifeanyi 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.
Current methodologies for RUL estimation often fall short
when it comes to capturing the holistic health state of a sys-
tem comprised of interconnected components. Some tech-
niques employ direct mapping of system inputs to outputs to
estimate system health or RUL, neglecting the nuanced degra-
dation dynamics within individual components (Behera et al.,
2021). Conversely, other methods focus on component-level
prognostics, predicting RUL for each component and aggre-
gating these predictions using tools like fault trees (Gomes et
al., 2013). However, this second approach becomes compu-
tationally intensive for larger systems due to the necessity of
establishing prognostic models for every individual compo-
nent.
The complexity of system-level prognostics is further com-
pounded by the intricate interdependencies between compo-
nents, where the degradation of one component can influence
and be influenced by others. These dependencies can lead
to unique degradation patterns within the system, requiring
a more sophisticated approach that accounts for these inter-
actions (Kim et al., 2021). Moreover, the necessity of con-
ducting system-level prognostics under various fault modes
remains a critical challenge (Kim et al., 2021). Different
fault modes can induce distinct degradation behaviors within
the system, necessitating the development of predictive mod-
els capable of adapting to and accurately forecasting RUL
across these diverse scenarios. Therefore, there is a clear im-
perative to develop novel methodologies that can effectively
integrate RUL information from individual components, ac-
count for system-level interdependencies, and accommodate
diverse fault modes to enhance the accuracy and applicability
of system-level RUL estimation techniques.
This paper seeks to explore these challenges and propose a
Graph Neural Network (GNN) based framework tailored to-
wards system-level RUL estimation, leveraging the inherent
relationships and dependencies between components to achieve
more robust and accurate prognostic outcomes. By address-
ing these limitations, the research aims to contribute towards
advancing the field of PHM and enabling more reliable main-
tenance strategies for complex engineering systems.
1
