Proactive Aircraft Engine Removal Planning with Dynamic
Bayesian Networks
Bharath Pidaparthi
†
, Ryan Jacobs, Sayan Ghosh, Sandipp Krishnan Ravi, Ahmad W Amer, Lele Luan, R Murali Krishnan,
Feng Zhang, Victor Perez, Liping Wang
General Electric Aerospace Research
1 Research Circle, Niskayuna, New York, 12309, USA
ABSTRACT
Aircraft engine removal for maintenance is an expensive or-
deal, and planning for it while balancing fleet stability ob-
jectives is a complex multi-faceted challenge. This is further
compounded by uncertainties associated with usage or just-
in-time maintenance approaches that are becoming prevalent.
Engine removal decisions rely on accurate estimation of dam-
age growth or remaining useful life of critical components
and a framework for aggregating these component-level esti-
mates (and their uncertainties) into an engine-level removal
forecasting model. An approach to this planning challenge
is to develop probabilistic prognostic digital twins tailored
to engine-specific operations and calibrate/update them with
inspection data from the field. To this end, this work out-
lines a framework involving: 1) building component-level
probabilistic models capable of forecasting damage growth
or remaining useful life, 2) aggregating the outputs of these
component-level models into a system-level view using a Dy-
namic Bayesian Network (DBN), and 3) updating the states
of the DBN with inspection information as and when they
become available.
Keywords: Prognostics, Probabilistic Digital Twins, Dy-
namic Bayesian Networks, Cumulative Damage Modeling
1. INTRODUC TION:
Aircraft engines are highly sophisticated systems, compris-
ing a multitude of interdependent components, each vulnera-
ble to various damage and failure modes. This inherent com-
plexity makes accurate prognostics challenging. To mitigate
these risks and ensure the highest standards of operational
safety, regulatory agencies impose stringent requirements for
periodic inspections and maintenance. These regulatory stan-
†
Lead & corresponding Author: bharath.pidaparthi@ge.com
Bharath Pidaparthi 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, pro-
vided the original author and source are credited.
dards encompass a wide spectrum of checks and procedures,
from routine Bore-scope Inspections (BSI) to comprehensive
engine overhauls. Often, these procedures necessitate the
removal of engines from the aircraft, leading to significant
downtime and potential revenue loss.
The financial and logistical impacts of unexpected mainte-
nance events can be especially severe. These events not only
result in significant costs but also create major challenges in
re-booking passengers and managing schedules, particularly
when spare engines or aircraft are not immediately available.
Additionally, unplanned maintenance is frequently hindered
by long lead times for replacement parts and limited avail-
ability of certified Maintenance, Repair, and Overhaul (MRO)
shop slots. These multifaceted challenges highlight the criti-
cal importance of efficient engine removal planning, which is
essential to strike a balance between maintaining operational
safety and ensuring engine availability.
Fleet-level engine removal statistics/models, based on histori-
cal data, are inadequate for making engine-specific decisions.
These models, by nature of their construction, fail to capture
differences in damage states from one engine to another – at-
tributed to individual engine-level variations in manufactur-
ing, material properties, operational profiles, and load fac-
tors. Building a prognostic digital twin, therefore, custom-
tuned/calibrated for each individual engine is essential for ef-
fective removal decision-making. These digital twins, as out-
lined in (Li, Mahadevan, Ling, Wang, & Choze, 2017; Li,
Mahadevan, Ling, Choze, & Wang, 2017), must be formu-
lated to accomplish the following:
• Incorporate various sources of uncertainty from hard-
ware manufacturing variations to loading/operational
differences.
• Integrate heterogeneous information – including oper-
ational data, laboratory data, physics-based/empirical
models, expert opinions, and more.
• Capable of updating the uncertainty in model parameters
(to reduce discrepancies between the digital twin and the
1
