PEER REVIEW
20
th
Australian International Aerospace Congress, 27 February to 1 March 2023, Melbourne
ISBN number: 978-1-925627-66-4
Change Detection for Improved Maintenance Notification
and Remaining Useful Life Calculation
Eric Bechhoefer
1
GPMS International Inc, 93 Pilgrim Park Rd, Waterbury, VT, 05676, USA
Abstract
The Defence Science and Technology Group (DTSG) Data Challenge provides an opportunity
to demonstrate essential HUMS (Health and Usage Monitoring Systems) technologies. In
general, HUMS aims to improve the safety and reliability of critical helicopter components.
However, widespread technology adoption requires HUMS to provide a return on investment
(ROI) by adding functionality and actionable maintenance information to improve asset
management. Providing an estimate of remaining useful life (RUL) enhances safety and allows
better control of assets, providing greater opportunities for revenue and improving ROI.
RUL estimation is an end-to-end process of feature extraction, threshold setting, extrapolation
of damage propagation, and validation. Feature extraction is the generation of condition
indicators (CI) that are representative of damage. Threshold setting triggers an alert as to when
it is appropriate to do maintenance. In this application, it defines the health indicator (HI). A
high cycle fatigue model can then be used to estimate the remaining cycles from the current HI
to the HI at which an alert is generated. As important as the estimate in the RUL is some
indicator of the confidence in the RUL. A high confidence RUL validates that the asset should
be removed from service and maintenance performed.
This paper describes the CIs used, the HI function, and the development of a configuration
process to support threshold setting. A subcritical crack growth model, based on a power law,
uses the HI time series to estimate the RUL and both the first and second derivatives of the
RUL. It is hypothesized that a well-modeled RUL estimation will have a first derivate of - 1
(that is, for each hour of usage, the RUL decrements by -1), and the second derivative of near
0 (the model is stable)..
Keywords: Condition indicator, Fatigue failure, Health indicator, RUL, Threshold setting.
Introduction
Helicopter drive systems play a signification role in the safety of rotorcraft. Vibration
monitoring systems have been developed to improve the safety and maintainability of these
drive systems. Because of their compact size and efficiency, the planetary transmission is used
in many rotorcrafts in the final stage of the drive. Planetary drive systems have posed a
challenge for vibration-based gear fault detection because the load is shared by multiple planets
operating synchronously, such that the other health gears may mask a gear fault.
Helicopter planetary gearbox faults are rare. Further, there are even fewer publicly available
data sets to develop new or test existing algorithms against. From an end-to-end perspective,
vibration-based component fault detection is part of a health and usage monitoring system
(HUMS) chain of events needed to trigger a maintenance event responsive to a potential fault.
As with any safety system, HUMS provides a balance between production/operations and
