Overview
Aircraft availability, in-flight reliability and low-cost maintenance are major concerns for helicopter operators. HUMS (Health Usage Monitoring System) implementing sensor-based monitoring is an enabling technology that seeks to provide a condition-based maintenance (CBM) relying on automated diagnosis/prognosis of the health of aircraft components. One challenge for HUMS was to implement automated low-cost CBM systems as an alternative to periodic physical inspections. Existing HUMS technologies tend to generate high rates of false alarms due to the use of fixed alarm thresholds. The automated analysis of fleet operating data on engine and rotating parts recorded by onboard sensors is a major scientific objective to reach adaptive, reliable, and low-cost HUMS systems. This objective was explored in this project by addressing:
- Performance of simultaneous oil debris monitoring (ODM) and vibration monitoring using available ODM and vibration sensors;
- Analysis of new physical models for ODM and vibration characteristics of helicopter rotating parts (gearboxes, bearings, etc.) to calibrate ‘ageing effects’ and ‘progressive emergence of failures’;
- Design and validation of innovative software tools dedicated to self-adaptive diagnosis/ prognosis of potential failures of helicopter rotating parts.
The project’s main goal was to enable ‘fleet-scale’ health monitoring for helicopters with robust failure diagnosis and prognosis, relying on multi-sensor monitoring and automated analysis of sensor-recorded data. This will reduce false alarm rates and maintenance costs and increase operational aircraft availability, enabling efficient scheduling of preventive maintenance.
The main scientific and technological objectives of this project were:
- To obtain a better understanding of the physical behaviour of ODM, vibration and acoustic sensors through new theoretical models and through a series of bench test experiments on helicopter gearboxes, especially in terms of ‘ageing effects’ and ‘progressive emergence of failures’ for rotating parts;
- To define innovative self-adaptive algorithms enabling data-driven automatic learning to analyse time evolutions of sensor data and to generate anticipated health diagnosis, taking account of ‘vehicle usage context variables’;
- To test these algorithms on helicopter fleet vibration data;
- To evaluate the feasibility of automated health monitoring of helicopter fleets by self-adaptive software analysis of (ODM + vibrations) data.
The project work breakdown structure included three sub-projects (SP):
- SP1 was concerned with project management, scope specification, results evaluation and dissemination towards potential end users.
- SP2 addressed experimental data acquisition and physical modelling of three key categories of sensors known to have discriminating capabilities to monitor the health of helicopter rotating parts: oil debris monitoring, vibrations and acoustic emissions. The main goal of SP2 was to reduce the rate of undetected faults.
- SP3 focused on innovative multi-sensor diagnosis software tools and explored the diagnosis potential of self-learning algorithms. It included five work packages addressing a helicopter fleet sensor database, external variable impact on vibration-based diagnosis, multi-sensor data fusion for diagnosis, automatic elimination of defective sensor data and the overall technical evaluation of the project outputs. The main goal of SP3 was to reduce the rate of false alarms.
Funding
Results
The main project outputs were:
- end-user needs / requirements and specifications for failure diagnosis and maintenance of helicopter gear boxes;
- extensive test results and data from laboratory gearboxes;
- advanced signal processing techniques applied to vibration and AE recordings;
- identification of diagnosis parameters extracted from data processing and performance assessment of these parameters;
- comparison of diagnosis indications from ODM, vibrations and AE;
- identification of contextual parameters (oil temperature, speed, load) effect on recordings;
- analysis and explanation of ODM behaviour;
- data from tribological analysis of gear contact mesh conditions in laboratory gear tests;
- modelling and analysis of gear mesh distress indicators from gear tests;
- comparisons of tribological analysis data and distress indicators with condition monitoring signal analysis;
- large base of vibration data recorded during flights of several helicopters in a period spanning over several years; recordings include data from healthy and faulty situations;
- methods to analyse large quantities of multi-sensor vibration data and to diagnose healthy / faulty situations; the methods use several techniques for data mining, automatic elimination of degraded recordings, data structuring, data aggregation, modelling of flight contextual variables, self learning based on healthy situations and diagnosis;
- software tool prototypes to implement the processing techniques and the diagnosis method;
- performance assessment of the developed methods and supporting tools;
- assessment of potential operational benefits of the project results and feasibility analysis of the industrialisation of the developed methods and software tools in the context of helicopter maintenance;
- preliminary plan for using and disseminating the project knowledge.
The project's results were disseminated through various activities, namely its website (http://www.adher-project.org - not maintained after the project completion), papers and press releases and contacts with potential end users.