Predictive maintenance seeks to forecast time-to-failure to minimize downtime, and thereby maximize equipment lifetime.
By Don Van Dyke
ATP/Helo/CFII, F28, Bell 222.
Pro Pilot Canadian Technical Editor
Predictive maintenance (PdM), on the other hand, bases aircraft maintenance on Big Data analytics and artificial intelligence (AI), which are used for operational simulations that often lead to insights that may be otherwise masked.
The main obstacles to PdM implementation for business aircraft operators have been difficulties assembling relevant Big Data, acquiring analytical processing capabilities, and justifying associated costs.
Since customer experience is crucially important in business aircraft operations, monitoring aspects of PdM must go beyond airworthiness to regard appearance, convenience, and esthetics in the cabin.
The continuing airworthiness of an aircraft or component is ensured through associated maintenance, repair, and overhaul (MRO) activities which are broadly related to aircraft make, model, age, and type of operation.
Traditionally, airworthiness management correlates with 3 main parameters – utilization (flight hours), cycles (landings), and calendar-based events (scheduled maintenance). Over time, aircraft maintenance accounts for as much as 35% of an aircraft’s annual operating budget.
Warranties. Original equipment manufacturers (OEMs) issue limited warranties on new products and components, so that related expenses are largely covered in the event of accident or equipment fault involving these proprietary items within linked periods.
Warranties add appreciably to the value of the purchase. According to Conklin & de Decker, labor and parts for new aircraft may cost 15% and 30% less, respectively, during the warranty period.
Post-warranty MRO. The post-warranty (aftermarket) MRO period is active and complex. OEMs continue to monitor reliability data so that maintenance programs prescribed for each aircraft, system, or component can be amended – subject to regulatory approval – to maintain airworthiness, yet avoid unnecessary upkeep.
Several OEMs offer extensions (eg, 2 years or 500 engine operating hours) to provide warranty-like protection beyond the original warranty period.
Viable maintenance strategies optimize component life and risk of failure, often maximizing operational utility by replacing parts that may still have significant remaining life, and by anticipating breakdowns based on experience.
Aftermarket mx strategies
Three primary aftermarket maintenance strategies are currently used – preventive (PM), predictive (PdM), and corrective (CM). Each of these has unique benefits and disadvantages, depending on the system monitored, its life stage, and the operational and business impact of downtime.
The main differentiators are their relative effectiveness in meeting airworthiness requirements, containing maintenance costs, and avoiding or minimizing operational interruptions.
PM relies on trend monitoring to avoid faults by inspecting or replacing components at scheduled maintenance intervals. PdM, on the other hand, is condition-based, and involves monitoring in-service component health and deterioration, using data and trends to determine when preemptive maintenance action should be taken. And CM is reactive and requires minimal effort to implement, but asks the self-insured owner to assume total risk and related costs.
Return on investment (ROI) in a given maintenance strategy reflects the value of meeting operational and utilization goals reliably against associated maintenance costs and asset management. Reliability-centered maintenance is an amalgam of these primary strategies which some authorities recommend be applied in the following indicated ratios: PM (25–30%), PdM (45–55%), and CM (10%).
Features of predictive mx
Wider application of information technology (IT) in aviation benefits from smarter, sensor-enabled assets; lower-cost sensors making more asset health data available, often in real time; cloud computing offering inexpensive connectivity and feature-rich data-management tools; and reliable analytics, AI, and machine learning (ML) technologies, leveraged for asset performance management (APM).
PdM is a main focus of future industry expenditures. It was cited as “extremely important” by 55% of respondents to a recent Honeywell business aviation survey, which concluded that this is the next frontier of connected technologies.
GE Aviation uses PdM to address significant maintenance challenges such as poor data visibility and insights, lack of internal resources and expertise, investigation and prioritization of maintenance issues, and ineffective maintenance fixes.
To date, its identified solutions have been applied to 13 business fleet types.
The IBM Watson Discovery Service, for example, combines AI and sophisticated analytical software. It performs as a question-answering machine providing precise responses to complicated questions with capabilities far superior to earlier maintenance expert systems.
Benefits. PdM tasks can be performed on an as-needed basis, while minimizing the risk of unplanned outages to increase dispatch reliability, repair/replace components prior to failure, avoid unscheduled maintenance events, improve maintenance efficiency, and optimize operating costs.
PdM employs non-intrusive testing techniques to evaluate asset performance trends. Additional methods used can include analysis of thermodynamics, acoustics, and vibrations, as well as infrared analysis.
Continuous developments in Big Data, machine-to-machine communication, and cloud technology have created new possibilities for investigating information derived from assets. PdM offers cost savings over routine or time-based scheduling, as well as proactive parts stocking and efficient scheduling of repair and manpower resources during downtimes.
Condition monitoring. Onboard condition monitoring serves to reduce the likelihood of catastrophic component failure. This involves 3 elements – relevant data collection from aircraft health-monitoring technologies, predictive analytics to anticipate and prevent inflight events, and asset maintenance workscopes based on utilization, environment, and operating practices.
Data sources include flight data recorder (FDR)/quick access recorder (QAR), manual recordings of cockpit instruments, and engine/airframe health monitoring (EHM/AHM) systems – often at the component level – via ACARS, sensors, actuators, and other controls.
Data are then interpreted, overlaid with historical information, and subjected to advanced analytics supporting models which predict fault behavior.
Related costs of required maintenance and engineering data processing can be prohibitive for some – a fact which encourages sharing related infrastructure. There are 2 major cost centers in which communal PdM could relieve costs and consequences – unscheduled maintenance and no fault found (NFF) events.
Unscheduled maintenance. These are particularly disruptive events since they often occur away from base and the means to quickly restore aircraft airworthiness. The traditional model requires parts replacement in the event of a fault – a reactive rather than proactive action that risks misidentification of both the cause and cure of the fault.
Failure prediction, fault diagnosis, failure-type classification, and recommended relevant maintenance actions are all hallmarks of the more forward-looking PdM methodology.
No fault found. This occurs when a part is suspected to be faulty but subsequent testing fails to confirm the diagnosis. NFF investigation reports often conclude “Cannot be replicated,” “No cause can be identified,” or “Retest OK.” NFF is a potentially hazardous condition with increased risk of maintenance-induced fault, especially when replaced parts are found to be airworthy.
Associated costs in terms of time and effort can be significant. Across both military and civil aviation, 3–5% of maintenance manhours are officially coded as NFF, but some sources contend that 10–15% may be more realistic.
PdM may prove particularly effective in dealing with NFF determinations, which are often time-consuming and expensive to characterize.
An important step is to identify the NFF components most often replaced – a task for which PdM is especially well-suited. ARINC Report 672, Guidelines for the Reduction of No Fault Found, suggests that operators should take a holistic view of the NFF problem, including its impact on design, documentation, training, testing, and communications.
A cargo carrier found, for example, that reported flight control computer failures were caused not by flaws in the computers, but rather by aging wires connected to the units.
In another case, a carrier and an OEM learned through destructive testing of circuits that, after 10 years, some microchips lost their moisture-repellent coating. With that knowledge, the vendor agreed to replace the microchips at specified intervals.
The future of PdM
Before PdM achieves wide acceptance, challenges pertaining to data sourcing, data ownership, connectivity, and regulatory support await comprehensive resolution.
Data sourcing. According to an Oliver Wyman survey, the global commercial aircraft fleet could generate 98 million terabytes (TB) of data by 2026. The large data set available for predictive analysis includes onboard sensor data, aircraft utilization, component removal and installation records, maintenance- and pilot-reported defects, base maintenance task card findings, and other similar sources.
The EASA Cockpit Voice and Data Recorder (CVDR) and Flight Data Recorder (FDR) mandate requires 25 hrs of datalink and 70 hrs of flight data recording for new aircraft with a maximum certified takeoff mass (MCTOM) greater than 27,000 kg (59,535 lb) entering service on January 1, 2021 or later. This will further increase the Big Data available to PdM systems.
Data ownership. The vast data generated and stored in pursuit of PdM comprise significant and valuable intellectual property.
A well-structured and managed PdM program can help to retain asset value by providing a secure, verifiable database of related maintenance.
One early approach ensures that the operator controls which datasets are shared, that identifiable data is owned by the operator, that it’s only visible to the operator and the OEM, and that it’s never shared outside of the OEM.
Data platforms. A data platform is a repository which exchanges data between applications and systems, such as between an electronic technical log (ETL) and the operator’s monitoring & evaluation system. Honeywell FORGE is an example which supports its Connected Maintenance application.
Initially, engine and airframe OEMs led the acquisition and processing of data, using proprietary software to manage data effectively. As aircraft become increasingly sophisticated, so does the need for solutions accommodating multiple fleet types involving different data standards and forms.
It is no longer prudent to keep Big Data in-house. OEMs, operators, and MROs now seek to share data to realize mutually beneficial goals.
Legal aspects. Many jurisdictions have legal restrictions on data use, sharing, and privacy. This may bring into question which laws and practices will normalize managing the quality, use, and sharing of aircraft-related data.
Poor response to PdM data could, in future, be regarded as malpractice, as when a doctor mismanages health data. Legal liabilities that could accompany the growth of PdM may lead to new requirements for data which must be shared with regulators.
And regulators may also have to decide what minimum MRO data technology will be required to hold certain licences or certifications.
Artificial intelligence. By integrating contextual data and component behavior, AI is an increasingly important influence on PdM to improve the analytical reliability of remaining life and recommended mitigation strategies. Combining PdM, AI and Big Data may highlight component replacement options and suggest other concurrent maintenance needs.
Business aviation maintenance planning and budgeting must regard airworthiness, operational demands, and cabin/galley esthetics, as well as account for cost-saving measures.
Predictive analytics lead to more timely maintenance that can be performed before an issue becomes hazardous. In turn, this means maintenance cost reductions, greater component reliability, lower inventory requirements, and shorter maintenance turn times.
The result is that PdM may offer business and operational benefits supplemental or alternative to those of PM and CM programs.
The potential benefits of PdM supporting evidence-based decision-making are most profound when considered collaboratively and appropriately with the other maintenance strategies.
This technology trend is just beginning to affect business aviation MRO widely. It is critical for OEMs and aircraft owners/operators to adopt an open interest in its ongoing development and application.
Don Van Dyke is professor of advanced aerospace topics at Chicoutimi College of Aviation – CQFA Montréal. He is an 18,000-hour TT pilot and instructor with extensive airline, business and charter experience on both airplanes and helicopters. A former IATA ops director, he has served on several ICAO panels. He is a Fellow of the Royal Aeronautical Society and is a flight operations expert on technical projects under UN administration.