It can be laborious and time-consuming to troubleshoot and repair systems on aircraft. Without granular data, it can be difficult for maintenance personnel to identify what part of a system is faulty and it can take a long time to remove, test, and reinstall parts of a system until the culprit is found. Worse, if a failure occurs that unexpectedly grounds an aircraft, it can mean severe disruption and expense for the airline.
Aircraft may have a huge amount of data available but this doesn’t mean it’s the right data to troubleshoot a problem. Typically, aircraft will have certain systems that are more problematic than others but there may not be sufficient instrumentation on these systems to indicate when a failure is impending or to help isolate what part in a potentially large system needs to be replaced.
This article discusses the problem of inefficient maintenance, lack of access to the right data, and how aircraft can be retrofitted with instrumentation to more quickly identify impending failures, rapidly repair them, and minimize aircraft downtime and maintenance time.
The Importance of Maintenance Costs
Commercial air travel is a highly competitive business with potential customers having a slew of price comparison websites to quickly and easily find the best prices for a particular journey. While different airlines have other benefits to offer (superior customer service, reliability, in-flight comfort and entertainment, more convenient flight times, etc.), price is a key consideration. Thus, airlines are always looking for ways to keep costs low so they can pass these savings on to their customers.
A big emphasis on controlling costs a few years ago was fuel efficiency. Fuel prices in the early 90s were often higher than $120 (a barrel), which strongly motivated carriers to look for more efficient aircraft and retire older models; however, this motivation has diminished as prices dropped dramatically in the mid-90s. This has resulted in airlines flying older aircraft that have more maintenance requirements. Keeping these maintenance costs low is now a vital part of maintaining a competitive edge.
Like all vehicles, aircraft require periodic maintenance to ensure they are operating safely and efficiently; however, some systems will need repair or replacement outside scheduled procedures. If an aircraft develops a fault during a flight that means it has to be grounded, an “unscheduled maintenance event” occurs, which will very likely result in delayed passengers or cargo. Such events have a significant impact on an airline’s competitive advantage. In fact, this phenomenon of passenger delays being caused by carrier delays accounted for over a quarter of aircraft delays from September 2009 to 2019 in the US.
Ideally, an aircraft will have instrumentation onboard to gather data and flag any impending faults to the maintenance crew via a system such as an Aircraft Communications Addressing and Reporting System (ACARS). This data allows the ground crew to schedule an aircraft for maintenance before a fault grounds it; however, there are no perfect aircraft and many systems either have no instrumentation or the data being collected is insufficient to accurately isolate the issue, resulting in lengthy and expensive repairs. This is particularly a problem on older aircraft.
Analyzing Data to Speed Up and Predict System Maintenance
Predictive maintenance relies on the availability of relevant data and although there is a significant amount of data available on traditional aircraft, it is not necessarily generated with predictive maintenance in mind. This is because designers must meet numerous objectives when developing aircraft such as keeping them light and affordable. Since adding data acquisition hardware will increase cost and weight, systems will only be installed if deemed necessary. Designers are also not primarily concerned with catching failures; they are trying to minimize their occurrence, i.e. to capture health status, not to catch all conceivable failure scenarios.
Typically, health monitoring systems will identify that there is a problem but they will not tell you where the problem is. So, for example, the maintenance system may flag that there is a problem in the hydraulic system but will not localize the problem to a specific part. The maintenance crew must take the aircraft out of service and begin stripping the hydraulic system to locate the fault.
This process of finding the cause by elimination usually results in “no faults found” (i.e. most parts will be functioning within specification). The longer it takes to identify and therefore fix the problem, the more resources are spent to fix the problem. Taking parts off an aircraft is difficult and disruptive as well as time-consuming. Any way to reduce this can result in significant savings for the operator.
It should be noted that in practice, a pattern of faults will emerge over an aircraft’s lifetime; for example, a part may have a short mean time between failures (MBTF) that results in a pressure drop in the hydraulic system. The maintenance crew may be familiar with the issue and make educated guesses about what repairs are most likely required. Artificial Intelligence (AI) software can also help predict which component is likely to be the cause but the correct data needs to be available.
Thus, when an aircraft is operated year after year, failure patterns start to emerge. If this wasn’t anticipated, then it’s very unlikely that there are sensors installed on the aircraft to gather the necessary information to feed the maintenance computers and inform maintenance personnel. Although this is more common with older aircraft (which have far less instrumentation), it’s also likely to be the case for current and next-generation aircraft.
Some failures can only be correctly diagnosed if the data is recorded at high enough rates but the rate selected during the design phase may be too low. Hence, a spike in a hydraulic, pneumatic, or electrical system may be missed by data that is recorded once per minute.
To install sensors and data acquisition equipment to cover every conceivable scenario is expensive, bulky, and complicates the aircraft; it also adds unnecessary weight. While best efforts can be made to anticipate systems that will need more maintenance than others, the reality is that it will only become clear after a number of years or even decades of operation, where sensors could have been most usefully placed. It’s also worth noting that how and where an aircraft is flown affects how systems may respond over time. An aircraft flown regionally in Scandinavia vs. a cargo plane in the Middle East will experience different operational and environmental conditions (moisture, sand, temperature, etc.)
Gathering Data from Problematic Systems
Given that aircraft OEMs simply don’t know what systems will become more failure-prone over time, they don’t know where to place sensors. Even if they do place sensors in problematic systems, those implemented by the manufacturer when it was designed several decades ago, or more, may not be providing the right data.
Predictive maintenance is a learning process. It can take time to discover what needs to be measured, how accurate the data needs to be, how often it needs to be sampled, and so on. It can be useful to have an adaptable system, at least during the development stage, so that the system can be altered to meet the program’s needs as it progresses.
You don’t need to install a solution to gather data from every system; rather, you seek to instrument only those that cause the most disruption in order to improve your chances of correctly diagnosing the problem. For one operator’s Boeing 737 fleet, 40% of the top 100 issues and six of the top ten issues were due to a problem in one of the systems identified in Table 1.
Determining the Cost Benefit
Implementing new instrumentation on every aircraft and setting up, maintaining, and operating the analysis of the resultant data will have a cost. The benefits of purchasing, installing, and operating new instrumentation, the associated data acquisition systems, and a method to offload and analyze the data cost-effectively depends on a number of factors. These include savings in reduced fleet maintenance time, increased fleet availability, and a reduction in unscheduled maintenance events and their cost (directly and to reputation).
For older aircraft designed in the 1980s or 90s (such as Boeing 737, 777, etc.), there may be a significant amount of instrumentation required to yield performance data that is useful for fuel burn analysis, health monitoring data, and so on. In such a case, the cost of instrumenting the fleet will increase but so will the benefit. A key consideration is how much longer the aircraft will be in service and to some extent, this is dependent on fuel prices that could stay low for a decade or double in a year, i.e. the future price is uncertain.
Operators also have to consider how to transport the new data, how to analyze the thresholds of the data to initiate a maintenance action, and assess the FAA/EASA/MAA design assurance level requirements of any modifications. The particular solution will also depend on the aircraft type and operator as follows:
Airline’s new aircraft: Don’t have the right data, may need limited supplementary hardware
Airline’s old aircraft: Not enough data, need larger data acquisition systems
Small fleet of business jets: May have no data-gathering or MRO infrastructure; needs consultancy, full hardware, installation, and software solution
Curtiss-Wright’s approach is to provide the data you don’t have, gather the right data with the right level of granularity to alert you of impending problems, and to provide data to help localize where issues are occurring to save time in repair. A typical data capture unit is shown in Figure 3 where multiple analog, discrete, and digital signals from a variety of sources are acquired. There are several features of this platform that are important for predictive maintenance implementation teams.
Flexibility is crucial during a prediction maintenance program, where measurement requirements can evolve and change during evaluation/prototyping phases. There are data acquisition units (DAUs), for example, that are completely modular and scalable, allowing for interface modules to be swapped in and out as the program requirements evolve. Additional options, such as user-programmable embedded DSP for real-time analysis and rapid detection of exceedances, remove some maintenance overhead while an easy path to certification for non-critical systems lowers installation costs.
High-fidelity data is required generally, but especially in a case where the data is being used to train a predictive maintenance neural network. The DAUs have a range of data transport options such as sending data to recorders, wireless routers, or existing HUMS solutions.
Many aircraft in commercial operation today do not provide sufficient data to support predictive maintenance or optimize maintenance operations. Despite living in an era of the “data rich” modern aircraft, even these don’t collect all the necessary parameters at the right sample rate. Many systems that are not critical for safe flight, but are critical for smooth operations, are not monitored in a useful way. Problematic systems such as air conditioning and auxiliary power are notorious for failing with little or no warning and are prime targets for predictive maintenance systems.
This article was written by Michael Doherty, Programs and Product Line Manager, and Stephen Willis, Marketing Portfolio Manager at Curtiss-Wright Defense Solutions, Ashburn, VA. For more information, visit here .