<img height="1" width="1" style="display:none" src="https://www.facebook.com/tr?id=334894647031124&amp;ev=PageView&amp;noscript=1">

The Ground Handling Blog

Mototok's blog for Hangar Professionals

Written by Mototok on June 24, 2019 // 1:00 PM

How aviation can profit from predictive maintenance

aviation maintenance

Predictive analytics in data analytics has proven to be a game changer in retail. It allows retailers to fine tune marketing strategies to almost unbelievable levels of accuracy, drilling down precisely what consumers are most likely to purchase and when.

But this only begins to scratch the surface of data analysis and analytics – predictive analytics, often used to determine customer behaviors, is also proving capable of predicting future events to include equipment failure. This sort of behavioral analysis is remarkable and directly useful to predicting change, breakage, and failure in just about any industry to include aviation equipment, both aircraft as well as ground equipment.

What is predictive maintenance with machine learning?

Predictive analysis is a non-invasive process which uses sensory equipment, connective devices or machines to send raw data into a cloud platform. As data is constantly added, statistical algorithms and machine learning are utilized to identify statistical trends over the duration.

Machine learning is a branch of artificial intelligence which teaches machines how to learn. It is essentially the process by which machines (complex computers) learn to sift through enormous amounts of data and come to a conclusion of its own accord without being specifically programmed to reach a decision. When we refer to machines in this context, it refers to highly advanced computer systems rather than common hardware and software systems such as PCs and smart devices. The programming in these common devices is developed for simplicity and to follow in designed parameters.

Complex machines built to analyze big data are then used to automate the creation of reliable and repeatable decisions based on amounts of data too vast for humans to process on their own. It can be applied to systems where the desired outcome is a known quantity (such as the fleet health of a given system).

Traditional preventative maintenance (PM) methods

This is a very simplified rendition of what data analytics is – it is unbelievably complex, building models from incomprehensible amounts of raw information. The traditional preventative maintenance theory and practice is a combination of processes which are either reactive (replacing a broken component), periodic (conducting repairs or inspection at a specified timeframe), or proactive (replacing highly-stressed or defective components at an early stage to prevent failure). These have proven fairly reliable over time, but they are also costly.

Reactive maintenance may come at the result of the failure of a benign component, or it could be potentially life threatening in an aircraft. Reactive maintenance can also end up meaning replacement of multiple components due to the failure of one. For example, a failed hydraulic pump can spread metal shavings throughout the entire system, which can then damage or destroy other components in the system.

Periodic maintenance is generally scheduled in measures of either equipment hours or isochronous. In the world of aviation, periodic maintenance entails an inspection component as well as maintenance tasks. The inspection portion is still ultimately reactive in nature because it is identifying areas which have already suffered damage. Areas of chronic pre-failure are catalogued and often a technical time-compliance inspection, repair, or modification is ordered. But again, this is still a reactive method.

Proactive maintenance is conducting maintenance prior to the mean time to failure (MTTF). This is a costly endeavor because it can include overhauling extremely costly items like engine and propellers significantly prior to when it is mechanically necessary. With overhaul costs well into the hundreds of thousands of dollars (or more), conducting proactive maintenance before it is mechanically necessary can add up in cost very quickly. If you are overhauling an engine at 4,000 hours but it is mechanically sound to 5,500 hours, you are wasting over 37% of the actual life cycle.

Predictive maintenance in aviation

A new cog in the PM process is emerging: Predictive Maintenance. Using sensory equipment to gather raw data from a variety of systems and subsystems. These are not necessarily there to determine system health; instead, they are simply there to collect. It is up to the complex machine learning systems to figure out how the data is interpreted and what to do with it, hence the predictive descriptor.

Vehicles and aircraft have used central maintenance systems and centralized fault identification systems for years now: The On-Board Diagnostic System (OBD) has been the standardized system for automobiles since the 1980s, starting with a series of lights which flash codes, and now uses a scanning tool to pull and erase codes from the control module. Aircraft have used similar, although much more complex, systems since around the same time. But these systems are still reactive at their core because the data is pulled manually after operation.

Emerging systems are able to transmit recorded data from the air, or immediately upon landing when it syncs automatically once wireless data is established. With aerial broadband, data can be streamed to support real-time predictive maintenance. This yields immediate benefits because ground technicians can now be on scene with parts awaiting the aircraft if it is determined that the part has either failed or is predicted to fail. This is where IoT sensor packs installed on the entire fleet are imperative: AI systems will establish baselines spanning across dozens-to-hundreds of aircraft which provides a highly accurate database for each system, allowing AI to determine component health and alert technicians of failing systems.

How predictive maintenance completes the PM process

Predictive maintenance is not going to usurp the other three modules of preventative maintenance; it is going to strengthen them, supplement them, and be used to make preventative maintenance as a whole far more effective. While it offers real-time or near real-time access to flight and operational data in aircraft (or any other machinery it is installed on), it is still ultimately a reactive measure if being used exclusively to determine component failure or pre-failure.

Predictive maintenance in the PM process is a fantastic addition to the process because it takes out a lot of the blind spots inherent to reactive and preventative maintenance strategies. The truth is that machinery requires all of these to function optimally. Let’s take a look at these:

  • Reactive maintenance:
      • Very simple; a component breaks and is repaired or replaced.
      • Easy to implement and run.
      • No costly software or sensor packages to purchase and train technicians on.
      • Unscheduled downtimes are totally unpredictable and will likely involve overtime labor.
      • Unexpected breakage reduces the life cycle of the machinery.
  • Proactive maintenance:
      • Leans forward to replace or service components before actual failure or MTTF. This significantly reduces overtime and unscheduled downtime.
      • This will take some time to implement fully, especially if there are a large amount of assets in the fleet.
      • A much more complex system than reactive maintenance and will take time to get off the ground.
      • Prone to excessive maintenance. While not damaging, it is costly because it requires additional downtime built into the schedule, as well as costly parts and supplies.
  • Predictive maintenance:
    • By anticipating failure based on hard data, unscheduled downtime will be significantly reduced.
    • You will have a real-time view of fleet health at any given moment. It greatly reduces time spent on troubleshooting maintenance.
    • On the downside, the upfront costs are high to purchase monitoring equipment, software, and sensors.
    • Requires specialized skills to interpret and utilize the information, which may take either specialized training or hiring additional staff.
    • This will take time to set up and integrate into maintenance practices.

There is no way to predict mechanical failures 100% of the time. Sometimes parts just fail without any indicators, so reactive maintenance will always be a pivotal part of the maintenance process. Overtime will always have to be built into the budgetary equation for equipment and fleet maintenance.

Proactive maintenance is equally important: systems subjected to high heat and harsh conditions (e.g., jet engines and their accessory drive systems) will fail prematurely if not serviced on a regular basis. Data from years and tens of millions of operational hours have provided consistent parameters which uses can expect MTTF. These are the tasks which ensure machines operate at their peak performance.

Predictive maintenance will be mutually beneficial to the other systems. It will indicate in real-time which components and systems are not operating in accordance with design engineer parameters, which will positively impact reactive maintenance by alerting technicians to a certain component which should be inspected immediately. Over time, predictive maintenance will boost preventative maintenance by honing life cycles of assets to much more precise measures than ever before.

Conclusion to predictive maintenance in aviation

In a nutshell, yes, data analysis and analytics play a big role in the prediction of machinery failure prior to actual failure, but it is not a simple solution. It is costly and time consuming, but the long term payoff is significant. It is also a significant step forward in fleet safety, regardless of whether it is a piece of ground support equipment or an airliner. Predictive component failure on items which can fail catastrophically and cause serious harm and death is of tremendous consequence. Obviously, there is the cost saving aspect as well: it is far cheaper to replace a failing component of a turbine engine rather than overhauling one after a full failure. This is truly just the beginning of what big data can offer aviation and aerospace, and the future is very bright and exciting to observe how the face of maintenance will adapt to this constantly evolving technology.

Find out how electric GSE can support predictive maintenance – request a free consultation today!

Comments

Not sure if this tug fits your specific aircraft?
Let’s have a look!

We will get back to you within one business day. (Probably quicker, we’re German.)