Over the last several years data generated by aircraft has increased exponentially thru introduction of new onboard equipment that allows for the so-called connected aircraft. In the wake of this data growth OEM’s and airlines alike the interest in data analytics and predictive maintenance has risen alike. According to the MRO survey 2017 by Oliver Wyman 77% of respondents want to implement predictive analytics in the coming three years. The MRO survey 2017 by Aircraft IT confirms this trend as most respondents answered that next to paperless maintenance predictive analytics will play a major role in the development the upcoming years. According to the latest MRO Survey by Oliver Wyman in 2018 predictive maintenance is even seen among airlines as a means to combat rising material costs and improve labour productivity and shortage in the future.
The aim of predictive maintenance
Let’s start at the beginning what is the aim of predictive maintenance?
Within aviation maintenance and engineering the aim of predictive maintenance is first to predict when component failure might occur, and secondly, to prevent the occurrence of the failure by performing maintenance. Monitoring for future failure allows maintenance to be planned before the failure occurs, thus reduce unscheduled removals and avoid AOG.
Benefits of applying predictive maintenance
- forecast inventory
- manage resources
- minimizing the time, the equipment is being maintained
- minimizing the production hours lost to maintenance, and
- minimizing the cost of spare parts and supplies.
Based on years of in-depth research EXSYN has identified three main phases, where the last defines the status of being able to become predictive:
Let’s have a closer look at the three types of airline:
The reporting airline
The reporting airline is characterized by producing monthly reliability reports, drawing from manual work of retrieving the various information such as aircraft utilization, Pilot reports, Maintenance findings and component removals and then consolidate this data into Excel.
This typically is a very manual labour-intensive process and puts the reporting airline in a situation that they are able to produce the reports on monthly bases but not action on these reports.
The monitoring airline
The monitoring airline, stands one level higher on the adoption success scale of becoming predictive. This type of airline typically already moved away from excel reports and uses an analytical tool and mostly also dashboards. This allows the airline to act more quickly on rising issues and monitor the fleet than rather reporting the status of their fleet.
However, it still draws on the same set of data, the SPEC2000 chapter 11 reliability data. Which ultimately prevents the airline of adopting any predictive maintenance models as data sets are to limit to become reliable enough for operational usage.
The data driven airline
The data driven airline, has the highest level of adoption success for predictive maintenance models. It used all character traits of the monitoring airline but has recognized the need of having access to larger sets of data beyond its own airline data, such as industry reliability data, Flight Data Recorder information, weather data as well as ADS-B transponder data. Ultimately the data driven airline implements a data driven platform in its organization that provides information to its operational units such as maintenance operations control to make informed decisions in day-of-operation itself.
The main question you have to ask yourself: 'Where are you currently standing: is your airline ready for predictive maintenance?'
In order to answers this question, you have to go thorugh below list of questions and answer these either with yes or no:
- Has your airline an integrated MRO software?
- Is the MRO software less than 8 years old?
- Is the MRO software updated at least once a year?
- Does your airline have interfaces built between various sources of information within your IT setup? (i.e. Records, Maintenance, Engineering, Inventory, Finance & Resource, Flight Operations data)
- Does your airline conduct data audits to secure that data is up-to-date?
- Does your airlinie have a high data integrity/quality?
- Does you cross-reference and update data against sources and revisions?
- Do your airline use a software solution (not EXCEL) to perform reliability analysis?
- Does your airline need to manipulate or correct data every time before the airline reliability report is published?
- Does your airline have access to weather data?
- Does your fleet use onboard data systems such as AHM or ACARS?
- Is your airline's flight department willing to share Flight Data Recorder information with other departments?
- Is your airline willing to share technical and operational data with other airlines that operate the same aircraft type(s)?
Where are you standing
Have you answered the first four questions with yes?
Your airline has entered the first stage of the process and is a reporting airline. This means you are able to produce analytical insight into your data, however it is mostly requiring some significant effort. Typically your reliability engineer(s) spend a lot of time on producing reports and do not have sufficient time for further analysis. Also, as data is still dispersed thru various departments, time is spent on consolidation and alignment of data.
In case “no” was the answer to any of the first four questions, chances are high that your current data and system landscape is not complete enough to start implementing anything close to predictive maintenance or predictive analytics. Due to the lack of data or reduced data quality any outcome would be below required thresholds for adequate prediction models. The first next step to undertake is to perform a full holistic data scan on your MRO data to identify the required actions to proceed and check the possibility to develop interfaces between various systems, so that double data entry is avoided (Check the blog: 'Human interaction with aircraft IT systems'). If you have answered question number one with no, it could be useful to establish a list with all the requirements for a MRO software you need to work with. The follwoing blog posts could be helpful to establish the list with requirements: '4 Practical Tips during a MRO software selection' and ' 7 tips how to implement a new MRO software system'.
Have you answered all first 9 questions with yes?
That's great your airline has entered the stage of being a monitoring airline. In this stage you are able to produce real-time insight into activities without manual intervention and can continue to focus on becoming a data driven airline. However, if not don't worry you are not the only one; Based on our experience we know that many airlines are in-between the first and second stage. Often EXCEL is still present and used for the realibility analysis but there is some sort of software in place that allows for direct analysis without data corrections. Highest risk in this stage is that is seems compelling to start looking at predictive maintenance as lots of data is available. However, albeit available it might still not be enough for proper predictions.
In case “no” was the answer to either question 5, 6 or 7 your first port of call will be your data quality. You have a suitable infrastructure in place, but not enough controls yet to ensure a proper level of data quality. Check out our blog post '7 practical tips how to improve data quality' and 'How to refine data capture and collection for implementing predictive maintenance'.
Have you answered all 13 questions with yes?
Congratulations, you are ready for the successful implementation of predictive models within your airline. In case your answer to one of the last 4 questions is “no” you’re missing a vital or multiple ingredient(s) for successful and reliable predictive models. Started working on predictive models with one of these questions still on no will set you up for a failed project and potentially high sunk costs as results will always look promising enough to continue justifying allocation of time, resources and funds. However, a reliable prediction cannot be obtained without these datasets missing.