Smarter maintenance planning at Fokker
Fokker Services Group
AI in aircraft maintenance and MRO
We combine maintenance history, sensor data and operational data to plan ahead for work orders, parts and hangar capacity at Fokker Services.
Smarter maintenance planning: we built a planning tool for Fokker Services using maintenance and sensor data
We combined maintenance data, sensor data and operational data in a predictive model on Databricks in Azure. The tool shows when maintenance is actually needed and helps planners plan work orders, parts and capacity in advance.
Why does our hangar schedule keep changing whenever an inspection turns up another finding?
The planner prepares the work packages from the maintenance programme, but an additional finding, a part that’s out of stock or one fewer technician disrupts the whole week. A model based on your historical data shows where bottlenecks are likely before they happen. The planner makes the decision.
From hangar planning to invoice checks: models built on data an MRO company already records
Maintenance at the right time
For Fokker Services, we combined maintenance, sensor and operational data into a model that shows when maintenance is needed.
One data platform for planning
For Fokker Services Group, we bring fragmented sources together in a medallion architecture on Databricks, with data quality checks for planning, inventory and costs.
Checking invoices by exception
For Fokker Services Group, a model predicts the expected invoice amount based on the work delivered. Staff check only significant discrepancies.
Answers from internal documents
For Fokker Services Group, we built Anthony Chat, which uses retrieval-augmented generation to find answers in internal documents.
One type of maintenance or one group of components first, then the entire fleet
-
Planning decision and data sources
Which decision needs improving, such as allocating hours to address findings or staging parts, and which work orders and sensor data are relevant?
-
Pilot model using completed work
We train the model on completed work packages and test it on a later period, with maintenance engineers assessing the results.
-
Recommendation for the planner
The recommendation becomes part of the planning process, with monitoring for drift. The planner still makes the planning decisions.
What maintenance data do we need before a model can tell us anything useful?
A good fit if
You’ve been recording work orders, findings and component replacements digitally for years, and the same decision keeps coming up, such as how many hours to reserve for findings during a heavy inspection.
Not a good fit if
Work packages are mostly on paper or in separate files without consistent codes for tasks, parts and records. In that case, you need to start with the data platform, and we’ll say so.
What determines the scope
The number of sources, such as your MRO system, ERP, inventory and sensor data; the number of aircraft types or components; whether it remains a pilot model; and who manages the model afterwards.
Does a model affect our maintenance sign-off?
No. A model makes a planning proposal. Maintenance sign-off remains with your Part-145 organisation, and changes to the maintenance programme go through your CAMO and the authority.
Off-the-shelf planning tool, bespoke model or data platform first: where do you start?
Planning tool
Choose this if you want to schedule maintenance around available capacity and operational needs, using fixed rules and priorities. Read more about our MRO planning tool.
Bespoke model
Choose this if you want to predict when maintenance or additional work will be needed. Our advice: start with one type of maintenance or group of components, as we did with Fokker Services, then expand.
Data platform first
Choose this if your maintenance data is spread across separate systems without shared codes. A data and AI platform makes every subsequent model easier to build.
Six projects for Fokker Services Group, from a data platform to a language model
Based in Amsterdam, Data Science Lab has worked with data and AI since 2016 and is a Databricks and Microsoft partner. For Fokker Services Group, we worked on maintenance planning, a data platform, invoice checks, KPI and market dashboards, and Anthony Chat.
A planning question from your hangar is enough for a first conversation
Stef Stoelwinder
Business DevelopmentStef is your first point of contact. In our first conversation, we’ll discuss one planning question and the maintenance data you already have.
E-mail info@datasciencelab.nl
Phone +31 (0)20 3636 724
Which maintenance decision do you want to support with your own data?
Describe the maintenance or planning question and the data you already have. We’ll reply by email or arrange an initial conversation.
Message sent
Thanks for your message. We’ll be in touch as soon as possible.
Frequently asked questions
MRO stands for maintenance, repair and overhaul of aircraft and components. AI in MRO means training a model on data the maintenance organisation already has, such as work orders, findings, component replacements, inventory and sensor data. This helps you plan maintenance, parts and capacity further ahead. It is one of the AI applications we build.
We have worked this way in aviation for Fokker Services Group, from maintenance planning to invoice checks.
Under point 145.A.47, a Part-145 organisation must plan the availability of personnel, tools, materials, maintenance data and facilities, while taking technicians’ fatigue into account. A model trained on completed work packages estimates how much work an inspection will generate, so you can plan accordingly in advance.
If you mainly need to allocate work across available capacity using fixed rules, our MRO planning tool is a better fit. Read more about optimising planning with AI.
Not on its own. Under M.A.302, the maintenance programme follows the design holder’s instructions for continuing airworthiness, usually those of the manufacturer, and must be approved by the authority. Longer intervals can be proposed on the basis of periodic reviews, but not through indirect approval for safety-related tasks. If the programme is based on MSG logic or condition monitoring, it must include a reliability programme.
A model can provide evidence for that programme. See predictive maintenance, and, for industrial installations, the process industry.
More than you might think. Under point 145.A.55, a Part-145 organisation must record and retain details of maintenance performed, including the release to service. There are also work orders and task cards in the MRO system, inventory and purchasing data in the ERP system, and sensor data from aircraft and components.
The challenge is that each source uses its own codes. For Fokker Services Group, we bring these sources together in a single data platform; see also our data and AI platform.
By using a language model that first retrieves relevant passages from your own documents and then formulates an answer. This is called retrieval-augmented generation (RAG). We built Anthony Chat for Fokker Services Group so employees can ask questions about internal documents.
Make sure every answer cites its source: approved maintenance data remains authoritative. Read how this works in our article on internal LLMs and RAG and on our generative AI page.
Yes, provided the work can be linked to the invoice. For Fokker Services Group, a model predicts the expected invoice amount from the work delivered and flags significant discrepancies in a dashboard. Employees only check those exceptions. For a ready-made option, see SmartInvoice.
For quotes, we connected the ERP system to external marketplaces in a single dashboard for each part.
EASA is developing guidance on AI in aviation. Its concept paper groups applications by level: human assistance, human–AI collaboration and more advanced automation. A proposal for the third edition was published in June 2026.
If AI is a safety component of an aviation product, the requirements of the AI Act are incorporated through aviation rules. Your legal counsel or compliance team determines whether your application falls within scope; we can help you assess it. See AI Act and Responsible AI.
When you want to know how many hours an inspection took, rather than how many the next one will take. A Power BI dashboard is more useful then, like the KPI dashboard we built for Fokker Services Group. If your planning is mainly a matter of rules and priorities, look at the MRO planning tool first.
If work orders, parts and records cannot be linked, start with the data foundation. If a simpler solution is a better fit, we will say so.
Where maintenance planning meets an off-the-shelf tool and Fokker’s approach
- Data and AI for mobility and logisticsHow we help planners spot deviations in schedules and capacity earlier.
- MRO planning toolOur off-the-shelf tool schedules maintenance based on capacity and operational priorities.
- Webinar: making maintenance and planning predictableHow Fokker Services Group moved from manual planning to a predictable operation, on the agenda for our webinar.