AI in Aviation

1. References

Artificial Intelligence:

1. Aurélien Géron, Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow: Concepts, Tools, and Techniques to Build Intelligent, O’REILLY, 2019
2. Max Bramer, Principles of Data Mining, Springer, 2016
3. Scott E. Page, The Model Thinker: What You Need to Know to Make Data Work for You, HB Group, 2018
4. Stuart Russel, Peter Norvig, Artificial Intelligence: A Modern Approach, Pearson Education Inc.
5. Yuli Vasiliev, Natural Language Processing with Python and SpaCy: A Practical Introduction, No Starch Press

Other:

1. Anthony Kenny, A New History of Western Philosophy, Clarendon Press – Oxford

2. Articles

1. AI In Aircraft Maintenance Planning
2. The New Aircraft Philosophy Doc.

Official Announcement:

Dear Civil Aviation Community,

Closing 2025, My New Year (2026) Wish:

Official Announcement:

As of 2025 it becomes more and more clear that we are stepping into the AI in Aviation era.
As matter of fact, it will require a lot of human and financial resources.
1. From the Human Resources prospective we will need not only SMEs from different aviation fields (Flight, Safety, Reliability, Maintainability, etc), but also we will need SMEs from different industries and sectors such as  Cybersecurity, Computer Science, Linguistics, Psychology, IT, etc.
2. From the Financial prospective AI in Aviation will attract trillions of dollars, investment will be required in aircraft reliability, safety, aircraft design, engine design, maintenance, ATC, etc.

How we can manage this type of phenomena? The answer is – Philosophy!
Since aviation is all about Safety, our record shows that any new “becoming/paradigm” in aviation always started with the Philosophy, and based on this philosophy we used to design the next aviation ecosystem, and we can use the same approach for AI in Aviation.

I would like to address the entire Civil Aviation Community:

– Governments
– Aviation Authorities (CAA, EASA, FAA, TCCA, NASA, etc.)
– Civil Aviation Organizations (IATA, ICAO, etc)
– Aircraft, Engine, and Parts Manufacturers (Airbus, Boeing, Bombardier, Honeywell Aerospace, GE Aerospace, Rolls-Royce, etc.)
– Airlines (American Airlines, Air Canada, Air France, Lufthansa Airlines, British Airways, United Airlines, etc.)
– Pilot and Maintenance Associations
– Civil Aviation Research Centres and Institutions
– Civil Aviation Universities and Colleges
– All Civil Aviation Entities (CA Publishing Companies, Airports, CA Conferences and Expo Organizers, etc.)
– Etc.

Let’s gather together for the first AI in Aviation Conference, by 2030.
And discuss the AI in Aviation Philosophy. Let’s clearly define from “As Is” to “To Be”.

Based on this Philosophy we can build the new aviation ecosystem – AI in Aviation Ecosystem.

Let’s officially open together the AI in Aviation Era.

Please like and share this post with your network.

Your comments, suggestions, opinions are highly appreciated.

PS The subject is to call the first AI in Aviation Conference, by 2030, and discuss AI in Aviation Philosophy.

Happy New Year!

Sincerely,
Ramaz Urushadze,
30-DEC-2025

2. The New Aircraft Philosophy Doc.

My Dear Aviation Enthusiasts,

This is a friendly reminder,

The “New-Old” Aircraft Design Starts with Philosophy.
The are 2 key points that need to be reflected in the Aircraft Design Philosophy:
First, shortly describe “As Is” Philosophy.
Second, describe “To Be” or new aircraft Philosophy.

In this post, I am going to describe shortly what needs to be in the New Aircraft design philosophy document from the maintenance prospective.

1. Phenomena Philosophy.
1.1 General Concept Philosophy. Describe the new aircraft concept.
1.2 What are we building?
Include – Aircraft Design and Development, Sales and Marketing, Operation, Airline Philosophy, Passenger Experience, Environment Impact, etc.
1.3 How are we building?
Include – Your Organization Management Philosophy, Engineering Philosophy, Quality, etc.

2. Safety Philosophy.
By all means, Safety is Always First in Aviation. Include Philosophy of 1309, CMR, AWL, etc. (or equivalent reqs, if you have some new solutions – describe them).

3. Reliability Philosophy.
From design and through the aircraft’s life. Aircraft Reliability, System Reliability, Parts Reliability, Structure Reliability, etc.

4. Maintenance Program (MP Development philosophy, aircraft life, types of inspections, etc.).

5. Maintenance (Maintainability) Philosophy.
Include all possible maintenance scenarios (Tasks, Labor Hours, etc.). Describe “New Design” for Maintenance. Inspections, NDTs, System Services, Fault Indication/Recognition/Diagnosis, TBS, LRU R&R, etc.

6. Aircraft Airworthiness Philosophy.

7. Aircraft Economics Philosophy. Aircraft, System, Parts, Maintenance, etc. Costs.

8. Factors.
Include all factors related to the safety, reliability, maintainability, economics, maintenance, etc., also include all know-hows that have impact on all of the above mentioned items/factors.
Include Company and M&E Structures and required skills, QA/QC, MTX Actions Rates (MAR). General numbers per optimal fleet size. In-House and Non-In-House works, etc.

9. Human Factors/Engineering. New Aircraft HF Philosophy.

The philosophy must also be described for the aircraft operation (in-service), Safe Operation, Airworthiness, Airline Operation, Passenger Experience, Airline Business and Maintenance Models, Parts/Logistics, etc.

This is a very short outline of the New Aircraft Design Philosophy Document.

I wish to all new aircraft design projects – success!

If you are building a new aircraft and need support to develop your aircraft maintenance and reliability philosophy from paper to successful EIS and Operation, please contact me and I will be happy to support your team.

(Reference: AC120-17, MTX HDBK-IATA, MTX Management Book).

Sincerely,
Ramaz

1. AI In Aircraft Maintenance Planning

My Dear Friends,

There are a lot of discussions about using Artificial Intelligence (AI) tool in aviation, and I would like to share my thoughts about AI in Aircraft Maintenance Planning (MP).

1.1 What is AI?

First let’s define what is AI. As per EASA, ARTIFICIAL INTELLIGENCE ROADMAP 2.0, Human-centric approach to AI in aviation (May 2023, Version 2.0), AI is –

“technology that can, for a given set of human-defined objectives, generate outputs such as content, predictions, recommendations, or decisions influencing the environments they interact with.”

For purpose of this article, let’s consider that AI is a tool that doesn’t need any human interreference after AI was trained (It is the only difference I see between “complex” AI tool and “simple” MS Excel (Simple? Kh-m! We still need to write Macros, use VBA, Power Query, etc.)).

1.2. Aircraft Availability – what we are expecting from AI

In aviation all our activates start from Aircraft Availability and end with Aircraft Availability.

(SAFETY is our TOP priority!)

Main contributor in Aircraft Availability, definitely, is – Unscheduled Events (UE).

Let’s see if we can use AI to predict Unscheduled Events and based on this prediction do planning (order parts, materials, tools, etc.).

1.3. AI Prediction Formula

How AI does the prediction? It is based on a mathematical formula. For example, the simplest one is Linear Regression:

Y = B + A*X

Where:

Y, X – Variable

A, B – Constants

B – describes bias constant (Scheduled Events) parameter

A*X – describes weight (UE) parameter

1.4. Failure Rate (FR) is Constant

When FR constant: A*X = 0 and B = Constant.

When B (FR) is constant it means that the event is Scheduled type and we don’t need to do any prediction.

1.5. FR is Non-Constant (UE)

As we mentioned AI is going to be really helpful if it can do accurate prediction of the failure (UE).

It means A*X is variable (not equals to 0) due to UE.

Let’s see how it works:

AMM (Aircraft Maintenance Manual) says that when we replace LRU, we have to check Seal/Packing and based on the condition replace or re-use (install) it.

This is when Seal/Packing failure becomes almost – unpredictable.

Upon Seal/Packing condition:

(a) sometimes we can wash, inspect seal and re-install it 2-3 times,

(b) sometimes just use it once,

(c) sometimes seal reliability is less then LRU reliability, etc.

1.6. How many parts do we need – can AI predict?

The question is how to predict the number of seals/packing we need for the next 5 YE?

And second question – How accurate this prediction is going to be, from operation perspective?

It is not clear, if when we replace LRU we also have to replace Seal/Packing. In other words, there is no linear dependency (between LRU and its seal replacement).

The seal replacement event is uncertain, nobody can tell us that we need to replace the seal, until mechanic removes LRU and checks the seal condition.

As matter of fact, after inspection mechanic decides to re-use the seal, he/she installs old seal,

but after LRU installation test, mechanic finds out oil leakage, it means he/she needs to install the new seal…

This is a small tip of the iceberg of UE FR.

1.7. What If

How many parts (Seal/Packing) AI is going to order for the next 5YE?

Is it okay if AI orders 50 Seals/Packings, for the next 5 YE based on provided data / prediction?

– What about UE described in 5 (a,b,c,)?

– What if the seal manufacturer (after 2YE, to improve reliability) decides to re-design the seal (new modification)? We have at least 25 more pre-mod seals in our store (company spent money).

– What if there is Airworthiness Directive (AD)? I know some companies spent millions of dollars to comply with AD.

– What if company decides to change operation model (change aircraft utilization, etc.)

– What if company decides to move from one type of aircraft to another.

– Etc.

All these “What If-s” will require human interference into the AI MTX Planning System.

1.8. If I had a million dollars

If your company has money definitely you can buy (try) an expensive tool, but you need to check how effective this tool is going to be compared to (let’s say) MS Excel ($400 – In 2020).

1.9. People – Training

Suppose, you decided to buy this (AI) tool, make sure to ask question – is this tool operator friendly? Do I need to train my people (planners, etc.) or create special group that supports AI tool?

IMPORTANT: Non-Trained (!!!) Tool Operators “mistake” can cause financial (million-dollar) problem.

(I am not mentioning – Aircraft Safety and Availability).

1.10. AI MTX Planning tool Adequacy, Accuracy, etc.

From the statistical and modeling point of view before using/buying the AI Aircraft Maintenance Planning tool to order parts, materials etc. I would find out:

– AI Model Adequacy (How it describes real life, in our case – planning)

– Numbers: Prediction vs Actual (difference in %) (3-7 YE Time Period)

– Error Accumulation (%, 3-7 YE Time Period)

– AI Model Stat Accuracy (%)

– Stat Confidence Levels (%), etc.

1.11. Where using AI can have advantage or where I would use AI

Answer is:

In the systems where input and output have linear dependence, based on the formula when bias B = Constant and A*X = 0.

For example, Planning C-Check (It is not a first C-Check (Aircraft type, Check number, etc.). Don’t forget, we have to train AI Tool – this is a given!):

– Selecting type of specialist (Avionics, mechanics, etc.)

– Recommending the number of specialists and grouping

– Assigning tasks to group and individual specialists

– Define working zones (Day 1 – Group1 works in FWD Avionic Compartment, Day 2 – Group1 works in AFT Avionic Compartment, etc.)

– Calculating Operational Maintainability characteristics,

– Inventory (parts) Management (movement), etc.

REMEMBER: All these tools don’t work without trained specialist, and it is true – Garbage-In-Garbage-Out (GIGO)! …

Sincerely,

Ramaz