Applications of AI in Aerospace and Defence Design: Intelligent Aerospace

Aerospace and defense companies are currently transitioning to new products and components. Concerns are practical, ranging from safety to fuel consumption optimization and full autonomy.

Furthermore, global trends such as emissions reduction vehicle electrification are forcing us to think beyond fuel efficiency and embrace a green transition in the industry. This promotes the search for a drastic increase in speed and efficiency of product design development.

In recent years, Artificial Intelligence (AI) and AI systems have become buzzwords in all industries, including the aviation industry and defense and space applications. Will AI systems facilitate the green transition and help design engineers to be more effective?

What Is Intelligent Aerospace?

An example is aerospace and defense companies adopting AI systems to speed up their design cycles. What is the purpose of AI technology, and what are the solutions? Why should business leaders in aerospace industries be involved?

Rather than a complete overview of all possible applications impacting operations, maintenance and impact on air traffic management and pilots, or factory automation; we will deep dive into Artificial Intelligence technologies that help the aerospace industry to take better decisions in the product development phase.

Application of AI to Aerospace and Defence Design

We will focus on how AI and machine learning can effectively learn from design systems such as CAD and CAE and produce “real-time” solutions with examples of using Artificial Intelligence to extract data while meeting data safety concerns.

After reviewing the role of technology startups and their practical success story with industry leaders, we will show how AI can reduce risk and cost and increment innovativeness by giving a tremendous boost to design engineers thanks to the power of AI and its novel predictive analytics capability.

AI Technological Startups, Aerospace and Defense Companies and Their Data

Many companies like Neural Concept in Switzerland were initially AI startups focused on developing high-end AI technology such as neural networks. AI and machine learning models are already helping the industry to facilitate an inefficient supply chain and address safety concerns with predictive tools based on big data, implementing automation of designs.

Pioneering applications showed how neural networks could optimize the whole shape of an unmanned aerial vehicle, thus incrementing the vehicle’s autonomy.

UAV being shaped by a neural network for optimal aerodynamics, while respecting constraints (box).
UAV being shaped by a neural network for optimal aerodynamics, while respecting constraints

Machine or, more specifically, Deep Learning means that Artificial intelligence (AI) works on data. The fit is perfect since the industry daily produces vast amounts of data.

The issues to overcome in data management are twofold:

  • data are often highly fragmented or poorly stored;
  • defense and aerospace industries have security and safety concerns related to data propagation in the supply chain or outside.

Data Extraction? Not an Issue.

AI’s ability to extract data can be implemented in any data-generating technology. Still, here we will focus on Computer-Aided Design (CAD), i.e. 3D representations of object shapes, and Computer-Aided engineering (CAE), i.e. 3D representation of object functionality.

All industries have safety issues, but aerospace and defense are particularly sensitive. While Neural Concept provides the Artificial Intelligence tool, the CAD and CAE data can entirely reside within the company to ensure safety of proprietary and sensitive data.

This is why Neural Concept was awarded substantial investments for propagating its technology in industry-grade implementations.

Pioneering AI Applications

Aerospace and Defense Companies Get More Design Speed with AI

AI can help to address organizational issues, such as a more efficient technological supply chain or technical issues, such as optimized fuel consumption, thanks to optimized designs.

Predictive analytics, in other words, using computer techniques to simulate events before they occur, can significantly help. What was hindering the application of predictive analytics was its sometimes inadequate response time.

In 2019, Neural Concept demonstrated with the aerospace industry leader Airbus an application of Artificial Intelligence to aircraft aerodynamics.

Using Neural Concept Shape as a surrogate of CAE (Computer-Aided Engineering), Airbus reduced the average time for its engineers to deploy predictive analytics to support design choices, and in 2021 it decided to renew its collaboration with Neural Concept.

Airbus Business Case, 10’000 X Speed-up in Design

Neural Concept now collaborates with Airbus to accelerate the engineering process and generate new design solutions across various design aerospace and defense problems in fluid dynamics, structural engineering, and electromagnetics.

Thanks to the implementation of AI, Airbus dramatically reduced the time it takes to predict the pressure field on the external body of aeroplanes.

The improvement was from an initial one-hour time with the traditional CAE approach down to 30 ms with machine learning. Thus, machine learning accelerated computational processes over 10’000 times.

Conversely, this means that given a certain amount of allocated time for design within an aerospace project, a product design team can explore 10'000 more design changes than before. Consequently, Airbus engineers approved the usage of Neural Concept Shape.

Intelligent Aerospace: Software and Hardware Investments

The Defense and Aerospace industry is known for investments in massive parallel computing. This is a way to address the need to accelerate processes in predictive analytics. In November 2022, an Italian aircraft industry's substantial private computational cluster ranked fourth globally.

Massive investment - GPU compute farm
Massive investment - GPU compute farm

What are the lessons from the Airbus case? It shows that even without massive hardware infrastructure investment, “Intelligent Aerospace” can leverage AI to bring design tools to engineers directly on their desk PC or notebook, without the need to invest in hardware compute farms for daily usage heavily.

Use Cases for Artificial Intelligence in Aerospace and Defense

We will review a few use cases of Artificial Intelligence and machine learning technologies for the aerospace and defense industry to increase product performance and for more innovative technology development.

Neural Concept Shape in action - aircraft aerodynamics in real time
Neural Concept Shape in action - aircraft aerodynamics in real time

Practical Use Case 1: Flight Envelope with Artificial Intelligence

As an example of “Intelligent Aerospace”, Neural Concept Shape is the first machine learning system to understand 3D shapes (CAD) and learn how they interact with the laws of physics (CAE).

Artificial Intelligence can emulate complex CAE simulators, giving predictions with AI in approximately 30 milliseconds versus hours or even days with previous approaches. The Neural Concept approach with AI allows engineers in the aerospace industry to extract maximum value from their data to explore designs with enhanced operational efficiency.

The specific AI systems utilized are Deep Learning and feature recognition.

Take the example of computing a flight envelope for a commercial aircraft: while most engineers usually analyze only a few operating conditions, Neural Concept makes it possible with AI to span the aircraft’s flight envelope in a few seconds over a range of velocities or angles of attack.

The customer experience with data-driven AI and machine learning differs from traditional approaches because operations become real-time.

Practical Use Case 2: Thermal Effects on Onboard Electronics with Artificial Intelligence

This aerospace project aimed to build an Artificial Intelligence-based system for thermal transfer simulations of satellite panels. The AI thermal simulation solution can help the satellite industry and, in general, any industry related to problems with the thermal management of electronic systems.

A traditional simulation takes about 20 minutes to run, while we expect Artificial intelligence to provide a result in tens of milliseconds.

The same concept of AI-based real-time simulation can be extended to other important engineering topics affecting satellites and their payloads, such as vibrations and harshness during the launch phase.

Engineers tuned the accuracy and speed of the AI tool in emulating the simulations despite varying topological changes, such as incrementing the number of chips.

Engineers evaluated the Neural Concept Shape (NC Shape) performance for this task and assessed its feasibility as a substitute for the entire simulation. They also tested NC Shape's ability to handle different topologies within a single database.

The network can successfully assign the corresponding heat flux to each component and propagate it within the panel.

The Technology Behind - Enabling Factors for Engineers

How is AI in Aerospace and Defense already of help in accelerating R&D cycles dramatically, enhancing product performances and solving the most complex engineering challenges?

Neural Concept Shape is a high-end implementation of deep learning. Deep Learning means that artificial neurons can understand three-dimensional shapes stored in the computer and learns how they interact with the laws of physics (CAE).

The AI approach can emulate professional CAE simulation software, giving predictions in approximately 30 ms versus minutes to hours (or even days) for CAE, with comparable professional quality.

Neural networks providing the same quality of predictions (right) than professional grade CAE (left)
Neural networks providing the same quality of predictions (right) than professional grade CAE (left)

Design Engineers can deploy Neural Concept Shape on their desks (without the need for compute farms or other expensive infrastructure for daily usage) to manually or automatically explore an unlimited number of designs, without calling back the resource-consuming, time-consuming CAE software.

Neural Concept Shape is the link between designers and simulation experts in the company, reducing lengthy iterations between teams.

Conclusions

The common objective of aerospace companies and Neural Concept is to keep pushing the deployment of new design methodologies based on machine learning systems.

This collaboration aims at making the power of machine learning accessible to design engineers for real-time simulation and interactive design optimization. 

This will help the aerospace industry accelerate and improve the engineering of next-generation aircraft and other AI in aerospace and defense applications.

About the author
Anthony Massobrio
Anthony is a CFD expert since 1990, working initially as a senior researcher, then moved to Engineering, acting also as technical director in a challenging Automotive Tier 1 supplier environment. Since 2001, Anthony has worked in Software & Engineering Consultancy as a Sales Engineer and manager. In 2020, Anthony fell in love with AI and has worked since then in the field of “AI for CAE” at Neural Concept and as an independent contributor.
About the author
Anthony Massobrio
Anthony is a CFD expert since 1990, working initially as a senior researcher, then moved to Engineering, acting also as technical director in a challenging Automotive Tier 1 supplier environment. Since 2001, Anthony has worked in Software & Engineering Consultancy as a Sales Engineer and manager. In 2020, Anthony fell in love with AI and has worked since then in the field of “AI for CAE” at Neural Concept and as an independent contributor.
About the author
Anthony Massobrio
Anthony is a CFD expert since 1990, working initially as a senior researcher, then moved to Engineering, acting also as technical director in a challenging Automotive Tier 1 supplier environment. Since 2001, Anthony has worked in Software & Engineering Consultancy as a Sales Engineer and manager. In 2020, Anthony fell in love with AI and has worked since then in the field of “AI for CAE” at Neural Concept and as an independent contributor.
About the author
Anthony Massobrio
Anthony is a CFD expert since 1990, working initially as a senior researcher, then moved to Engineering, acting also as technical director in a challenging Automotive Tier 1 supplier environment. Since 2001, Anthony has worked in Software & Engineering Consultancy as a Sales Engineer and manager. In 2020, Anthony fell in love with AI and has worked since then in the field of “AI for CAE” at Neural Concept and as an independent contributor.
About the author
Anthony Massobrio
Anthony is a CFD expert since 1990, working initially as a senior researcher, then moved to Engineering, acting also as technical director in a challenging Automotive Tier 1 supplier environment. Since 2001, Anthony has worked in Software & Engineering Consultancy as a Sales Engineer and manager. In 2020, Anthony fell in love with AI and has worked since then in the field of “AI for CAE” at Neural Concept and as an independent contributor.
About the author
Anthony Massobrio
Anthony is a CFD expert since 1990, working initially as a senior researcher, then moved to Engineering, acting also as technical director in a challenging Automotive Tier 1 supplier environment. Since 2001, Anthony has worked in Software & Engineering Consultancy as a Sales Engineer and manager. In 2020, Anthony fell in love with AI and has worked since then in the field of “AI for CAE” at Neural Concept and as an independent contributor.
About the author
Anthony Massobrio
Anthony is a CFD expert since 1990, working initially as a senior researcher, then moved to Engineering, acting also as technical director in a challenging Automotive Tier 1 supplier environment. Since 2001, Anthony has worked in Software & Engineering Consultancy as a Sales Engineer and manager. In 2020, Anthony fell in love with AI and has worked since then in the field of “AI for CAE” at Neural Concept and as an independent contributor.
About the author
Luca Zampieri
Luca joined the Application team in 2018, aiming to build the next generation Deep-Learning tool dedicated to CAD and CAE.
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
About the author
Thomas von Tschammer
Thomas joined the team in 2018 as Director of Operations, aiming to empower engineers with next a next generation Deep-Learning tool dedicated to CAD and CAE.
About the author
About the author
About the author
About the author
About the author
About the author
About the author