UK-based Aerogility uses model-based AI to run simulations of complex programs. Aerogility’s chief scientific adviser is Michael Luck, professor of computer science and director of the UKRI Centre for Doctoral Training on Safe and Trusted Artificial Intelligence. He said people assume that model-based AI is the same as ML, because ML has become so powerful and successful in a wide range of applications.
But while ML can analyze images or translate what people are saying, in many cases it doesn’t actually understand the significance of the content. In contrast, model-based AI links various specialist databases (also called intelligent agents) in such a way that “what if” scenarios can be generated. This also has the advantage that, if conditions change, the model can be run again.
A good example is provided by the company’s longest-standing airline client, low-cost carrier easyJet, which has used the system since 2018 to organize winter base maintenance programs and engine and landing gear overhauls. Aircraft utilization is used to calculate when the next check is due and then correlated with hangar slots at various MRO facilities to find the best match. A hangar visit is a good time to swap landing gear, so this can also be planned.
In 2019, SAS adopted the system to handle powerplant shop visit scheduling. The SAS fleet is much more diverse than easyJet’s, including Airbus A320 family aircraft, A350s and Boeing 737s. As the system was being introduced, it also had to handle the phasing out of some older 737s and introduction of new Airbus A320/321neo aircraft.
Gary Vickers
Aerogility
Gary Vickers, CEO of Aerogility, said that while big data and predictive maintenance are becoming more common, they are based on history, extrapolating trends from collected information. That can throw up new problems that have never been detected before. Using Aerogility, airlines can run realistic simulations to establish the likely effect on their operations and then develop the best solution. The simulations can be run again to see how they matched up to the real world and modified if necessary.
One of the most important aspects of model-based AI, Vickers said, is that it is understandable. There is an element of trust when you’re working with huge amounts of data that are too large for human comprehension. On the other hand, Aerogility is composed of readily understandable modules, even though the setup is a complex process.
Looking forward, he sees a new area where model-based AI can have a big impact. It could be used by airlines to simulate various methods of reducing carbon emissions, to identify the best possible outcomes for both their business and their sustainability targets before implementing them in real life. That could be done across an entire fleet while looking ahead to see how these decisions will affect their operations over the next months, years or even decades.
Aiir Innovations says users of any model of borescope can simply drag and drop the video file of their inspection into their cloud-based platform via a dedicated internet portal. Multiple parties can then view, comment on and share the findings allowing for quick decisions to be made by the group. Aiir Innovations images.
Last year, easyJet continued its commitment to advanced maintenance technology by partnering with Amsterdam-based Aiir Innovations to explore how computer vision and artificial intelligence can speed up borescope inspections and cut out errors by providing automated damage detection.
Aiir Innovations
Aiir Innovations was formed in 2016 by an assistant professor in computer vision and five graduates in artificial intelligence. They had been invited by the AFI KLM E&M engine shop at Amsterdam-Schiphol to see if they could develop a system to automatically analyze borescope video streams to identify faults such as cracks, scratches and dents.
Bart Vredebregt
Aiir Innovations
Bart Vredebregt, CEO and co-founder (and one of the students), said initial results were promising but it took a few years to return to AFI KLM E&M with a viable product.
The Aiir software, which includes automated blade-counting, uses image analysis to very quickly generate a report. Damage is flagged before the camera probe has left the engine, while historical footage can be reviewed online.
This last feature was important for their customer MTU Maintenance Lease Services. A problem engine could be anywhere in the world, at an MRO facility or even an AOG at a remote airport. With travel restrictions during the pandemic, the data could be reviewed by all interested parties on Aiir Innovation’s cloud-based platform via a dedicated internet portal. With no room for doubt, quick decisions could be made on rectification and liability.
The latest development, earlier this year, was a technology partnership with Waygate Technologies, which will incorporate a version of the advanced software into its Everest Mentor Visual iQ VideoProbe. In this case, the software will provide automatic defect detection on still images taken during inspections. This transforms the borescope into a true digital assistant, capable of spotting tiny defects that human eyes can easily miss and helping to improve inspection reliability and efficiency.
Vredebregt said many AI projects fail because they are “innovation for innovation’s sake” and they fail to take enough account of human involvement, especially when there is no associated legislation in place. As a result, while prototypes may be easy to create, they are difficult to get accepted by workshop personnel. He is very proud of the fact that the system at AFI KLM E&M is in daily operation and fully accepted by the technicians and seen as a backup to their experience. Having determined the problem, they check the software report in case something has been missed.
Ramco Systems
For Ramco Systems, ML is being used to develop value-added packages for its enterprise resource planning software, said Saravanan Rajarajan, director of aerospace and defense solution consulting and presales.
Saravanan Rajarajan
Ramco Systems
The company’s innovation lab in Singapore is working on a number of use cases that could eventually be combined as an optional package. For example, if a mechanic encounters a technical problem, they can consult the system, which will use historical data to identify the most likely cause, with a probability of around 95%. It can also provide details of the parts required for rectification. If the advice is accepted, the ERP system can then automatically process the request, including delivery, inventory management and finance. Rajarajan said the human decision is essential and avoids concerns about replacement by a machine. Of course, if there are any changes to the process — replacement part numbers, for example — they can easily be incorporated as an update.
Ramco is considering image analysis, but the use case is based on scanning packaging labels in Goods Inward and using Optical Character Recognition to enter the details automatically in the ERP system. The company is an ERP specialist and this will always take priority, he said.
flydocs says their recently launched Component Management software enables buyers to procure, sell and lease parts faster, backed by a digital trace. The solution is powered by AI, ML and blockchain. flydocs images.
OCR is also being used to analyze teardown reports. Rajarajan said they contain lots of useful information but are rarely studied in detail. High-value components can be identified and the system scans the report and places the details in the ERP system.
It can also be used to analyze invoices. A price range is selected for each component and any outliers (high or low) are flagged up.
flydocs
At flydocs, Mark Bunting, product director of asset management, component management and machine learning, said the company originally started by scanning maintenance records and using OCR to analyze them. Since the company’s takeover by Lufthansa Technik, there has been a shift in emphasis to transfer the technology into other areas. As a result, it has been a busy year.
Yet again, easyJet is involved, with a 10-year deal signed in February that will use the flydocs integration with AMOS MRO software from Swiss AviationSoftware to digitize the records and asset management of its entire fleet of over 300 aircraft. This will include lease transitions. Another AMOS-related deal, an extension for five years, was signed two months later with Wizz Air, to continue to digitize records management and technical services for over 140 aircraft.
In April, flydocs launched its Component Management software, which will enable buyers to procure, sell, and lease parts up to 50% faster, backed by a digital trace. The solution is powered by AI, ML and blockchain. The company has previously partnered with Honeywell on its GoDirect Trade marketplace as well as with the IATA MRO SmartHub.
That was quickly followed by an MoU with Pratt & Whitney’s Commercial Serviceable Assets business, a provider for serviceable material, engines and tailored solutions, to use the software to
provide an optimized solution for inventory and document management as well as a tailored inventory for existing and prospective customers.
The latest development, and another new direction, was an MoU in June with Conduce Group to develop an interface with the latter’s eTechLog8, allowing common clients to keep a central repository for their e-signed tech log pages.
Bunting also highlighted aircraft teardowns as a challenge, especially, he said, the Airbus A380.
You must be logged in to post a comment.