Bluefruit Software take the edge in edge AI with AeroSpace Cornwall funding
It’s estimated that more than 750 million edge AI chips will be sold in 2020. Just got your head around AI? Edge AI is Artificial Intelligence-driven computing on embedded systems and the growth potential in the UK is immense. Now AeroSpace Cornwall, who is part funded by the European Regional Development Fund and Cornwall Council, has awarded Bluefruit Software - leaders in providing high-quality embedded software - £100k of funding for critical research and development into this mind-blowing technology, which is forecasted to expand by 20% over the next five years.
Set to revolutionise heavily regulated sectors like medical, aerospace and space beyond recognition, edge AI works by helping machines make intelligent decisions and provide critical feedback whilst keeping the data on board a local device. This means systems can still run despite restricted connectivity and without data centre costs. It also limits safety, data privacy and security concerns.
Bluefruit’s ‘edge’ - and the reason they have been awarded this funding - is their approach to the application of edge AI for diagnostics, building upon existing image-based systems to provide real-time machine learning applications for data, ranging from audio and vibrations to ultrasonic and motion diagnostics. Bluefruit has taken on the additional challenge of trying to make their work within AI compliance friendly, so their advancements can be applied to the highly regulated medical and aerospace sectors. The need to find a way to solve the AI and compliance puzzle is one of the barriers to innovation in this space and the Bluefruit team feel their work around continuous documentation and validation could help work towards solutions that make both innovation teams and compliance teams happy.
Bluefruit's R&D project started with a proof-of-concept phase, in which the team created a machine learning solution that achieves high levels of accuracy with small amounts of sample data - one of the main barriers to applied AI. While Bluefruit built on existing image-based AI development, they intentionally wanted to explore other sources of data that haven’t been invested in as much previously. Their first prototype, the Audio Classification Equipment (ACE) project is an embedded software application that can take pre-recorded audio or live audio and categorise the sample into pre-trained categories. Making it possible to diagnose an internal machinery fault through sound recognition using real-time AI solutions.
The Bluefruit team are currently working on a second wave of prototypes, exploring other data sources within machinery and devices that can be gathered and diagnosed on the devices themselves, not requiring any cloud connections. There are now three different feasibility projects underway within life sciences, manufacturing, and defence. For these projects, Bluefruit is deploying AI on embedded ARM processors creating small diagnostic solutions that can be built into or bolted onto the equipment.
A key factor to the success of this project was the ability to use the supercomputing facilities at Goonhilly Earth Station’s Space AI Institute. The initial training of any AI algorithm is a critical part of the R&D process, but it’s so computationally intensive that each iteration can take hours, days or even weeks to achieve on standard hardware. With the added help of the Goonhilly workstations each training iteration is reduced to a few minutes meaning that Bluefruit were able to complete all of the neural network R&D and training, including porting of the algorithm to the standalone embedded edge device, within the first eight weeks of the project. Bluefruit and Goonhilly are now looking at ways to expand the relationship, working together on key applications for the space sector and contributing to the growing list of leading AI companies in the UK.
The development of this project has already caught the attention of companies in the biopharma, aerospace, and energy industries who have approached Bluefruit for consultation, proving their decision to work within this space has met an industry need.
Caitlin Gould, Director, Bluefruit Software said: “We believe using AI in this way, built into equipment, can make the entire device or machine more intelligent, more effective, and possibly safer for end users. For the companies we talk to, these are business-changing conversations. They would have the ability to have products that could predict faults before they become a catastrophic failure. This creates not only a better machine; it saves the company huge amounts in maintenance and diagnostics costs. We genuinely believe the application of edge AI in places where AI isn’t at the minute - places without the internet even - will enable it to change our daily lives; that is why we are so keen to explore and develop this technology. AI shouldn’t be reserved for the tech giants of the world. It should be something that helps improve products for smaller companies as well, and we think we can help do that. We are so excited to see what the future holds.”
Ian Jones, Goonhilly Earth Station commented: “We’re delighted to be working with the team at Bluefruit Software demonstrating the value of a collaborative way of working. Our supercomputers make a huge difference in reducing development times for our clients. Of course, the large cloud platforms also provide pure processing power alone. However, by working together we open up opportunities to create new products and markets that wouldn’t necessarily happen otherwise.”
Gail Eastaugh, Director at AeroSpace Cornwall said: “This incredible project taps into the market potential of edge AI, unlocking the innovation it can bring to the space and aerospace industry supply chains, particularly in the South West. We are excited to see Bluefruit’s AI capabilities grow, and their offering scale within this sector and beyond.
“As demand increases for edge AI applications, it is the responsibility of companies like Bluefruit Software to continue to innovate in this growing space and provide intelligence on embedded systems to future-proof industries which will benefit from this groundbreaking technology. It's our responsibility to make sure they have the funding they need.”