AI Pushes the Envelope on Pack Performance
An Industry Fellowship between Cranfield University and Delta Cosworth has proved a concept that could reduce manufacturing costs and improve safety in high performance vehicles.
Car and motorbike battery pack designers are always trying to set new boundaries for performance, cost and safety. This pressure to improve is felt acutely at Delta Cosworth as it develops battery packs for the likes of Jaguar Land Rover, Norton motorcycles, and the British Touring Car Championship.
Since the beginning of the 2022 season, the racing cars are hybrid, offering drivers a boost button to get a little extra speed from a battery that recharges for the rest of a lap, until enough charge is available for the next burst of speed.
For both its commercial clients and its racing applications, Delta Cosworth wanted to discover if there is a way to get more performance from its battery packs, with a lower cost of production and with higher safety levels. It was a conversation with Abbas Fotouhi, then a Lecturer, and now Senior Lecturer, in Vehicle Engineering and Transport Systems at Cranfield University, that sparked an idea to investigate how AI might be used to test heat and performance parameters within a battery to deliver these improvements.
|By the numbers|
|30||kW of extra power a Delta Cosworth battery provides to a BTCC car|
|15||Number of seconds ‘boost’ the battery provides per lap|
|2||Faraday Institution Industry Fellowships awarded to Dr Fotouhi|
|20:2||The number of sensors before and after using AI|
|2021||The year Dr Fotouhi earned promotion to Senior Lecturer, which the Industry Fellowship helped to secure|
|2025||The possible launch of the Zero Emission Norton electric motorbike concept|
But why is AI necessary when a battery pack could be developed with multiple temperature sensors for each cell to give real time measurements? Matt Lowe, Head of Battery Management Systems at Delta Cosworth, explains that there are serious constraints to the complexity, placement, and cost of manufacturing that means using fewer sensors is hugely beneficial.
“Ideally you’d have many sensors but it’s not always feasible, at reasonable cost, to put them in every part of the battery,” he says. “We want to simplify manufacturing by building in fewer sensors and instead use AI to predict temperatures so we know if we can safely push performance harder. Batteries have a thermal inertia; you can’t always see they are going to get too hot until they are approaching unsafe levels. So, AI is helping us predict how we can push performance boundaries safely by predicting if there will be a problem in advance.”
To ensure the AI algorithm he was building was used to different scenarios, Dr Fotouhi worked with Delta Cosworth on a wide range of tests, looking at the data available from discharge and charging batteries of different chemistries at different rates and different aging profiles.
One of the most important parts of the tests was to equip packs with multiple sensors so the researchers were not relying on AI predictions alone – that they could be validated against what a physical sensor was measuring.
“Although we were trying to cut down on the number of sensors in the final manufactured products, we used a lot in our test packs to validate the computer’s predictive results,” Abbas Fotouhi reveals.
“We’ve shown in our testing that we can get down from twenty sensors per pack to just two. This means Delta Cosworth can use AI to estimate the impact of discharging and recharging a battery so they can get that extra bit of performance out of it, without impacting safety. It’s also very important that we’ve shown it’s possible to achieve a reduction in the cost of manufacturing.”
This successful academic/industry collaboration was enabled by a 2020 Faraday Institution Industry Fellowship. Initially for a single year, Cranfield University and Delta Cosworth were able to demonstrate sufficient progress to ensure a second application, for a further year, was successful. The next step for the partners is to seek further funding to establish if their discoveries can be commercialised.
“We’d been talking about working with one another but without the Fellowship paying for my time with Delta Cosworth, I’m not sure how we’d have been able to launch the project. From a career perspective, it’s also great. Because it’s such a prestigious award, I used it in my application for promotion and was appointed as Senior Lecturer as a result.”
Matt Lowe reveals how the Industry Fellowship also empowered Delta Cosworth to turn an early conversation into an ongoing knowledge exchange with Cranfield University.
“The Industry Fellowship worked well for us because, as a research and development company, we enjoy collaborating with academic institutions.
“We like to stay in touch with institutions who can show us the latest technology and help feed in new ideas. At the same time, we also talk with a lot of companies, so we have a pretty good idea of what industry needs, which we can feed that back into to academic institutes.”
The work so far has proven the concept that AI can replace sensors in a battery pack by helping to predict the impact of a cell’s age, chemistry as well as charge and discharge rates.
The next stage is to fund research into getting the finding commercialised and installed in a real battery pack. While the company cannot always reveal whom it is working with, one option for commercialisation is likely to be the Zero Emission Norton. The company is working with the iconic British motorbike brand on its new all-electric, high performance motorbike concept, which could see the new technology rolled out as soon as 2025.
Success story published March 2023