Virtualisation and software-orientated design processes could make the transition away from internal combustion engines more efficient. By Christopher Dyer
The automotive industry is steadily moving away from internal combustion engines (ICEs) in the wake of more stringent regulations. Some industry watchers regard electric vehicles (EVs) as the next step in vehicle development, despite high costs and infrastructural limitations in developing markets outside Europe and Asia. However, many markets remain deeply dependent on the conventional ICE vehicle. A 2020 study by Boston Consulting Group found that nearly 28% of ICE vehicles could still be on the road as late as 2035, while EVs may only account for 48% of vehicles registered on the road by this time as well.
For manufacturers, this represents a huge and multi-faceted challenge. There are not only the industry’s looming and ambitious environmental targets to consider but also the drive for CASE (Connected, Autonomous, Shared and Electric) vehicles is increasing design and development complexity. Also, there are the bottom-line pressures where European R&D spend has already increased by 75% between 2011 and 2019. Enter Secondmind, a machine learning company based in the UK. The company works with automotive engineers, helping them to use data-efficient transparent machine learning that combines the subject matter expertise of today’s engineers with algorithmic intelligence. Secondmind’s Chief Executive Gary Brotman argues that this new breed of machine learning is required to efficiently streamline the vehicle development process, helping automotive companies accelerate the transition away from ICE and ensure sustainable design and development engineering.
How is the upcoming ICE ban in some markets affecting manufacturers’ approaches to developing more environmentally friendly vehicles?
Secondmind sees the challenge as the speed at which the automotive industry can transition to greener mobility—balancing environmental and business sustainability objectives with the complex reality of having to optimise the production of ICE vehicles in the short-term while at the same time shifting more investment to electrification and what will eventually be environmentally friendly passenger vehicles.
OEMs and Tier 1s have different design strategies and tools to address these growing requirements. Many manufacturers are headlong into the transition away from the conventional automotive business model and shifting to more software-defined electrified platforms. The approaches being taken vary greatly and it’s difficult to identify any one new thing in particular that the industry has embraced to radically change the design and development lifecycle.
What more can be done to optimise ICEs?
All OEMs have the opportunity to adapt or make ICE systems more efficient through the design and development process using advanced technologies, like machine learning. There’s quite a bit of inefficiency embedded in current legacy hardware processes that can be wrung out with the right machine learning. Secondmind uses a technique called Active Learning, which is also the name of our platform. With Active Learning, automotive engineers across the ecosystem can radically reduce data dependencies and experimentation time at the calibration phase of engine development, for example. This approach can also result in a reduction in the use of raw materials and the costs that come with the manufacturing of prototypes, which are often damaged using standard calibration processes and tools.
OEMs have the opportunity to adapt or make ICE systems more efficient through the design and development process using advanced technologies, like machine learning
What technology could be employed to prolong the life of ICEs as harsher legislation is put in place?
One of the things we’re focused on is optimising ICE while it is still a necessary component in passenger vehicles, not extending the life of these complex systems any longer than necessary. OEMs with ‘ICE vehicles’ in their portfolios, which will be many in number through 2035, must ensure environmental and business sustainability in the face of increasing design complexity resulting from tighter emissions regulations—a problem compounded by variations in regulations across different regions.
The complexity in the design of ICE and hybrid components is off the charts, and legacy processes and tools for calibrating and building an engine buckle under the pressure. Current machine learning techniques, which have made autonomous driving and ADAS a reality, struggle when it comes to data efficiency, which is critical for tackling high-dimensional engine design and calibration problems. We made a conscious decision to focus on a problem fewer companies are addressing—cleaning up what’s there while also helping the automotive industry cope with the even greater engineering complexity of the software defined car.
What role could machine learning play?
Practical machine learning will help automotive engineers build better cars, faster, with trust and transparency being key to accelerated adoption—something we recognised and have designed for since day one. Our Active Learning platform can intelligently automate the engine and EV motor calibration process and it does so in collaboration with the expert—the engineer. We believe the best outcomes will result from the blend of the domain expertise of the engineer with the intelligence of our algorithms.
Is it not too late for ICEs to be zero-emissions as manufacturers shift towards EV development?
There’s a fair amount of experimentation in biofuels and hydrogen, but these are not areas where we’ve spent as much time or R&D. Our goal is to minimise the impact of ICE today, and helping our customers create efficiencies in their businesses that can fuel investment in EV, hybrid and other strategic CASE applications, to get to the other side of the green transition faster.
Where do you see the biggest challenges with this technology, and what’s Secondmind’s industry outlook for the next decade?
On a macro level, the inflationary, post-pandemic environment that we’re in presents considerable uncertainty and it also makes it more challenging to raise capital at the same valuations as a year ago when money was more or less free to borrow. Like any start-up with technology capable of addressing problems in multiple industries, we have to ensure we don’t spread ourselves too thin. The key for us is to grow our business as fast as possible, while being disciplined and laser focused on the problems we are good at solving in an industry that needs us most—automotive.
The complexity in the design of ICE and hybrid components is off the charts, and legacy processes and tools for calibrating and building an engine buckle under the pressure
As CEO, what is your strategy for ensuring ICEs remain profitable?
My primary goal is not to make ICE profitable in isolation, but rather to ensure that my customers have the tools they need to drive design and development efficiencies in their ICE portfolios that can result in increased investment in EV, greener tech, and surviving the tidal wave of engineering complexity that awaits in the software defined car with the help of our Active Learning Platform.
What is your outlook for ICE development over the next couple of years?
Technology like ours creates a rising technology tide, that is raising the boats of a few key partners today and has the potential to raise all boats across the industry in the relatively near future. If we do our job right, this technology will help minimise the investment needed to deal with today’s complexities in ICE and make the transition to zero emissions mobility faster than analysts are currently projecting.
Where do you see the company developing over the next five to 10 years?
The approach we use today for calibrating engines in design and development will be critical for optimising the performance of engines and other complex systems throughout the lifecycle of the vehicle. The versatility of our Active Learning technology is also proving to be highly efficient in optimising the design of complex systems such as hybrids early in the design and simulation phase of development. With the data efficiency of our platform and its ability to virtualise the design and calibration of complex systems and components, we can help OEMs develop highly precise prototypes in less time and with little rework later in the development cycle.
100% virtualisation is the ultimate goal.
Keyword: Can machine learning clean up the last days of ICE?