Simulation has been a key technology of the automotive industry for many decades. Predictive simulations enable engineers and designers to understand how parts of a vehicle, or even entire vehicles, will operate under a multitude of conditions before ever building them physically. This has proven especially useful in the cases of complex issues such as external aerodynamic wind tunnel or crashworthiness tests. Why expend time and physical resources performing these difficult assessments when they can be completed much faster virtually?
Vehicles being designed today, however, are remarkably different and more complex than vehicles designed, say, a decade ago. Whereas past vehicles traditionally have been mechanically centric products, today’s vehicles are multi-physics products in numerous ways. Many models are replacing internal combustion engines with electric batteries, as well as becoming software-defined vehicles, incorporating software, sensors, and electronics that gather real-time data across the vehicle.
These electrically powered vehicles generate heat, requiring new cooling systems to prevent the batteries from overheating. The amount of new physics to be calculated from these technologies can be challenging for traditional simulation software, risking the creation of bottlenecks in the design process.
Yet there is another technology that has also been on the rise lately that has the potential to alleviate these bottlenecks, and that technology is artificial intelligence (AI). AI can work at incredible speeds, generating results in seconds instead of hours or days, but as AI is still in its infancy, this may be at the cost of sacrificing accuracy initially. Fortunately, combining AI with simulation can help achieve the best of both worlds. By leveraging rich simulation data to train AI models, which can then be used to quickly explore a design space and guide further simulations, the existing product design process can be enhanced to become faster and more flexible.
The State of AI in Automotive
At the time of writing this article, the footprint of AI and its influence in automotive design is small but growing. After all, AI is still rather new, and trying to apply it to situations it is not trained in is going to come with its own challenges, such as the risk of reduced accuracy.
If there is one thing AI is good at, however, it is how to learn quickly from historical data. Engineers can educate and train AI using data from vehicles and their simulations, enhancing the AI’s knowledge over time. Eventually the AI can learn enough to capture key aspects of previous vehicle designs and usage to produce more accurate simulations, enhancing the design process for the next generation. AI’s influence in automotive design and simulation is quite likely to increase with each new generation of vehicle.
Why AI?
While some might question the value of using AI while it is still in its infancy, there are plenty of reasons to begin using AI in automotive simulations now. In addition to training it early to increase its expertise down the line, current AI models can still aid engineers and designers by accelerating the completion of complex simulations.
Consider the largest, time-consuming obstacles of simulation today: the creation of the simulation model, and then the amount of time required to generate and compute data from the simulation. Now apply these to some of the most complex simulations in the automotive industry, such as the external aerodynamics and crashworthiness tests mentioned before. In crash tests, for example, engineers potentially can run many different crash simulations in the design space, resulting in a lot of data to process. The last thing engineers want to do is spend their days interrogating every single simulation and create another bottleneck.
AI can automate simple tasks such as identifying the most relevant simulation for further analysis to save engineers and designers immense amounts of time. For example, one thing AI has already proven to be good at is recognizing data patterns, which can be especially helpful with simulations that deal with complex geometry such as crash tests. The AI can take all the simulations generated in the design space and cluster them into distinct types of crash system behavior, narrowing the field of investigation and presenting key insights.
This way, engineers and designers can focus on key insights and skip the time-consuming process of sifting through volumes of data and, in turn, allow them to move to the next steps of the design process faster. Furthermore, the more AI evolves, the better it can handle physics and geometric data of higher complexity, expanding the kinds of simulations it can enhance.
Copilots You Can Trust
The examples described above are only the beginning. AI also has immense potential to become copilots and knowledge banks that can bolster engineers’ expertise and creativity.
There are two general areas these kinds of applications fall under. The first is developed as a copilot internal within a particular piece of software. This copilot can then become very well educated on that piece of software, becoming a valuable assistant to its users. Simulation tools are a prime target for copilots, and in companies where more experienced engineers are retiring, the copilots can help newly hired engineers become more familiar with the simulation software.
The other area of copilot application involves building the AI deeper into a software platform to the point of governing a whole process through an agentic workflow. Engineers can instruct these agents directly when they want to, for example, make a change to a particular part of their product’s structure. The AI can perform the change on its own (e.g. change the simulation model then run the simulation). This can accelerate the design process significantly, automating many of its mundane operations and giving engineers more time to explore and simulate different design options and build their creativity.
As previously mentioned, AI is still in its infancy, and the full extent of its capabilities are yet to be commonplace. What is certain, however, is that it is learning fast. AI technology is progressing alongside the automotive industry, becoming a powerful tool in the engineer’s toolbox that can help balance simulation software against the growing multi-physics complexity of today’s vehicles. By investing in AI early, companies can reap its rewards as it matures, accelerating simulation capabilities as well as the digitalization of the automotive industry.
This article was written by Royston Jones, Global Head of Automotive & Transportation, Siemens Digital Industries Software (Plano, TX). For more information, visit here .
