Interview with Eric Link, Siemens Industry Software
“Fast charging and its infrastructure will be crucial in Europe”
Eric Link holds an M.Sc. in Sustainable Energy Systems in Mechanical Engineering from HAW Hamburg and a B.Eng. in Mechanical Engineering from Karlsruhe University of Applied Sciences.
Marc Jaspers
Battery development is becoming more complex as OEMs balance performance, safety, ageing and charging requirements. Eric Link of Siemens Industry Software explains how 3D, 1D and AI methods can be combined in simulation workflows.
Battery simulation increasingly has to cover very different levels of detail – from high-fidelity 3D models to system-level simulation and AI-assisted methods. Eric Link, Technical Account Manager at Siemens Industry Software, has worked for the company for 12 years and focuses on simulation workflows in engineering.
At the Automotive Battery Conference 2026, he will present the keynote “Battery Simulation Workflow – 3D, 1D and AI Methods”, addressing model interoperability, the transition from 3D CFD to 1D simulation and the use of AI to accelerate studies. Ahead of the event, we spoke with him about where simulation can reduce development effort, where physical fidelity remains essential and which challenges still stand in the way of continuous simulation workflows.
The Automotive Battery Conference 2026 agenda spans resilience, scalable manufacturing, battery safety, lifecycle value and charging. Which issue will most decisively shape the battery and e-mobility industry over the next five years, and why?
In my opinion, coming from a development background, I believe lifecycle value as well as charging will become increasingly important in the coming years, especially in Europe. Lifecycle value is largely determined in the early development phases. This means engineering the whole lifecycle of a battery will be an additional design challenge that has to be addressed. With increasing regulatory requirements, such as recyclability, the maximum use of the battery has to be taken into account in order to design a valuable product for the automotive aftermarket as well as the second-life market for batteries. For this, durability, high cycle numbers and residual capacity are design parameters that have to be taken into account at an early stage.
Simulation plays a major role here in enabling engineers to predict battery behaviour over the full lifecycle. Fast charging and its infrastructure will be crucial for adoption in many European countries that currently have a small EV footprint in their fleets. Fast charging cannot remain a luxury feature. New cell chemistries promise to allow fast charging in mid-budget cars. In my opinion, simulation again plays a major role from an engineering point of view. Finding optimal charging strategies with minimal impact on ageing factors, such as lithium plating, under all operating conditions is a perfect use case for a large simulation and optimisation study requiring detailed cell and thermal system models.
Your agenda explicitly names exchangeable model formats. What is the bigger obstacle to a continuous 3D/1D/AI workflow today: format interoperability, IP protection, parameter governance or ownership across engineering teams?
Format interoperability is widely covered by existing exchange formats such as FMUs. There is almost no commercial solver that does not support FMUs anymore. The tricky part is preparing the model so that it can run in different environments from a numerical point of view. And this touches on all the other aspects you mentioned. Take parameter governance, for example: a parameter in an exporting environment can be a continuous signal, while on the importing side it could be a discrete signal. This may seem like a small issue, but it can easily lead to failures.
So clearly defining all parameters and their boundaries is key. IP protection is also a major issue as soon as models cross company boundaries. Troubleshooting in a black box is almost impossible. So defining the model's fidelity upfront is important and should be an integral part of the process before exchanging models. Ownership across engineering teams is the most obvious, but sometimes also the most complex issue. Especially in large organisations, I often see different simulation departments working in a very isolated way from each other. They are treated as individual “verification services” rather than as a unified simulation workforce, leading to incompatible methodologies and rework of modelling activities.
In a battery simulation workflow spanning detailed 3D CFD, reduced-order models and 1D system simulation, where should engineers deliberately retain high physical fidelity and where is simplification usually safe?
This highly depends on the use case. But I think it is safe to say that 3D CFD can be the most accurate of those mentioned. However, it is up to the engineer to decide where fidelity and accuracy are needed and where speed and wider system boundaries are beneficial. A good example is the thermal behaviour of a cell. It is very important to understand where hot spots can form in order to optimise the cooling system. This is a typical 3D CFD task requiring a high level of detail. Once the hot spots are known, it can be good practice to reduce the order of the model in order to simulate the whole pack over a long drive or charging cycle.
Another example is early design variations. A detailed model does not necessarily have to be correct. A detailed model requires a lot of detailed parameter information. If you do not have this information at an early stage, your detailed model might be just as “wrong” as a lower-fidelity model. The lower-fidelity model, however, would allow you to move faster through design variations while understanding the qualitative implications of the changes.
What information is most often lost when models move between 3D and system-level environments, and how do you preserve boundary conditions, parameter definitions and model validity across different tools and engineering teams?
Again, this highly depends on the use case. And I would not say that the information is lost; the engineer decides what kind of information is relevant for the given use case. Take an electrochemical 3D CFD cell model that has been built for lithium plating estimation under fast-charging conditions. This model serves a very specific purpose. If you are now tasked with running a simulation to determine the potential range of the vehicle under different ambient conditions, you do not need this level of detail. So you could use your detailed model to characterise a simplified version of the cell that is better suited for this purpose.
AI-assisted simulation acceleration is on your agenda. What guardrails are needed to detect when a surrogate model leaves its valid training space before a plausible but physically wrong result influences a safety-critical design decision?
In my opinion, this is a mix of engineering experience and tool guidance. An engineer needs to know the intended operating range of the model. In the end, a surrogate model can only work within valid boundaries. Hence, choosing the right training and validation data is key. Your training data should cover the range of operating conditions, while your validation data needs to be selected in order to challenge your surrogate model within these boundaries. Good surrogate or model order reduction tools support engineers in choosing the right amount and type of training data.
With experimental battery-safety work also on the programme, what level of test correlation, uncertainty quantification and model traceability is required before simulation can replace physical design iterations rather than merely complement them?
That is a very important question. I firmly believe that simulation will not fully replace physical testing in the foreseeable future. However, simulation will be absolutely necessary to keep the amount of physical testing and prototyping acceptable and efficient. Physical testing efforts should be focused on homologation for relevant checks, while other development milestones could be carried out virtually. I believe model traceability is one of the biggest challenges.
Compared with, for example, mechanical design data, simulation models are relatively ungoverned within organisations. Without a governance layer that allows simulation engineers to have reliable and traceable input resources for their simulations, while at the same time working in a managed environment in which all changes can be assessed by the affected parties, it will be difficult to rely solely on simulations for virtual verification. Processes and tools need to come together in a verification environment that protects both the process and the engineer.
Which other agenda topic intersects most with your own, and what would you ask that speaker or panel?
For sure, “eBus in the Fast Lane: MAN’s Successful Transition from Diesel to Battery Power” from Fabian Schöffel. I would be very interested to learn whether and how testing and simulation were orchestrated in a joint workflow.
