How crucial are the innovative methods used for model-based DoE in accelerating and refining the battery testing process? Could you explain the significance of intelligent DoE in ensuring efficient and meaningful testing by avoiding redundant experiments?
Let me give you an example on this. Some common questions that we address with model-based Design on Experiments (DoE) could be the following: What kind of experiments need to be conducted in order to improve our knowledge of a specific parameter of the tested battery. By naively running many simulations or physical experiments on this battery, we might obtain an answer on this question but also wasted a lot of time and money because of redundancies in our virtual and physical experiments. Intelligent DoE aims at speeding up this process, reducing redundancies and simultaneously discovering those experimental designs that are optimal in terms of time and costs.
How do these methodologies align with the requirements derived from battery testing in different use cases?
The different use cases (energy storage, automotive, off-road mobile) in FASTEST have very distinct characteristics in the way the battery is being used. This includes their charging and discharging patterns but also operating conditions like the temperature at which the systems may be running. Especially in terms of battery ageing, such details can have a huge impact on the outcome and the performance of the underlying models. Therefore, we focus on DoE methodologies that are robust and able to provide good experimental designs for all these different scenarios. This is also an important prerequisite for automatic task distribution.
Considering the development of concepts for automatic task distribution: How do these concepts aim to integrate physical and virtual testing efficiently?
While model-based DoE can be a powerful tool to reduce time and costs in battery testing, we still need some level of physical testing to validate that the virtual models remain a reliable description of reality. Concepts for automatic task distributions aim at finding the sweet spot between sound experimental validation and more efficient virtual testing. For this purpose, we develop tools that help us in deciding and quantifying to which extent we can trust the outputs of our models. If there is some uncertainty such tasks need to be distributed automatically to the physical testbench.