Publication
Exploring model complexity: A model comparison
Pascal Marijan; Sebastian Igel; Tatjana Legler; Martin Ruskowski
In: AIP Conference Proceedings, Vol. 3381, No. 1, Page 020001, API Publishing, 3/2026.
Abstract
System identification involves describing processes based on observed data, with model accuracy heavily reliant on data quality. In industrial material processing, data acquisition can be challenging, making it crucial to choose the right model complexity. This study conducts a model comparison to evaluate the performance of models with varying complexities in the context of a vertical roller mill (VRM). The analysis includes both stationary operation points and transient data areas of the VRM. Given the constraints of limited sensors and low-frequency data, this study seeks to determine which type of model best captures the system’s behavior in each state. The results provide insights into the balance between model complexity and performance in data-limited scenarios offering a framework for the selection of suitable models, in this case regarding the optimization of energy consumption in industrial grinding processes.