Browsing by Author "Röder, Benedict"

Now showing 1 - 2 of 2

Open Access
Hybrid modeling of multibody systems : comparison of two discrepancy models for trajectory prediction
(2024) Wohlleben, Meike; Röder, Benedict; Ebel, Henrik; Muth, Lars; Sextro, Walter; Eberhard, Peter
This study focuses on hybrid modeling approaches that combine physical and data‐driven methods to create more effective dynamical system models. In particular, it examines discrepancy models, a type of hybrid model that integrates a physical system model with data‐driven compensation for inaccuracies. The study applies two discrepancy modeling methods to a multibody system using discrepancies in the state vector and its time derivative, respectively. As an application example, a four‐bar linkage with nonlinear damping is investigated, using a simplified conservative system as a physical model. The comparative analysis of the two methods shows that the continuous approach generally outperforms the discrete method in terms of accuracy and computational efficiency, especially for velocity prediction and prediction horizon. However, scenarios, where input signals for training and testing differ, present nuanced findings. When the continuous method is trained on complex signals (sine) and tested on simpler ones (stair), it struggles to deliver satisfactory results, exhibiting notably higher root mean square error (RMSE) values, particularly in angular velocity prediction. Conversely, training on simple signals (stair) and testing on complex ones (sine) surprisingly yields low RMSE values, indicating the continuous method's adaptability. While the discrete method aligns more closely with expectations and performs better in certain scenarios, its results are consistently moderate, neither exceptional nor particularly poor. The study also introduces a selection framework for choosing the most suitable algorithm based on the specific characteristics of the modeling task. This framework provides guidance for researchers and practitioners in leveraging hybrid modeling effectively. Finally, the study concludes with an outlook on future research directions.
Open Access
Towards intelligent design assistants for planar multibody mechanisms
(2023) Röder, Benedict; Ebel, Henrik; Eberhard, Peter
General‐purpose mechanisms can perform a broad range of tasks but are usually rather heavy and expensive. If only particular movements need to be executed, more efficient special‐purpose mechanisms can be employed. However, they typically require an expert to design the system based on manual inspection of simulations and experimental results. This procedure is not only time‐consuming, but the outcome also depends on the expert's experience. Hence, the design process stems from subjective criteria while only a limited number of structurally different mechanisms can be considered. In contrast, a design assistant can consider a broad range of mechanisms and leverage multi‐objective optimization to retrieve optimal designs for the given task. Due to the systems being synthesized based on mathematical functions rather than individual experience, the assistant allows a more transparent development of optimal problem‐specific mechanisms compared to the conventional process. Experts can then fine‐tune and analyze the proposed designs to compose the final system. In recent years, neural networks have been utilized to directly learn the inverse mapping from a trajectory to a mechanism design. This requires some parameterization of the trajectory to be fed into the network. In this work, we evaluate various preprocessing methods for the trajectory on a simple mechanism design model problem. We assess multiple configurations such as different neural network sizes, applying input‐output normalization, and varying the number of features. Consequently, we investigate and compare the trends and robustness of the implemented methods.