Categorizing Simulation Models Using Convolutional Neural Networks 2023-01-1217
Whether as an optimization problem or a development tool, neural networks help engineers to work more efficiently. This paper’s central aspect is to add metadata to the core files of the project simulation data. To understand the project and its simulation models, a pre-processing methodology and convolutional neural network architecture are presented. With the added labels, it is possible to access the content of the model files of an engine performance simulation tool without examining them. At first, a pre-processing approach and its design are introduced to extract and filter the desired data from the XML data structure. Then, the data is split into sequences and paired with labels. Expert knowledge is used to label the models. These labels are further paired with the extracted sequences. In addition, a convolutional neural network design with a single convolutional layer and three dense layers is presented to add the defined labels that characterize the powertrain architecture to these sequences. Furthermore, this paper shows the advantages and disadvantages of the pre-processing methodology and the architecture of the convolutional neural network. The results show that the model can categorize the extracted sequences with a very good accuracy. At last, suggestions for improvement for the pre-processing and additional studies are presented.
Citation: Grbavac, A., Angerbauer, M., Grill, M., Itzen, D. et al., "Categorizing Simulation Models Using Convolutional Neural Networks," SAE Technical Paper 2023-01-1217, 2023, https://doi.org/10.4271/2023-01-1217. Download Citation
Author(s):
Andrija Grbavac, Martin Angerbauer, Michael Grill, Dirk Itzen, Sasa Milojevic, Timo Hagenbucher, André Kulzer
Affiliated:
FKFS, IFS, University of Stuttgart
Pages: 8
Event:
23rd Stuttgart International Symposium
ISSN:
0148-7191
e-ISSN:
2688-3627
Related Topics:
Neural networks
Simulation and modeling
Architecture
Optimization
Engines
Powertrains
SAE MOBILUS
Subscribers can view annotate, and download all of SAE's content.
Learn More »