Presentation Information
[1H02]Frequency-Based Reduced-Order Models for Gas-Solid Flows with Heat Transfer
*Michael Tan Castro1, Shuo Li1, Hiroki Imai1, Kai-en Yang1, Toshiki Imatani1, Mikio Sakai1 (1. UTokyo)
Keywords:
discrete element method,computational fluid dynamics,proper orthogonal decomposition,non-intrusive reduced-order model,heat transfer
Reduced-order models (ROM) based on proper orthogonal decomposition (POD) have been developed for accelerating CFD-DEM simulations of gas-solid flows. However, the irregularity of gas-solid flows necessitates the use of complex and elaborate machine learning models for this purpose, and these can be difficult to fine-tune as the number of hyperparameters and combinations thereof grows large. In this study, we introduce the frequency-based ROM, which replicates gas-solid flow dynamics by extracting the dominant frequencies from POD coefficients. This approach involves minimal fine-tuning and does not incur the large training time of machine learning methods. Our ROMs can also replicate the spatiotemporal characteristics of the FOM while accelerating simulations by several orders of magnitude.