Coupling numerical simulation and artificial intelligence prediction: A computational fluid dynamics–discrete element method and deep learning approach to gas–solid flows DOI
Haolei Zhang, Ji Xu, Li Guo

et al.

Physics of Fluids, Journal Year: 2024, Volume and Issue: 37(5)

Published: May 1, 2024

With the development of computing technology and numerical methods, a coupling discrete element method (DEM) with computational fluid dynamics (CFD) is becoming an effective simulation for deeper understanding large-scale gas–solid flow systems. Traditionally, cost DEM part usually much higher than CFD due to explicitly tracking more solid particles grids. However, coarse-graining high-performance computing, actually wall-time bottleneck in many cases. To mitigate this problem, deep-learning-based fluid-dynamics prediction (FPM) proposed replace or accelerate CFD. Accurate CFD–DEM results provide dataset deep learning using artificial neural network (ANN) UNet architecture, time-series local, neighboring, global information consider its critical spatiotemporal heterogeneity feature. The FPM can be incorporated into (FPM–CFD–DEM) simply (FPM–DEM). Simulations have demonstrated that FPM–DEM 10-fold faster original under similar accuracy, while FPM–CFD–DEM only double speed but predict reasonable longer time. Future work may exploit remarkable potential intelligence techniques developing efficient methods other multiphase system.

Language: Английский

A mPOD-based Reduced-order Modelling Approach for Fast Gas-solid Flow Simulations DOI Creative Commons
Huiting Chen, Wangyan Li, Jie Bao

et al.

Chemical Engineering Science, Journal Year: 2025, Volume and Issue: unknown, P. 121155 - 121155

Published: Jan. 1, 2025

Language: Английский

Citations

1
A LSTM-enhanced surrogate model to simulate the dynamics of particle-laden fluid systems DOI

Arash Hajisharifi,

Rahul Halder, Michele Girfoglio

et al.

Computers & Fluids, Journal Year: 2024, Volume and Issue: 280, P. 106361 - 106361

Published: July 6, 2024

Language: Английский

Citations

3
A data-driven method for fast predicting the long-term hydrodynamics of gas–solid flows: Optimized dynamic mode decomposition with control DOI
Dandan Li, Bidan Zhao, Shuai Lu

et al.

Physics of Fluids, Journal Year: 2024, Volume and Issue: 36(10)

Published: Oct. 1, 2024

Data-driven methods are of great interest in studying the hydrodynamics gas–solid flows. In this paper, we developed an optimized dynamic mode decomposition with control (DMDc) method for long-term and fast prediction one physical field aid another field. Using computational fluid dynamics-discrete element (CFD-DEM) simulation results as benchmark, ability standard DMDc is evaluated. It was shown that superior when order magnitude predicted data much larger than auxiliary data, which cannot be addressed by using scaled or dimensionless instance, gas pressure solid volume fraction; on other hand, both can reasonably predict behavior flows, elements comparative to This study proposes a relatively accurate predicting flows known

Language: Английский

Citations

3
Enhanced Long-Time prediction of Gas-Solid flows via Integrating Reduced-Order model with Cluster-Based network model DOI
Tingting Liu, Yu Jiang, Yuanye Zhou

et al.

Chemical Engineering Science, Journal Year: 2025, Volume and Issue: unknown, P. 121634 - 121634

Published: April 1, 2025

Language: Английский

Citations

0
Advanced graph neural network-based surrogate model for granular flows in arbitrarily shaped domains DOI Creative Commons
Shuo Li, Mikio Sakai

Chemical Engineering Journal, Journal Year: 2024, Volume and Issue: 500, P. 157349 - 157349

Published: Nov. 1, 2024

Language: Английский

Citations

3
Coupling numerical simulation and artificial intelligence prediction: A computational fluid dynamics–discrete element method and deep learning approach to gas–solid flows DOI
Haolei Zhang, Ji Xu, Li Guo

et al.

Physics of Fluids, Journal Year: 2024, Volume and Issue: 37(5)

Published: May 1, 2024

With the development of computing technology and numerical methods, a coupling discrete element method (DEM) with computational fluid dynamics (CFD) is becoming an effective simulation for deeper understanding large-scale gas–solid flow systems. Traditionally, cost DEM part usually much higher than CFD due to explicitly tracking more solid particles grids. However, coarse-graining high-performance computing, actually wall-time bottleneck in many cases. To mitigate this problem, deep-learning-based fluid-dynamics prediction (FPM) proposed replace or accelerate CFD. Accurate CFD–DEM results provide dataset deep learning using artificial neural network (ANN) UNet architecture, time-series local, neighboring, global information consider its critical spatiotemporal heterogeneity feature. The FPM can be incorporated into (FPM–CFD–DEM) simply (FPM–DEM). Simulations have demonstrated that FPM–DEM 10-fold faster original under similar accuracy, while FPM–CFD–DEM only double speed but predict reasonable longer time. Future work may exploit remarkable potential intelligence techniques developing efficient methods other multiphase system.

Language: Английский

Citations

1