# American Institute of Mathematical Sciences

doi: 10.3934/era.2020128

## Efficient computations for linear feedback control problems for target velocity matching of Navier-Stokes flows via POD and LSTM-ROM

 Department of Mathematics, Ajou University, Suwon, 16499, Republic of Korea

* Corresponding author

Received  August 2020 Revised  September 2020 Published  December 2020

Fund Project: The author is supported by Basic Science Research Program through the National Research Foundation of Korea NRF-2016R1D1A1B03932219 and NRF-2019R1F1A1050231

An efficient computing method for a target velocity tracking problem of fluid flows is considered. We first adopts the Lagrange multipliers method to obtain the optimality system, and then designs a simple and effective feedback control law based on the relationship between the control ${{\boldsymbol f}}$ and the adjoint variable ${{\boldsymbol w}}$ in the optimality system. We consider a reduced order modeling (ROM) of this problem for real-time computing. In order to improve the existing ROM method, the deep learning technique, which is currently being actively researched, is applied. We review previous research results and some computational results are presented.

Citation: Hyung-Chun Lee. Efficient computations for linear feedback control problems for target velocity matching of Navier-Stokes flows via POD and LSTM-ROM. Electronic Research Archive, doi: 10.3934/era.2020128
##### References:

show all references

##### References:
] (left) and our results (right) where $Re = 1$">Figure 3.  $|| {\boldsymbol u}(t) - {\boldsymbol U}(t) ||_2$, Compare between Figure 1 in [19] (left) and our results (right) where $Re = 1$
Discrepancy of the trajectories between the FOM, ROM and least squares projected solutions after Galerkin projection
Controlled (left) and target (right) flows at $t = 0.0, 0.050,0.057,0.059$ (left pair of columns) and $t = 0.060, 0.061,$ $0.064, 0.5$ (right pair of columns) for $\gamma = 10$
$|| {\boldsymbol u} - {\boldsymbol U}||_{{\bf L}^2({\Omega})}$ for different control values $\gamma$
The POD reduced basis of cardinality 8 ($\gamma = 10$ Upper two rows and $\gamma = 40$ Down two rows). Basis 1, 2, 3, 4 from the left to the right (Top) and Basis 5, 6, 7, 8 from the left to the right (Bottom) in each two rows
The singular values of the snapshot data matrix (right) for different control values $\gamma$
RMSE of Galerkin ROM (Upper 4 lines) and LS Projection (Lower 4 lines)
LSTM Cell (left) and LSTM Training where $r = 5$
Training and validation loss
Temporal evolution of the 8 POD modal coefficients predicted by LS Projection, GP-ROM and GP-LSTM for $\gamma = 50$ using the basis from $\gamma = 40$
RMSE of ROMs predicted by LS Projection, GP-ROM and GP-LSTM for $\gamma = 50$ using the basis of $\gamma = 40$
The first 8 singular values of the snapshot matrix
 singular vales RIC($\gamma=40$) singular vales RIC($\gamma=40$) 1 4.83907e+02 99.2476% 2 2.41493e+00 99.8174% 3 3.98155e-01 99.9207% 4 2.02669e-01 99.9731% 5 6.01101e-02 99.9909% 6 1.65390e-02 99.9964% 7 6.20453e-03 99.9985% 8 2.51213e-03 99.9994%
 singular vales RIC($\gamma=40$) singular vales RIC($\gamma=40$) 1 4.83907e+02 99.2476% 2 2.41493e+00 99.8174% 3 3.98155e-01 99.9207% 4 2.02669e-01 99.9731% 5 6.01101e-02 99.9909% 6 1.65390e-02 99.9964% 7 6.20453e-03 99.9985% 8 2.51213e-03 99.9994%
A list of hyperparameters utilized to train the LSTM network for numerical experiments
 Variables Hyperparameters Number of hidden layers 2 Number of neurons in each hidden layer 120 Number of lookbacks 5 Batch size 32 Epochs 1000 Activation functions in the LSTM layers tanh Validation data set 20% Loss function MSE Optimizer ADAM
 Variables Hyperparameters Number of hidden layers 2 Number of neurons in each hidden layer 120 Number of lookbacks 5 Batch size 32 Epochs 1000 Activation functions in the LSTM layers tanh Validation data set 20% Loss function MSE Optimizer ADAM
CPU time (in second) comparison for the different ROM frameworks investigated in this study and RMSE at $t = 0.25$
 Framework Times RMSE(0.25) FOM $1695.7$ 0 LS-Proj 1.34053e-05 GP-ROM(8) 0.3268 3.77711e-03 GP-ROM (16) 1.7012 1.28362e-03 GP-LSTM(8) 0.6372 3.20970e-04
 Framework Times RMSE(0.25) FOM $1695.7$ 0 LS-Proj 1.34053e-05 GP-ROM(8) 0.3268 3.77711e-03 GP-ROM (16) 1.7012 1.28362e-03 GP-LSTM(8) 0.6372 3.20970e-04
 [1] Yueqiang Shang, Qihui Zhang. A subgrid stabilizing postprocessed mixed finite element method for the time-dependent Navier-Stokes equations. Discrete & Continuous Dynamical Systems - B, 2021, 26 (6) : 3119-3142. doi: 10.3934/dcdsb.2020222 [2] Derrick Jones, Xu Zhang. A conforming-nonconforming mixed immersed finite element method for unsteady Stokes equations with moving interfaces. Electronic Research Archive, , () : -. doi: 10.3934/era.2021032 [3] Daoyuan Fang, Ting Zhang. Compressible Navier-Stokes equations with vacuum state in one dimension. Communications on Pure & Applied Analysis, 2004, 3 (4) : 675-694. doi: 10.3934/cpaa.2004.3.675 [4] Thomas Y. Hou, Ruo Li. Nonexistence of locally self-similar blow-up for the 3D incompressible Navier-Stokes equations. Discrete & Continuous Dynamical Systems, 2007, 18 (4) : 637-642. doi: 10.3934/dcds.2007.18.637 [5] Cheng Wang. Convergence analysis of Fourier pseudo-spectral schemes for three-dimensional incompressible Navier-Stokes equations. Electronic Research Archive, , () : -. doi: 10.3934/era.2021019 [6] Xin-Guang Yang, Rong-Nian Wang, Xingjie Yan, Alain Miranville. Dynamics of the 2D Navier-Stokes equations with sublinear operators in Lipschitz-like domains. Discrete & Continuous Dynamical Systems, 2021, 41 (7) : 3343-3366. doi: 10.3934/dcds.2020408 [7] Ling-Bing He, Li Xu. On the compressible Navier-Stokes equations in the whole space: From non-isentropic flow to isentropic flow. Discrete & Continuous Dynamical Systems, 2021, 41 (7) : 3489-3530. doi: 10.3934/dcds.2021005 [8] Xin Zhong. Global strong solution and exponential decay for nonhomogeneous Navier-Stokes and magnetohydrodynamic equations. Discrete & Continuous Dynamical Systems - B, 2021, 26 (7) : 3563-3578. doi: 10.3934/dcdsb.2020246 [9] Mirela Kohr, Sergey E. Mikhailov, Wolfgang L. Wendland. Dirichlet and transmission problems for anisotropic stokes and Navier-Stokes systems with L∞ tensor coefficient under relaxed ellipticity condition. Discrete & Continuous Dynamical Systems, 2021  doi: 10.3934/dcds.2021042 [10] Oleksandr Boichuk, Victor Feruk. Boundary-value problems for weakly singular integral equations. Discrete & Continuous Dynamical Systems - B, 2021  doi: 10.3934/dcdsb.2021094 [11] Francis Hounkpe, Gregory Seregin. An approximation of forward self-similar solutions to the 3D Navier-Stokes system. Discrete & Continuous Dynamical Systems, 2021  doi: 10.3934/dcds.2021059 [12] Lunji Song, Wenya Qi, Kaifang Liu, Qingxian Gu. A new over-penalized weak galerkin finite element method. Part Ⅱ: Elliptic interface problems. Discrete & Continuous Dynamical Systems - B, 2021, 26 (5) : 2581-2598. doi: 10.3934/dcdsb.2020196 [13] Huancheng Yao, Haiyan Yin, Changjiang Zhu. Stability of rarefaction wave for the compressible non-isentropic Navier-Stokes-Maxwell equations. Communications on Pure & Applied Analysis, 2021, 20 (3) : 1297-1317. doi: 10.3934/cpaa.2021021 [14] Rui Hu, Yuan Yuan. Stability, bifurcation analysis in a neural network model with delay and diffusion. Conference Publications, 2009, 2009 (Special) : 367-376. doi: 10.3934/proc.2009.2009.367 [15] Marcel Braukhoff, Ansgar Jüngel. Entropy-dissipating finite-difference schemes for nonlinear fourth-order parabolic equations. Discrete & Continuous Dynamical Systems - B, 2021, 26 (6) : 3335-3355. doi: 10.3934/dcdsb.2020234 [16] Armin Lechleiter, Tobias Rienmüller. Factorization method for the inverse Stokes problem. Inverse Problems & Imaging, 2013, 7 (4) : 1271-1293. doi: 10.3934/ipi.2013.7.1271 [17] Tao Wu, Yu Lei, Jiao Shi, Maoguo Gong. An evolutionary multiobjective method for low-rank and sparse matrix decomposition. Big Data & Information Analytics, 2017, 2 (1) : 23-37. doi: 10.3934/bdia.2017006 [18] Reza Lotfi, Yahia Zare Mehrjerdi, Mir Saman Pishvaee, Ahmad Sadeghieh, Gerhard-Wilhelm Weber. A robust optimization model for sustainable and resilient closed-loop supply chain network design considering conditional value at risk. Numerical Algebra, Control & Optimization, 2021, 11 (2) : 221-253. doi: 10.3934/naco.2020023 [19] Zengyun Wang, Jinde Cao, Zuowei Cai, Lihong Huang. Finite-time stability of impulsive differential inclusion: Applications to discontinuous impulsive neural networks. Discrete & Continuous Dynamical Systems - B, 2021, 26 (5) : 2677-2692. doi: 10.3934/dcdsb.2020200 [20] Xiaomao Deng, Xiao-Chuan Cai, Jun Zou. A parallel space-time domain decomposition method for unsteady source inversion problems. Inverse Problems & Imaging, 2015, 9 (4) : 1069-1091. doi: 10.3934/ipi.2015.9.1069

Impact Factor: 0.263

## Tools

Article outline

Figures and Tables