An extended physics informed neural network for preliminary analysis of parametric optimal control problems

In this work we propose an extension of physics informed supervised learning strategies to parametric partial differential equations. Indeed, even if the latter are indisputably useful in many applications, they can be computationally expensive most of all in a real-time and many-query setting. Thus, our main goal is to provide a physics informed learning paradigm to simulate parametrized phenomena in a small amount of time. The physics information will be exploited in many ways, in the loss function (standard physics informed neural networks), as an augmented input (extra feature employment) and as a guideline to build an effective structure for the neural network (physics informed architecture). These three aspects, combined together, will lead to a faster training phase and to a more accurate parametric prediction. The methodology has been tested for several equations and also in an optimal control framework.

[1]  Tomás Roubícek,et al.  Optimal control of Navier-Stokes equations by Oseen approximation , 2007, Comput. Math. Appl..

[2]  A. Quarteroni,et al.  A reduced computational and geometrical framework for inverse problems in hemodynamics , 2013, International journal for numerical methods in biomedical engineering.

[3]  G. Karniadakis,et al.  Conservative physics-informed neural networks on discrete domains for conservation laws: Applications to forward and inverse problems , 2020 .

[4]  Barak A. Pearlmutter,et al.  Automatic differentiation in machine learning: a survey , 2015, J. Mach. Learn. Res..

[5]  Gianluigi Rozza,et al.  Reduced order methods for parametrized non-linear and time dependent optimal flow control problems, towards applications in biomedical and environmental sciences , 2019, ENUMATH.

[6]  Mohammad Motamed.  A multi-fidelity neural network surrogate sampling method for uncertainty quantification , 2019, ArXiv.

[7]  G. Karniadakis,et al.  Physics-Informed Neural Networks for Heat Transfer Problems , 2021, Journal of Heat Transfer.

[8]  Gianluigi Rozza,et al.  Reduced order methods for parametric optimal flow control in coronary bypass grafts, toward patient‐specific data assimilation , 2019, International journal for numerical methods in biomedical engineering.

[9]  Paris Perdikaris,et al.  Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations , 2019, J. Comput. Phys..

[10]  Tsuyoshi Murata,et al.  {m , 1934, ACML.

[11]  Jan S. Hesthaven,et al.  Multi-fidelity regression using artificial neural networks: efficient approximation of parameter-dependent output quantities , 2021, Computer Methods in Applied Mechanics and Engineering.

[12]  George Em Karniadakis,et al.  Locally adaptive activation functions with slope recovery for deep and physics-informed neural networks , 2020, Proceedings of the Royal Society A.

[13]  George Em Karniadakis,et al.  Learning nonlinear operators via DeepONet based on the universal approximation theorem of operators , 2019, Nature Machine Intelligence.

[14]  Gianluigi Rozza,et al.  Reduced basis approximation of parametrized optimal flow control problems for the Stokes equations , 2015, Comput. Math. Appl..

[15]  Andreas Griewank,et al.  Trends in PDE Constrained Optimization , 2014 .

[16]  Liu Yang,et al.  B-PINNs: Bayesian Physics-Informed Neural Networks for Forward and Inverse PDE Problems with Noisy Data , 2020, J. Comput. Phys..

[17]  A. Quarteroni,et al.  Numerical modeling of hemodynamics scenarios of patient-specific coronary artery bypass grafts , 2017, Biomechanics and Modeling in Mechanobiology.

[18]  Annalisa Quaini,et al.  Numerical Approximation of a Control Problem for Advection-Diffusion Processes , 2005, System Modelling and Optimization.

[19]  D. Rovas,et al.  Reliable Real-Time Solution of Parametrized Partial Differential Equations: Reduced-Basis Output Bound Methods , 2002 .

[20]  Paris Perdikaris,et al.  Learning the solution operator of parametric partial differential equations with physics-informed DeepONets , 2021, Science advances.

[21]  George Em Karniadakis,et al.  hp-VPINNs: Variational Physics-Informed Neural Networks With Domain Decomposition , 2020, Computer Methods in Applied Mechanics and Engineering.

[22]  George Em Karniadakis,et al.  Hidden fluid mechanics: Learning velocity and pressure fields from flow visualizations , 2020, Science.

[23]  Yi Luo,et al.  All-optical machine learning using diffractive deep neural networks , 2018, Science.

[24]  Pavel B. Bochev,et al.  Least-Squares Finite Element Methods , 2009, Applied mathematical sciences.

[25]  A. Quarteroni,et al.  Numerical Approximation of Partial Differential Equations , 2008 .

[26]  Max Gunzburger,et al.  Perspectives in flow control and optimization , 1987 .

[27]  George Em Karniadakis,et al.  Adaptive activation functions accelerate convergence in deep and physics-informed neural networks , 2019, J. Comput. Phys..

[28]  Stefan Ulbrich,et al.  Optimization with PDE Constraints , 2008, Mathematical modelling.

[29]  Stefan Wendl,et al.  Optimal Control of Partial Differential Equations , 2021, Applied Mathematical Sciences.

[30]  Chris H. Q. Ding,et al.  Multi-class protein fold recognition using support vector machines and neural networks , 2001, Bioinform..

[31]  Jimmy Ba,et al.  Adam: A Method for Stochastic Optimization , 2014, ICLR.

[32]  Wei-Hsin Liao,et al.  Modeling and control of magnetorheological fluid dampers using neural networks , 2005 .

[33]  Annalisa Quaini,et al.  Reduced basis methods for optimal control of advection-diffusion problems ∗ , 2007 .

[34]  J. Hesthaven,et al.  Certified Reduced Basis Methods for Parametrized Partial Differential Equations , 2015 .

[35]  Ivan P. Gavrilyuk,et al.  Lagrange multiplier approach to variational problems and applications , 2010, Math. Comput..

[36]  Luca Dedè,et al.  Optimal flow control for Navier–Stokes equations: drag minimization , 2007 .

[37]  Gianluigi Rozza,et al.  Model Reduction for Parametrized Optimal Control Problems in Environmental Marine Sciences and Engineering , 2017, SIAM J. Sci. Comput..

[38]  A. Kılıçman,et al.  Analytical Solutions of Boundary Values Problem of 2D and 3D Poisson and Biharmonic Equations by Homotopy Decomposition Method , 2013 .

[39]  Geoffrey E. Hinton,et al.  ImageNet classification with deep convolutional neural networks , 2012, Commun. ACM.

[40]  Fredi Tröltzsch,et al.  Optimal Control of the Stationary Navier--Stokes Equations with Mixed Control-State Constraints , 2007, SIAM J. Control. Optim..

[41]  W. Hartt,et al.  Data-driven physics-informed constitutive metamodeling of complex fluids: A multifidelity neural network (MFNN) framework , 2021 .

[42]  George Em Karniadakis,et al.  PPINN: Parareal Physics-Informed Neural Network for time-dependent PDEs , 2019, Computer Methods in Applied Mechanics and Engineering.

[43]  Guofei Pang,et al.  nPINNs: nonlocal Physics-Informed Neural Networks for a parametrized nonlocal universal Laplacian operator. Algorithms and Applications , 2020, J. Comput. Phys..

[44]  George Em Karniadakis,et al.  NSFnets (Navier-Stokes flow nets): Physics-informed neural networks for the incompressible Navier-Stokes equations , 2020, J. Comput. Phys..