Physics-informed machine learning with conditional Karhunen-Loève expansions

• Published in: Journal of computational physics Vol. 426; no. C; p. 109904
• Main Authors: Tartakovsky, A.M., Barajas-Solano, D.A., He, Q.
• Format: Journal Article
• Language: English
• Published: Cambridge Elsevier Inc 01.02.2021; Elsevier Science Ltd; Elsevier
• Subjects: Computational physics; Conditional Karhunen-Loéve expansions; Covariance; Diffusion; Diffusion coefficient; Machine learning; Mathematical models; Model accuracy; Model inversion; Neural networks; Parameter estimation; Parameter identification; Partial differential equations; Physics; Conditional Karhunen-Loéve expansions; Parameter estimation; Model inversion; Machine learning
• ISSN: 0021-9991; 1090-2716
• DOI: 10.1016/j.jcp.2020.109904

•Conditional Karhunen-Loève expansion (cKLE) representation of unknown parameters and states in PDE models.•Solving a PDE inverse problem by computing weights in the cKLEs that minimize the PDE residuals.•For considered problems, PICKLE is more accurate than existing methods such as MAP and PINN. We present a new physics-informed machine learning approach for the inversion of partial differential equation (PDE) models with heterogeneous parameters. In our approach, the space-dependent partially observed parameters and states are approximated via Karhunen-Loève expansions (KLEs). Each of these KLEs is then conditioned on their corresponding measurements, resulting in low-dimensional models of the parameters and states that resolve observed data. Finally, the coefficients of the KLEs are estimated by minimizing the norm of the residual of the PDE model evaluated at a finite set of points in the computational domain, ensuring that the reconstructed parameters and states are consistent with both the observations and the PDE model to an arbitrary level of accuracy. In our approach, KLEs are constructed using the eigendecomposition of covariance models of spatial variability. For the model parameters, we employ a parameterized covariance model calibrated on parameter observations; for the model states, the covariance is estimated from a number of forward simulations of the PDE model corresponding to realizations of the parameters drawn from their KLE. We apply the proposed approach to identify heterogeneous diffusion coefficients in diffusion equations from sparse measurements of the diffusion coefficient and the solution of the diffusion equation. We find that the proposed approach compares favorably against state-of-the-art point estimates such as maximum a posteriori estimation and physics-informed neural networks.