Search

In this article, we present one numerical approach to infer the model parameters and state variables of acoustic wave equations. The method we consider is based on the recently proposed method-the so-called hidden physics model. With placing Gaussian process (GP) prior on the state variables, the structure and model parameters of acoustic wave equations are encoded into the kernel function of a multioutput GP. The purpose of this article includes: 1) testing the applicability of hidden physics model to infer the velocity, density, and state variables of the acoustic wave equation, which is important for many applications in geophysics; 2) adapting the method to handle for both homogeneous and heterogeneous media that are of practical interest; and 3) exploring efficient sequential sampling methods to improve the sampling efficiency. We suggest that the expected-improvement-based sequential sampling method would be effective for most practical problems related to acoustic wave propagation. Besides, we demonstrate the performance of the proposed scheme via several benchmark problems. [ABSTRACT FROM AUTHOR]

A numerical approach based on the hidden physics model to estimate the model parameters of elastic wave equations with the sparse and noisy data is presented in this paper. Through discretizing the time derivatives of elastic wave equations and placing the priors of the state variables as Gaussian process, the model parameters and structure of elastic wave equations are encoded in the kernel function of a multi-output Gaussian process. In the learning stage, a parameter bound constraint condition is incorporated to enforce the physical bound of the model parameters. The numerical results from several benchmark problems, including homogeneous media, layer media, anisotropic media, and homogeneous model with an inclusion, demonstrate the feasibility and performance of the hidden physics model. • A hidden physics model is developed for parameter estimation of elastic wave equations. • A Physics-informed kernel is derived to encode the structure and model parameters of elastic wave equations. • An active learning strategy is suggested to improve the efficiency. • The performance of the hidden physics model is demonstrated by several benchmark problems. [ABSTRACT FROM AUTHOR]

Abstract We develop a numerical framework to implement the cumulative density function (CDF) method for obtaining the probability distribution of the system state described by a stochastic kinematic wave model. The approach relies on the computation of the fine-grained CDF equation of system state, as derived by the CDF method, via the method of characteristics. Due to its linearity, the fine-grained CDF equation is solved far more efficiently than the direct simulation of the kinematic wave model. Since the ensemble mean of the fine-grained CDF solutions is the probability distribution of the original system state, the proposed scheme requires less realizations than direct Monte Carlo simulations (MCS) of the kinematic model and thus converges relatively quickly. We verify the accuracy and effectiveness of our procedure via comparisons with direct MCS of several examples, including a particular kinematic wave system, the Saint-Venant equation. Highlights • Numerical framework of the cumulative density function (CDF) method for the stochastic kinematic wave model. • Computation of the fine-grained CDF equation from the CDF formulation via method of characteristics. • Numerically more efficient than direct simulation of the nonlinear kinematic wave model at each realization. • Less realizations are needed than direct simulation of the stochastic kinematic wave model for a desired accuracy. [ABSTRACT FROM AUTHOR]