Herein, we focus on using deep learning algorithms to find unknown parameters in differential equations. This kind of study is called “Scientific Machine Learing”, which aims to solve differential equations with limited input data and the properties of differential equations such as initial conditions and boundary conditions. Our goal is to find the water transfer coefficients in the porous mechanics, which cannot be easily obtained from experimental or clinical measurements. We modified the framework to the form of a multilayer perceptrons (MLP) as shown in figure 1. The results show that our framework is truly feasible.
