Pinn-based dam-break simulation method and system

Abstract

The present application relates to the technical field of numerical simulation of hydraulic engineering, and particularly relates to a dam-break simulation method and system based on Pinn, which comprises the following steps: constructing a data set and training a first prediction model by numerically simulating different river channels, constructing an error matrix based on the model prediction results and fitting the correlation between time and space and error, using the relationship to construct an error-weighted quadratic training loss function to optimize and train the model, and finally screening the best dam-break simulation model to achieve accurate prediction of dam-break water flow and water depth in the target scene. The present application combines physical information neural network with numerical simulation data, physical constraints and error-weighted quadratic training, which not only gets rid of the grid dependence of traditional numerical methods, improves the calculation efficiency and scene adaptability, but also solves the problem of physical deviation caused by data scarcity, significantly improving the prediction accuracy and reliability under extreme dam-break working conditions.

Images