Method for constructing neural network force field model based on global optimization algorithm

Abstract

The invention discloses a method for constructing a neural network force field model based on a global optimization algorithm, and relates to the technical field of deep learning, and the method comprises the following steps: S1, collecting effective material simulation data according to a research target, carrying out the single-point calibration, classification and screening, and storing the data in a database; s2, according to a research target, selecting energy/energy and force from a database as an evaluation function of a neural network, and training to obtain a neural network model; s3, based on the neural network model obtained through training, performing data sampling through a genetic algorithm, and search of a global space is achieved; s4, performing validity evaluation on the sampling data, automatically submitting the valid sampling data to a supercomputing server for first principle calculation calibration, and merging the valid sampling data into a database to complete a round of iteration process; and S5, repeating the steps S1 to S4 to obtain the neural network force field model adaptive to the research task. According to the invention, automatic construction of the neural network force field model can be realized.

Description

technical field [0001] The invention relates to the technical field of deep learning, in particular to a method for constructing a neural network force field model based on a global optimization algorithm. Background technique [0002] In the field of material simulation, the traditional first-principles calculation software (VASP) can accurately describe the thermodynamic and dynamic properties of materials, but the calculation is time-consuming and requires extremely high supercomputing resources. The traditional molecular dynamics software, such as lammps and gromacs, although the calculation speed is acceptable, but there are problems of poor precision and too many empirical parameters. [0003] With the gradual recognition of neural networks in material simulation, the status of material big data in traditional material simulation research is increasing. The key to the construction of the neural network force field model lies in the selection of data sets, which makes ...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the needs and deficiencies of the current technological development, the present invention provides a method for constructing a neural network force field model based on a global optimization algorithm. This method makes full use of historical simulation data and takes into account the accuracy of the first-principle calculation software and the general molecular The speed of dynamics software can well solve the sampling problem in material simulation data

Images