A machine learning accelerated distributed task management system (Malac-Distmas) and its application in high-throughput CALPHAD computation aiming at efficient alloy design
Jianbao Gao, Jing Zhong, Guangchen Liu, Shenglan Yang, Bo Song, Lijun Zhang, Zuming Liu
Abstract
High-throughput calculations/simulations are the prerequisite for the efficient design of high-performance materials. In this paper, a machine learning accelerated distributed task management system (Malac-Distmas) was developed to realize the high-throughput calculations (HTCs) and storage of various data. The machine learning was embedded in Malac-Distmas to densify the output data, reduce the amount of calculation and achieve the acceleration of high-throughput calculations. Based on the Malac-Distmas coupling with CALPHAD software, HTCs of thermodynamics, kinetics, and thermophysical properties, including Gibbs free energy, phase diagram, Scheil-Gulliver solidification simulation, thermodynamic properties, thermophysical properties, diffusion simulation, and precipitation simulation, have been performed for demonstration. Furthermore, it is highly anticipated that the Malac-Distmas can also be coupled with any calculation/simulation software/code, which provides a console model to achieve different types of HTCs for efficient alloy design.