Open AccessPFC model parameter calibration using uniform experimental design and a deep learning network
Author(s) -
Shijie Zhai,
Jiewei Zhan,
Yiding Ba,
Jianping Chen,
Yuchao Li,
Zhihai Li
Publication year - 2019
Publication title -
iop conference series. earth and environmental science
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.179
H-Index - 26
eISSN - 1755-1307
pISSN - 1755-1315
DOI - 10.1088/1755-1315/304/3/032062
Subject(s) - normalization (sociology) , calibration , macro , artificial neural network , computer science , sensitivity (control systems) , design of experiments , algorithm , artificial intelligence , control theory (sociology) , electronic engineering , mathematics , engineering , statistics , control (management) , sociology , anthropology , programming language
A macroscopic parameter calibration method for the particle flow code (PFC) using a uniform experimental design and deep learning network is proposed to explore the relationship between the micro parameters and macro response of rocks. First, a sensitivity analysis was used to identify which micro parameters influence the macro factors. A uniform experimental design was applied to achieve this goal in this study. Then, a neural network based on batch normalization was constructed. Through a large number of numerical simulation tests, the relationship between the input (micro parameters) and output (macro response) can be obtained by deep neural network (DNN). Finally, the DNN model was used to perform validation by using marble in Jinping. The average error for the best results was only 4.211%. The main advantage of this method is that it can achieve fast parameter calibration compared with traditional calibration methods, and the calibration results can be obtained in only a few seconds through the designed program.