Deep learning-based molecular dynamics simulation reconfiguration of efficient heat energy transport of Gra/h-BN heterointerface

We investigate the reconfiguration scheme of efficient heat energy transport of the Gra/h-BN heterointerface by a hybrid enhanced machine learning method combining an integrated iterative method, automatic modeling, non-equilibrium molecular dynamics calculation, and convolutional neural network (CNN). The results show that the method identifies the optimal defect distribution in gra/h-BN from tens of millions of possible defect configurations, and the interfacial thermal conductivity (ITC) of the vdW gra/h-BN heterointerface in the optimal defect distribution is 97% higher than that in the original gra. Furthermore, the heat transfer transformation of the vdW gra/h-BN heterointerface with different defect distributions at room temperature can be observed efficiently. The dependency law between the defect distribution and the ITC is revealed by combining the CNN model of the ResNet network. The efficient heat energy transport design can promote the sustainable service life of materials and structures.