基于神经网络的石墨烯弹性参量识别方法研究

Idenfication of elastic parameters method for graphene based on neural network

石墨烯的弹性参量是准确研究其力学性能的前提和基础.将神经网络的BP算法应用于石墨烯弹性模量和剪切弹性模量的预测,考虑石墨烯薄膜的长度、宽度、长宽比、手性、层数和温度6个影响因素,通过选取84组训练和检验样本,建立了石墨烯弹性参量的BP神经网络预测模型.将预测结果进行误差分析,其平均相对误差均小于3%,从而验证了该方法的适用性和可行性.将训练好的网络模型进行扩展计算,基于L25（5^6）正交表试验理论分析了石墨烯弹性参量对各影响因素的敏感性.为同类材料性能的预测提供了参考.

Elastic parameters of graphene is the premise and foundation for the research of its material mechanics performances. The BP neural network is used to predict the elastic modulus and shear modulus of graphene. Considering the length, width, aspect ratio, chiral, layers and temperature of graphene as the main influence factors and choosing 84 groups of data as training and forecasting sample, BP neural network model is established. The errors of forecasting results are analyzed, and the average relative errors are less than 3 %, which proves the applicability and feasibility of this method. Based on the calculation results, the sensitivity of influence factor to the graphene elastic parameter is analyzed by using L25（5^6） orthogonal table, which may provide a reference to the performance prediction of similar material.